Latest from Tom's Hardware in Amd-ryzen-7-1800x

MORE: All CPUs Content

Power Consumption

Our power measurements, which we recently migrated towards a more efficiency-focused approach, revealed some interesting tidbits. Particularly in regards to both Precision Boost Overdrive and manual overclocking.

Power consumption measurements are always a bit tricky. But as long as your 12V supply (EPS) readings, motherboard power supply sensor values, and voltage transformer losses plausibly coincide, everything is fine. Therefore, we're using pure package power to avoid possible influences from our motherboard. Results from the PWM controller are very reliable if you take them as averages over a few minutes.

AMD's Precision Boost Overdrive (PBO) is an adaptive overclocking approach that allows the processor to communicate with the platform to modulate performance based on the motherboard's power delivery subsystem and thermal dissipation capabilities. The processor monitors Package Power Tracking (PPT), which is total socket power, and the Thermal Design Current (TDC) variable, which is the motherboard's maximum available sustained current. Electrical Design Current (EDC) also indicates the maximum current possible from the VRMs during peak/transient conditions.

105W CPU Limits	PPT
AMD IPM	142W	95A	140A
MSI X570 Godlike	1000W	490A	630A

AMD enables two options for PBO: IPM is AMD's default PBO setting, which is activated if you leave the PBO setting to 'Auto' in the Godlike's UEFI. But you can select 'Enabled' to activate a profile that's dictated by the maximum limits of the motherboard's power delivery subsystem. These limits vary by motherboard and are defined by the vendor. We chose the latter to unlock the full potential of PBO. As you can see in the table, this enables the most robust power delivery options available, kicking the socket's maximum power delivery up to 1000W, and is designed to offer the best of increased multi-core boost clocks while retaining the high single-core boost clocks.

You can also further tune the power delivery options with an Auto OC (AOC) feature. This new feature grants you some control over the maximum attainable boost clocks by allowing you to add up to an extra 200MHz to the maximum boost clock, but it isn't guaranteed that the processor will reach those speeds at all times, or under all conditions. Instead, the processor will still respect the limits imposed by the motherboard maker.

Unfortunately, we've found that the PBO+AOC feature often comes at the expense of performance in single-threaded workloads even though it is billed as retaining, and even heightening, single-core boost clocks. From a performance and power consumption standpoint, it is more comparable to our all-core 4.3 GHz overclock. This might be an issue with early AGESA versions, but for now, the feature actually results in less performance in single-threaded workloads with MSI's X570 Godlike motherboard.

We chose to go with the limits of the motherboard, noted on the charts below as PBO, and the all-core 4.3 GHz overclock we outlined on the prior page. It's noteworthy that our all-core overclock comes at the expense of the single-core 4.5 GHz boost, which was a relatively rare occurrence during our testing in both stock and PBO trim.

We began with the non-AVX stress test in AIDA64 and found that the Ryzen 7 3800X draws very little power at stock settings, nearly matching the stock Ryzen 7 3700X's power draw. We activated Precision Boost Overdrive, and the processor's power draw only increased marginally to 95W. Our all-core overclock, however, sucked down 126W on average, but hit 163W peaks during the test.

You'll notice the Ryzen 7 3700X draws more power with PBO+AOC active than the 3800X's PBO configuration, but that's due to its all-core overclock-like nature. As such, our PBO+AOC results for the Ryzen 9 3900X and Ryzen 7 3700X are more comparable to the 3800X's all-core overclock.

The y-cruncher benchmark computes pi using a heavy multi-threaded AVX workload and also generates a performance measurement that we can use for efficiency metrics. We're also adding in HandBrake in x264 and x265 flavors. The latter uses a heavier distribution of AVX instructions than the former, but both transcoders are great for stressing the processor with a real-world workload.

We see muted increases in the 3800X's power consumption from the PBO option, and as a result, we also see minor performance improvements. Conversely, the all-core overclock consumes much more power but doesn't yield big performance benefits over the PBO settings. It appears that, for stock operation, AMD has tuned the processors right at the point of the voltage/frequency curve that provides the maximum frequency possible that can still provide great efficiency metrics. This PBO configuration also seems to retain some of those same characteristics, but that doesn't leave much headroom for explosive performance gains.

Plotting power consumption over our performance measurements highlights some of the gains we recorded with both overclocked settings. Here we can see the 3800X's 4.3 GHz all-core overclock regularly outperform the stock and PBO configurations, but its performance is very similar to the Ryzen 7 3700X with PBO. The overclocked 3800X's wins over the tuned 3700X weighed in at 1.91% for y-cruncher, 2.6% for x265, and 3.2% with x264 encoding. That shows that the Ryzen 7 3700X, while slower than the 3800X at stock settings by a larger margin, is competitive in threaded workloads after we tune both processors. Let's see how they compare in games.

MORE: Best CPUs

MORE: All CPUs Content

VRmark, 3DMark and AotS: Escalation

We tested the Threadripper 2920X in Game Mode for our game suite, and Creator Mode for our application tests.

VRMark, 3DMark

We aren't big fans of using synthetic benchmarks to measure performance, but 3DMark's DX11 and DX12 CPU tests provide useful insight into the amount of horsepower available to game engines.

The DX11 and DX12 CPU test results expose the full threaded heft of the Ryzen 3000 series processors. The Ryzen 9 3900X dethrones the overclocked Core i9-9900K in both tests. After overclocking, the Ryzen 7 3800X posts a virtual tie with Intel's flagship in the Fire Strike physics test, and the all-core overclock provides a comparable level of performance in the Time Spy test.

The VRMark test benefits heavily from per-core performance, and the Ryzen 3000 processors have made great strides compared to the first- and second-gen models. The Ryzen 3000 series processors have also become much more competitive in this benchmark, with the 3800X beating the stock Core i9-9900K and landing within a few FPS of the Core i7-9700K.

Ashes of the Singularity: Escalation

Ashes of the Singularity: Escalation is a computationally intense title that scales well with thread count, but clock speeds and per-core performance play a big role, as evidenced by the overclocked Core i9-9900K's spot at the top of the chart. The Ryzen 3000-series processors notch impressive gains, and we can see that the title responds well to our manual all-core overclock and the automatic PBO overclock. At stock settings, the Ryzen 7 3800X slots in right where we'd expect, and the Intel processors can only challenge if we apply an overclock.

MORE: Best CPUs

MORE: All CPUs Content

Civilization VI Graphics and AI, Dawn of War III

Civilization VI AI Test

The Ryzen 7 3800X handily beats the previous-gen models in this test of AI performance in a turn-based strategy game. This test is highly dependent on per-core performance. At stock settings, the 3800X matches the 3700X. As expected, the automated PBO overclocking feature yields similar results, but the processor doesn't fare as well with the all-core 4.3 GHz manual overclock because it loses the advantage of the single-core boost frequency of 4.5 GHz.

Civilization VI Graphics Test

The Ryzen 9 3900X is incredibly impressive in this benchmark, but the Ryzen 7 3800X requires tuning to climb the ranks. The 3800X ties the 3700X at stock settings, but cranking up the voltage gives it a slim ~1 FPS lead over the tuned 3700X.

The all-core overclock, which is manually dialed into a static 4.3 GHz, does provide a 3 FPS advantage in our 999th percentile frame rates over the auto-overclocked PBO configuration. We noticed this almost imperceptible difference between the two processors in several tests, indicating PBO's dynamic clock adjustments might have an impact on smoothness, although it's fair to say it is likely imperceptible.

We also see a 1 FPS difference in the less-stringent 99th percentile measurement, but the differences are slight. The advantage goes either way, either for or against manual tuning, throughout our suite.

Warhammer 40,000: Dawn of War III

The Warhammer 40,000 benchmark responds well to threading, but it's clear that Intel's clock speed advantage has an impact. The Ryzen 7 3800X gains 1 FPS over the 3700X at stock settings, and roughly 2 FPS after tuning. In either case, the Intel processors lead the chart by a large margin after tuning.

MORE: Best CPUs

MORE: All CPUs Content

Far Cry 5 and Final Fantasy XV

Far Cry 5

The Ryzen 7 1800X reminds us how far AMD has come since its first-gen products, and even the overclocked second-gen model lags behind the stock Ryzen 3000's by a decent margin.

After tuning, the 3800X ranks as the fastest AMD processor, and there is little difference between the manual and auto-overclocked configurations. Less than ~1 FPS separates the overclocked Ryzen 7 3800X and 3700X, which has become a theme. Intel asserts itself with faster performance across the board.

Final Fantasy XV

We run this test with the standard quality preset to sidestep the impact of a bug that causes the game engine to render off-screen objects. We see the slightest of gains for the manual overclock over the PBO configuration, but that's with expected variance for this benchmark, so consider this a tie.

Again, the Ryzen 7 3700X with PBO applied lands within striking distance of the overclocked 3800X.

MORE: Best CPUs

MORE: All CPUs Content

GTA: V and Hitman 2

Grand Theft Auto V

Grand Theft Auto V favors Intel architectures and, more generally, multi-core designs with high clock rates. The 3700X with its combination of PBO and an all-core overclock again offers roughly the same performance as the tuned 3800X.

Hitman

Hitman finds an oddity: The tuned 3700X actually ekes out a slim win over both overclocked 3800X configurations, but the differences are slight.

MORE: Best CPUs

MORE: All CPUs Content

Project Cars 2, The Division 2, and World of Tanks enCore

Project CARS 2

Although Project CARS 2 is purportedly optimized for threading, clock rates obviously affect this title's frame rates. Intel's per-core performance, which is a mixture of IPC and frequency, pays big dividends in this title.

A glance at the Ryzen 7 1800X reminds us of the giant strides AMD has made in just three generations.

The Division 2

The Ryzen 7 3700X continues to be a thorn in its sibling's side, once again offering nearly the same performance after we engage automatic overclocking.

World of Tanks enCore

World of Tanks closes out our game testing with much the same pattern we observed in other titles.

MORE: Best CPUs

MORE: All CPUs Content

Office, Web Browser, and Productivity

Web Browser

The Krakken suite evaluates JavaScript performance using several workloads, including audio, imaging, and cryptography. Like most browser-based benchmarks, single-threaded performance reigns supreme.

Browsers tend to be impacted more by the recent security mitigations than other types of applications, so Intel has taken a haircut in these benchmarks of fully-patched systems.

AMD's improved IPC comes into play as the processors post surprisingly good results that we wouldn't have expected with prior-gen models. The tuned Ryzen 7 3800X beats its more expensive counterpart the 3900X in several of these tests, but we see the impact of the 3700X's PBO+AOC configuration hamper its performance in lightly threaded tests. The same applies to the 3800X in some benchmarks even though we aren't engaging the AutoOC feature, meaning there appears to be some work left on the firmware side. Meanwhile, both processors provide great performance in stock trim. The 3800X notches complete victories over the -9700K in Krakken and Speedometer 2, but the -9700K takes the win in WebXPRT 3.

Microsoft Office

Microsoft's office suite makes its debut in our test suite via PCMark 10's new application test. This benchmark tests with real Microsoft Office applications, and we can see that the Ryzen 3000 series processors are very competitive in Excel, the Edge browser, and Word.

Core i7-9700K benefits massively from its overclock in some tests, but in stock form, it often trails the Ryzen 7 3800X. That trend is evident in the overall score.

Productivity

The application start-up metric measures load time snappiness in word processors, GIMP, and Web browsers under warm- and cold-start conditions. Other platform-level considerations affect this test as well, including the storage subsystem. The 3000-series processors occupy the middle of the ranks, but it will be interesting to revisit this benchmark with a PCIe 4.0 SSD in the future.

The photo editing benchmark measures performance with Futuremark's binaries using the ImageMagick library. Common photo processing workloads also tend to be parallelized, which plays well to Ryzen's multi-threaded heft. It takes a hefty overclock to boost the -9700K into contention with the 3800X. Again we can see some of the 3700X's performance sapped by the PBO+AOC configuration.

MORE: Best CPUs

MORE: All CPUs Content

Rendering, Encoding, Compression, Encryption

Rendering

It shouldn't surprise anyone to see the 12-core 24-thread 3900X bring a taste of HEDT performance to the multi-threaded rendering benchmarks. Ryzen 3800X performs well throughout the suite, particularly in the threaded Cinebench R15 test where it even beats the overclocked Core i9-9900K. And it only lags that overclocked processor slightly in the threaded Cinebench R20 benchmark. The 3800X also matches the 3900X in several of the single-core rendering tests, but we can see the 3800X's all-core 4.3 GHz result tumble in the rankings because the processor loses the benefit of its boost clocks. That doesn't have as much of an impact in the threaded workloads where it largely matches or exceeds its own PBO configuration. The tuned 3700X also posts solid scores in those threaded benchmarks, often slotting in slightly below the Ryzen 7 3800X.

With only eight threads, the -9700K is out of its element in the threaded tests, so the wins go Ryzen 7 3800X's way convincingly. The 3800X also beats the -9700K in the single-threaded Cinebench R20 test, but the -9700K excels in single-threaded work.

Encoding and Compression

Core i9-9900K traditionally leverages high frequencies to dominate the HandBrake x265 test, which relies heavily on AVX instructions, and the H.264 test. The Ryzen 7 3700X is impressive against its price-comparable competition, notching impressive leads over the Core i7-9700K and nearly matching the stock -9900K.

We can also see the vast improvement in Ryzen's AVX performance in the y-cruncher tests: We would never have imagined such a massive generational leap, particularly in single-threaded performance. The work AMD has done here is truly impressive and benefits a wide range of professional applications.

Our threaded compression and decompression metrics work directly from system memory, removing storage throughput from the equation. Ryzen 3900X's greatly improved memory performance, along with the generous helping of cores, plays a big role in its commanding lead, but the Ryzen 7 3800X is equally impressive given its single-die design. Here we can see it outperform Intel's -9700K and -9900K across the board.

LAME is the quintessential single-threaded application, and here the Core processors take the lead.

MORE: Best CPUs

MORE: All CPUs Content

Conclusion

We came away impressed with the Ryzen 7 3800X, but motherboard firmwares are still raw, leaving us wondering if there is more performance lurking underneath the heat spreader. Particularly with the Precision Boost Overdrive and Auto Overclock features. If history is any indication, we could see that situation improve as AMD and motherboard vendors work out the kinks. For now, these results reflect our experiences with a chip purchased at retail and with publicly-available firmwares, but your mileage might vary.

In the chart below, we plot gaming performance with both average frame rates and a geometric mean of the 99^th percentile frame times (a good indicator of smoothness). It's worth noting that AMD's previous-gen line-up is heavily discounted, so we’re departing from our standard practice of using official price lists. Instead, we’re using average pricing found online (temporary sales excluded). Volatility applies.

For gaming specifically, we did see slightly larger gains with the Ryzen 7 3700X via its PBO feature than we did with the Ryzen 7 3800X, allowing that processor to nearly match the more-expensive Ryzen 7 3800X in average framerates in all configurations. That isn't entirely unheard of – After overclocking, we saw the same trend of less expensive previous-gen models (non-X) offering similar performance to AMD's pricier X-series processors.

As we've seen, gaming remains an advantage for Intel, so if squeezing out every last frame is all you care about, Intel's processors are a good choice. Much of that performance advantage will be less noticeable when gaming at higher resolutions, or if you pair the processors with a lesser graphics card.

But, like most humans, if you do things other than gaming, the Ryzen 7 3800X offers a better mixture of performance in single- and multi-threaded applications. The 3800X offers twice the threads of the price-comparable Core i7-9700K, and it wields them to great effect in threaded workloads. As such, rendering and encoding remain a strong suit of the Ryzen chips, and AMD's improvements to AVX throughput have yielded impressive results.

AMD's platform also supports the PCIe 4.0 interface that provides twice the throughput of the previous-gen standard. The extra throughput doesn't equate to improved performance in gaming, but it does speed up everyday tasks like file transfers and will unlock more performance in storage-bound applications. That's an advantage that Intel can't match.

Unfortunately, the faster interface does result in higher-priced X570 motherboards, but AMD's continued support for the X470 motherboard ecosystem could help blunt the blow. Motherboard partners continue to offer X470 motherboards, and they are cheap and plentiful. You'll lose access to the PCIe 4.0 interface in exchange for lower pricing, but you'll also have a compelling upgrade path in the future.

AMD's Ryzen 7 3800X comes with a bundled cooler while competing Intel processors come with a cardboard box. That's an added value that you should take into consideration, but you might need a beefier cooling solution if you plan on pushing the limits with overclocking.

For now, we don't see massive gains in performance for the 3800X from the automated overclocking features, and AMD says we shouldn't expect too much manual overclocking headroom. Instead, you should look to tuning the Infinity Fabric and memory, which both raise in lockstep to a sweet spot of DDR4-3600. AMD also cites using PBO to boost performance, but in our test environment, that doesn't have a dramatic impact on performance. Given that we overclocked our memory in tandem with activating PBO, it's possible some of those slim gains actually stem from increased memory performance.

That means running the processor cores at stock settings paired with overclocked memory could offer roughly equivalent performance gains in some workloads. We're impatiently awaiting new firmware revisions to gauge how they react, but we have tested a multitude of PBO configurations with the Ryzen 7 3800X, of which there are many potential options, with varying levels of success. None seem to confer large speedups for gaming, though we did see some better uplift in traditional applications.

In either case, AMD is wringing a surprising amount of performance from its comparably lower clock speeds, largely due to the big generational uplift in IPC. In our tests, the Ryzen 3000 series is also more power efficient in terms of the amount of energy required to complete a task, which also equates to lowered thermal output.

Value seekers who aren't afraid to press the Precision Boost Overdrive button and have sufficient cooling should look to the Ryzen 7 3700X for roughly equivalent performance to the 3800X, particularly if gaming factors heavily into the buying decision. That could save you $70, reinforcing our decision to give the Ryzen 7 3700X an Editor's Choice award.

Out of the box, the Ryzen 7 3800X is a better all-arounder than the Core i7-9700K and offers incrementally higher performance than its downstream counterpart. The bundled cooler reduces platform costs, and a wide array of motherboards with both X470 and X570 chipsets offers plenty of choices for builders.

Photo Credits: Tom's Hardware

MORE: Best CPUs

MORE: CPU Benchmark Hierarchy

MORE: All CPUs Content

AMD Ryzen 7 2700 Review: The Non-X Factor

palcorn@outlook.com (Paul Alcorn) — Wed, 16 May 2018 13:00:00 +0000

Go X Or Go Home

AMD's Ryzen 7 2700 includes the same eight cores as its more expensive Ryzen 7 2700X flagship, plus simultaneous multi-threading technology that allows each core to work on two software threads at the same time. But its clock rates are trimmed back to create a $30-cheaper model sporting a little less performance. Even though the 2700 loses its enthusiast-targeted X modifier, AMD still arms the chip with an unlocked ratio multiplier for flexible overclocking. And this less expensive CPU should hit nearly the same frequencies as the Ryzen 7 2700X we like so much.

Last generation, plenty of overclocking headroom and lower prices earned AMD's non-X Ryzen SKUs praise up and down the stack. Much of that was based on the company's bundled coolers, though. For example, the then-flagship Ryzen 7 1800X launched at $500 with no thermal solution at all. Meanwhile, the Ryzen 7 1700 sold for $330 with a cooler in the box. Now, AMD bundles a heat sink and fan with all of its new Ryzen chips. Moreover, it only sells 2700 at a $30 discount. Worse, although it's possible to match Ryzen 7 2700X's performance after a bit of overclocking, you need an aftermarket cooler to get there. AMD's freebie won't cut it. This time around, there's not much reason to step down a tier.

Ryzen 7 2700

With its 2000-series Ryzen processors, AMD was challenged to deliver more than the incremental improvements we've been seeing from Intel lately. To begin, the company swapped out its 14nm manufacturing process with a 12nm node, enabling higher clock rates at the same power consumption levels. AMD also optimized the Zen architecture by adding more sophisticated multi-core boost algorithms and lowering cache and memory latencies. Together, those changes enable speed-ups in pretty much every type of workload we test with, and they're all baked in to Ryzen 7 2700.

	AMD Ryzen 7 2700X	AMD Ryzen 7 1800X	AMD Ryzen 7 2700	Ryzen 7 1700	AMD Ryzen 5 1600X	AMD Ryzen 5 2600X	AMD Ryzen 5 2600	Intel Core i7-8700K	Intel Core i5-8600K	Intel Core i5-8400
MSRP	$329	$349	$299	$299	$219	$229	$199	$359	$257	$182
Cores/Threads	8/16	8/16	8/16	8/16	6/12	6/12	6/12	6/12	6/6	6/6
TDP	105W	95W	65W	65W	95W	95W	65W	95W	95W	65W
Base Freq. (GHz)	3.7	3.6	3.2	3.0	3.6	3.6	3.4	3.7	3.6	2.8
Precision Boost Freq. (GHz)	4.3	4.1	4.1	3.8	4.0	4.2	3.9	4.7	4.3	4.0
Cache (L3)	16MB	16MB	16MB	16MB	16MB	16MB	16MB	12MB	9MB	9MB
Unlocked Multiplier	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No
Cooler	105W Wraith Prism (LED)	-	95W Wraith Spire (LED)	95W Wraith Spire	-	95W Wraith Spire	65W Wraith Stealth	-	-	Intel

AMD rates its Ryzen 7 2700 at a much lower 65W thermal design power than its 105W Ryzen 7 2700X. That's partly why the 2700's base frequency is a much more conservative 3.2 GHz, while its maximum Precision Boost clock rate tops out at 4.1 GHz. You'd think, then, that the bundled 95W Wraith Spire thermal solution with LED lighting would be beefy enough for aggressive overclocking. It's not, though. If you're really after Ryzen's peak potential, purchase a more capable aftermarket heat sink/fan combination or closed-loop liquid cooler.

All 2000-series Ryzen CPUs are compatible with motherboards sporting new X470 or older 300-series chipsets. You can even overclock the new processors on value-oriented B-series platforms. While lower-cost 400-series chipsets aren't available yet, we're counting on them to offer a more affordable option for enthusiasts looking to tune 2000-series Ryzen CPUs.

The Ryzen 7 2700 supports up to DDR4-2933 memory, just like the 2700X. Just be aware that you'll only get those data rates with single-rank modules installed in a maximum of two slots. Even then, it takes a motherboard with six PCB layers to operate at 2933 MT/s stably. AMD uses Indium solder between its Ryzen 7 die and heat spreader, improving thermal transfer performance compared to Intel CPUs reliant on paste instead.

Like all 2000-series models, the Ryzen 7 2700 comes with StorMI Technology, which is a software-based tiering solution that blends the low price and high capacity of hard drives with the speed of an SSD, 3D XPoint, or even up to 2GB of RAM.

Precision Boost 2 And XFR2

AMD's previous-gen Ryzen processors include Precision Boost functionality, which is similar to Intel's Turbo Boost technology. They also sport a feature called eXtended Frequency Range (XFR), which enables higher clock rates when it's determined that your cooling solution has thermal headroom to spare.

The new Precision Boost 2 (PB2) and XFR2 algorithms improve performance in threaded workloads by raising the frequency of any number of cores. AMD doesn't share a list of specific multi-core PB2 and XFR2 bins because the opportunistic algorithms accelerate to different clock rates based on temperature, current, and load. However, we collected our measurements on a motherboard with solid voltage regulation circuitry and a good cooler, two requirements for optimal frequencies.

The Ryzen 7 2700 offers an impressive 4.1 GHz clock rate benchmarked in a single-threaded workload. Try as we might, though, the CPU wouldn't exhibit the same graceful downward frequency slope as AMD's 2700X as we increased the test's thread count. No doubt, Ryzen 7 2700 is a step up compared to the previous-gen Ryzen 7 1700, but its frequency drops further and faster than the 2700X.

MORE: Best CPUs

MORE: All CPUs Content

Overclocking & Test Setup

Overclocking

Like the Ryzen 7 2700X, we overclocked AMD's 2700 to 4.2 GHz. While the lower-end chip did need a bit more voltage to get there, our 1.4V Vcore, 1.2V SoC, and Level four Loadline Calibration (LLC) settings are still within the safe zone for AMD's processors. They just make the CPU operate a little warmer than it would otherwise. Top the 2700 with a capable closed-loop watercooler and thermals won't be an issue. If you're stuck with AMD's bundled heat sink and fan, though, expect Ryzen 7 2700 to fall short of its best possible overclock. Our sample topped out at 4.05 GHz as it exceeded 95°C.

We did match the 2700X's overclocked memory settings (DDR4-3466 at 14-14-14-34 timings) with minimal effort. That's particularly encouraging since memory tuning dramatically improves Ryzen's gaming performance.

Precision Boost Overdrive

AMD isn't giving us much detail about its Precision Boost Overdrive feature, though we know it increases maximum boost voltage and boost duration. However, we also know that Precision Boost Overdrive is an AMD-sanctioned feature. Because this is a standard capability for Ryzen 2000-series processors, we leave it enabled. Conversely, we disable the Multi-Core Enhancement BIOS option found in many Intel-based motherboards because it overclocks beyond the company's specifications.

MSI X470 Gaming M7 AC

Our MSI X470 Gaming M7 AC has a PCI Express 3.0 slot with a x16 link, a slot with a x8 connection, and another PCI Express 2.0 slot with a four-lane link for graphics cards. Its four RAM slots support DDR4-2933 and scale quite a bit higher through overclocking.

The motherboard also provides two M.2 slots with PCIe connectivity. The I/O panel has a USB 3.1 Type C connector. The USB 3.1 and USB 3.1 Gen 2 ports support fast charging for smartphones and tablets. If RGB is your thing, MSI has you covered. The integrated RGB Mystic lighting allows customizable effects with several software-controlled zones.

Comparison Products

Intel Core i5-8400

Intel Core i5-8600K

Intel Core i7-8700K

Test Systems

Test System & Configuration
Hardware	AMD Socket AM4 (400-Series)AMD Ryzen 7 2700, Ryzen 7 2700X, Ryzen 5 2600XMSI X470 Gaming M7 AC2x 8GB G.Skill FlareX DDR4-3200 @ DDR4-2933, DDR4-3466Intel LGA 1151 (Z370):Intel Core i7-8700K, i5-8600K, Core i5-8400MSI Z370 Gaming Pro Carbon AC2x 8GB G.Skill FlareX DDR4-3200 @ DDR4-2400, DDR4-2667, DDR4-3466AMD Socket AM4 (300-Series)AMD Ryzen 7 1800X, 1700, Ryzen 5 1600XMSI X370 Xpower Gaming Titanium2x 8GB G.Skill FlareX DDR4-3200 @ DDR4-2667, DDR4-3200All EVGA GeForce GTX 1080 FE 1TB Samsung PM863 SilverStone ST1500-TI, 1500W Windows 10 Creators Update Version 1703 - All Spectre and Meltdown mitigations
Cooling	U.S.Corsair H115i

MORE: Best CPUs

MORE: All CPUs Content

VRMark, 3DMark & AotS: Escalation

VRMark & 3DMark

While synthetic benchmarks scale well with increased host computing resources, those gains don't always translate to real-world gaming performance. Rather, these benchmarks give us a solid measure of theoretical horsepower available to game engines.

All of our test processors easily passed the 109 FPS threshold that UL sets as the minimum for a smooth virtual reality experience.

The stock Ryzen 7 2700 nearly matched AMD's previous-gen 1800X. However, it also trailed the 2700X by a significant margin. A bit of tuning goes a long way though, and the overclocked 2700 came close to tying our overclocked Ryzen 7 2700X. Also interesting, the stock Ryzen 5 2600X beat Ryzen 7 2700 out of the box.

Ashes of the Singularity: Escalation

The 2700's stock performance was impressive, highlighting the advantage of lower memory latency and enhanced multi-core turbo algorithms. Overclocked, Ryzen 7 2700 exceeded the performance of a stock 2700X.

Ashes of the Singularity: Escalation scales well with additional threads, so six-core models without Intel's Hyper-Threading Technology like the Core i5-8400 and -8600K languish at the bottom of our chart. Overclocking propelled the Core i5-8600K into contention with a tuned 6C/12T Ryzen 5 2600X.

MORE: Best CPUs

MORE: All CPUs Content

Civilization VI Graphics & AI, Dawn of War III

Civilization VI AI Test

Civilization's AI test measures performance in a turn-based strategy game and tends to favor per-core performance. Therefore, its comes as no surprise that a stock Ryzen 7 2700 delivered the turn time we expected, given fairly conservative clock rates.

An unlocked ratio multiplier is a mighty equalizer, though. After overclocking, the 2700 outpaced Ryzen 7 2700X operating at its default clock rates and nearly matched the Core i7-8700K. Of course, tuning the Intel processors widened their lead considerably.

Civilization VI Graphics Test

Ryzen 7 2700 beat the Ryzen 7 1800X in stock trim, though tuning improved its outcome significantly.

Clearly, Intel's Core i5-8400 is a formidable competitor. It achieved almost the same performance as an overclocked Ryzen 7 2700. Too bad, then, that Intel locks the chip's multiplier, preventing enthusiasts from making it any faster. Stepping up to an overclockable Core i5-8600K yielded a big performance boost, though that also compels you to spend big on a Z-series motherboard and aftermarket cooler.

A tuned Ryzen 7 2700 trailed the 2700X at a similar 4.2 GHz, if only slightly.

Warhammer 40,000: Dawn of War III

Warhammer 40,000: Dawn of War III responds best to the high clock rates of Intel's tuned CPUs. This time around, however, an overclocked Ryzen 7 2700 nearly matched the stock Core i7-8700K.

MORE: Best CPUs

MORE: All CPUs Content

Far Cry Primal, GTA: V & Hitman

Far Cry Primal

Our Far Cry Primal benchmark used to swing entirely in Intel's favor. But AMD's 2000-series Ryzen processors are much more competitive than their predecessors. Again, a stock Ryzen 7 2700 easily outpaced AMD's Ryzen 7 1800X and 1700. And this time around, a bit of overclocking even pushed the 2700 beyond AMD's flagship at 4.2 GHz.

Grand Theft Auto V

Grand Theft Auto V favors Intel architectures and, more generally, multi-core designs with high clock rates. The tuned Ryzen processors, which effectively tied each other, go a long way to improve AMD's competitive position.

Hitman

Hitman's Game of the Year update imposed a 90 FPS cap, so this title no longer scales well with high-end PCs. Unfortunately, some popular AAA titles employ similar frame rate limits, so we leave this result in our suite to show that not all titles respond to faster components.

Overall, the results are unsurprising given IO Interactive's newly-imposed performance ceiling. However, the Core i5's results stand out for their low 99th percentile frame rates.

MORE: Best CPUs

MORE: All CPUs Content

Shadow Of War & Project CARS 2

Middle-earth: Shadow Of War

Middle-earth: Shadow of War doesn't scale as well as some of our other benchmarks, and it certainly isn't as sensitive to IPC throughput and clock rate as Shadow of Mordor.

All of the processors we benchmarked offered a similar gaming experience, though the Ryzen 7 1700 obviously lagged the group due to its lower frequency and lack of multi-core turbo.

Project CARS 2

While the tuned Ryzen 7 2700 effectively tied Core i7-8700K, Intel's Core i5-8400 remained a formidable challenger. Overclocking helped the Core CPUs to some extent, though as we saw from Intel's Core i5-8600K, these numbers are indicative of a graphics bottleneck.

MORE: Best CPUs

MORE: All CPUs Content

Office & Productivity

Adobe Creative Cloud

The Ryzen 7 2700 lagged behind AMD's 2700X flagship at stock settings in our Creative Cloud suite, particularly during the benchmark's lightly-threaded tests. Overclocking helped achieve near-parity, though.

Intel's processors benefited from increased overclocking headroom and the ability to get more done per clock cycle, easily outpacing the rest of our CPUs.

Web Browser

The Krakken suite tests JavaScript performance using several workloads, including audio, imaging, and cryptography.

AMD's processors typically lag Intel's in Web browser benchmarks due to their lower per-core performance. So it comes as little surprise that a stock Ryzen 7 2700 suffered in these metrics due to its low frequency range. Of course, overclocking helped rectify this.

Productivity

AMD's Ryzen 7 2700 performed admirably right out of the box. While overclocking yielded a big boost, the 2700 still fell behind the pricier Ryzen 7 2700X.

Our video conferencing suite measures performance in single- and multi-user applications that utilize the Windows Media Foundation for playback and encoding. It also performs facial detection to model real-world usage. This workload benefits from parallelism, so the 8C/16T Ryzen models naturally enjoyed the largest advantage from their architecture. Although an overclocked Ryzen 7 2700 slipped past AMD's 2700X, its lead fell within the margin of error.

MORE: Best CPUs

MORE: All CPUs Content

Rendering, Encoding & Compression

Rendering

Ryzen 7 CPUs find plenty of fans among the rendering crowd. AMD's Ryzen 7 2700 represents a solid step beyond the company's 1000-series processors in our multi-threaded tests. In fact, it often provided a similar performance level as the previous-generation 1800X flagship in our tests.

The Intel processors demonstrated their per-core advantage in our lightly threaded tests. Meanwhile, a stock Ryzen 7 2700X proved faster than the same chip overclocked in many of the same benchmarks, since its 4.3 GHz dual-core boost frequency is higher than our 4.2 GHz all-core overclock.

Encoding & Compression

LAME is a quintessential single-threaded workload that typically illustrates Intel's per-clock advantage. This time around, though, AMD's Ryzen 7 2700X and Ryzen 5 2600X landed on top of the mighty Core i7-8700K in stock form.

Our threaded compression and decompression tests work directly from system memory, removing our SSDs from the equation. The Ryzen 7 2700 put its eight cores to good use, beating most of its competition.

There's a larger delta between Intel and AMD processors in our HandBrake x265 test compared to the x264 benchmark, given the former's heavier reliance on AVX instructions. AMD looks much better in both metrics than it used to, though. The improved multi-core Precision Boost 2 frequencies help level the playing field.

MORE: Best CPUs

MORE: All CPUs Content

Final Analysis

AMD's Ryzen 7 2700 boasts many of the same features as the flagship 2700X, including an eight-core architecture and 16MB of L3 cache. But it loses quite a bit of steam in our benchmarks due to its lower clock rates. Overclockers can get comparable performance out of both CPUs, but they need to replace the stock heat sink and fan with higher-end aftermarket cooling first. This sullies any value advantage the 2700 might have enjoyed.

How does a frequency disadvantage affect the 2700's viability in a gaming PC versus AMD's Ryzen 7 2700X? The charts below plot performance using average frame rates and a geometric mean of the 99^th percentile frame times (a good indicator of smoothness), which we then convert into a FPS measurement. We also present price-to-performance charts that get split up to include CPUs-only and extra platform costs. For the models that don't come with a bundled cooler, we add an extra $25 for a basic heat sink. We also add $20 if overclocking requires a more expensive motherboard (as is the case for Z370).

When it comes to gaming, an overclocked Ryzen 7 2700 provided a statistically insignificant ~1 FPS advantage over the 2700X in 99th percentile frame rates. In essence, both tuned processors offer a very similar experience through our suite. It's unfortunate, then, that you need to buy an aftermarket cooler in order to overclock Ryzen 7 2700 to its potential. So much for the $30 savings versus Ryzen 7 2700X.

That means you get more value from AMD's flagship than the stepped-down Ryzen 7 2700. In fact, at stock settings, even a Core i5-8400 is roughly equivalent to the 2700. And it costs $110 less. And you can drop it into affordable motherboards. Intel's stock cooler is fine, too. You get the picture.

Similar trends surfaced in our desktop productivity tests. Ryzen 7 2700's frequency deficit resulted in notably less performance than a stock Ryzen 7 2700X across the board. Although Ryzen 7 2700 is faster than Intel's Core i5-8400 and -8600K in threaded workloads thanks to its eight-core design, AMD's own Ryzen 7 2700X is more compelling if you're interested in those types of applications.

If you tune the memory, the Ryzen 7 2700X doesn't gain much from overclocking the cores. That means you can pair the 2700X and its stock cooler with a capable memory kit and get impressive performance. In contrast, you'll have to invest in a more capable cooler to achieve the same level of performance with the Ryzen 7 2700. That largely negates the 2700's scant $30 price advantage.

Gone are the days of AMD selling its X-series Ryzen CPUs at higher prices and without thermal solutions. Newer models like the Ryzen 7 2700X only cost a bit more than the 2700, plus they also include a heat sink and fan. We'd rather spend the extra $30. Of course, gamers might want to go another direction. Intel's Core i5-8400 costs less and is every bit as quick in our favorite titles. And then there's the Ryzen 5 2600X, which is surprisingly fast across a broader range of workloads and much more affordable.

MORE: Best CPUs

MORE: All CPUs Content

AMD Ryzen 5 2600X Review: Spectre Patches Weigh In

palcorn@outlook.com (Paul Alcorn) — Fri, 27 Apr 2018 13:00:00 +0000

Ryzen To The Mainstream

AMD's 2000-series Ryzen CPUs are already available, challenging the Coffee Lake-based Core line-up from Intel. As we found in our Ryzen 7 2700X review, a host of improvements made possible by 12nm manufacturing, such as higher frequencies and Precision Boost 2, add more performance in threaded apps. Meanwhile, lower system memory and cache latencies augment AMD's showing in lightly-threaded apps like games. Unlocked multipliers, backward compatibility with older Socket AM4 motherboards, a beefy bundled cooler, and a $330 price tag combine to leave us impressed. The Ryzen 7 2700X offers a great alternative to Intel's Core i7-8700K, which costs more, doesn't come with a thermal solution, and drops into more expensive motherboards (at least if you want to overclock).

Similarly, Ryzen 5 2600X targets Intel's enthusiast-oriented Core i5-8600K, leveraging similar advancements and a more attractive $230 price tag. As we'll see, it's even faster than the first-gen flagship Ryzen 7 1800X in many workloads.

But First, Spectre Variant 2

Unfortunately, due to a lack of communication from AMD, we weren't told that the company had rolled its Spectre Variant 2 patch into shipping X470 platforms. As a result, our Ryzen 7 2700X launch day coverage didn't include Intel CPUs tested with their corresponding patches. Today's review does, however, feature results generated on Intel-based systems with the latest Spectre microcode updates.

Ryzen 5 2600X

Ryzen 2000-series processors, otherwise known by their Pinnacle Ridge code name, are based on the same basic Zen core design as previous-gen models (though AMD now uses Zen+ nomenclature to reference the architecture's various improvements). The CPUs still utilize a dual-CCX configuration, tied together with Infinity Fabric, yielding eight physical cores. The flagship Ryzen 7 2700X comes with all eight of its cores active. For Ryzen 5 2600X, AMD turns two off, creating a six-core, 12-thread configuration with an unlocked ratio multiplier.

As mentioned, Ryzen 5 2600X sells for $230, replacing the $220 Ryzen 5 1600X. It slots into the gap between Core i5-8600K and the Core i5-8400, forcing the chip to contend with Intel's recently-announced Core i5-8600. While we don't have that model in our lab yet, we do have the two nearest Coffee Lake-based competitors in today's benchmark charts.

What do you get, performance-wise, for the extra $10? Ryzen 5 2600X sports the same 3.6 GHz base clock rate and a slightly higher 4.2 GHz Precision Boost 2 frequency (+200 MHz) than 1600X. That might seem minor, but as our benchmarks show, the gains are quite pronounced in threaded workloads. Like its predecessor, the 2600X also features 16MB of L3 cache and a 95W TDP.

	AMD Ryzen 7 2700X	AMD Ryzen 7 1800X	AMD Ryzen 7 2700	AMD Ryzen 5 1600X	Ryzen 5 1600	AMD Ryzen 5 2600X	AMD Ryzen 5 2600	Intel Core i5-8600K	Intel Core i5-8600	Intel Core i5-8400
MSRP	$329	$349	$299	$219	$189	$229	$199	$257	$224	$182
Cores/Threads	8/16	8/16	8/16	6/12	6/12	6/12	6/12	6/6	6/6	6/6
TDP	105W	95W	65W	95W	65W	95W	65W	95W	65W	65W
Base Freq. (GHz)	3.7	3.6	3.2	3.6	3.2	3.6	3.4	3.6	3.1	2.8
Precision Boost Freq. (GHz)	4.3	4.1	4.1	4.0	3.6	4.2	3.9	4.3	4.3	4.0
Cache (L3)	16MB	16MB	16MB	16MB	16MB	16MB	16MB	9MB	9MB	9MB
Unlocked Multiplier	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	No
Cooler	105W Wraith Prism (LED)	-	95W Wraith Spire (LED)	-	95W Wraith Spire	95W Wraith Spire	65W Wraith Stealth	-	Intel	Intel

Although AMD didn't include thermal solutions with its original Ryzen X-series processors, the company does bundle coolers with its pricier models now. On one hand, it's nice that the 95W Wraith Spire cooler neatly matches the 2600X's thermal design power. On the other, we're not expecting much overclocking headroom from the combination.

Ryzen 5 2600X can drop into either new X470 or older 300-series motherboards. As usual, AMD allows you to overclock on value-minded B-series boards, too. And even though 400-series B-models aren't available yet, they'll undoubtedly offer a lower-priced alternative for overclocking.

Officially, the Ryzen 5 2600X supports up to DDR4-2933 memory, just like Ryzen 7 2700X. This trumps Coffee Lake's Intel-specified DDR4-2666 ceiling (with a few caveats). AMD also sticks with Indium solder between Ryzen 5's die and heat spreader, improving thermal transfer performance. And as we mentioned in our Ryzen 7 2700X review, these new CPUs also include StorMI Technology, which is a software-based tiering solution that blends the low price and high capacity of a hard drive with the speed of an SSD, 3D XPoint (including Intel's Optane parts), or even up to 2GB of RAM.

Precision Boost 2 and XFR2

In a nutshell, AMD is leveraging GlobalFoundries' 12nm process to enhance its design, rather than shrink it. The enhancements offer higher performance or lower power consumption at any given frequency, giving AMD headroom for other improvements.

The company's previous-gen Ryzen processors have Precision Boost, which is similar to Intel's Turbo Boost technology, and eXtended Frequency Range (XFR), capable of delivering a frequency uplift when your cooling solution has thermal headroom to spare.

The new Precision Boost 2 (PB2) and XFR2 algorithms improve performance in threaded workloads by raising the frequency of any number of cores. AMD doesn't share a list of specific multi-core Precision Boost 2 and XFR2 bins because the opportunistic algorithms accelerate to different clock rates based on temperature, current, and load.

AMD gave us a graph of the PB2 frequencies for Ryzen 7 2700X, but we followed up with our own measurements to compare the current and previous-gen Ryzen 5 models. As you can see, Ryzen 5 2600X offers more robust multi-core frequencies than its predecessor, and our Ryzen 7 2700X measurements largely mirror AMD's. We tested both CPUs with AMD's Precision Boost Overdrive active. The Ryzen 7 2700X does have a higher TDP rating that some older motherboards may struggle with, so PB2 performance will vary based upon the power delivery subsystem.

MORE: Best CPUs

MORE: All CPUs Content

AMD Ryzen 7 2700X

AMD Ryzen 5 2600X

Patching Up, Overclocking & Test Setup

Patching Up

The Spectre and Meltdown security vulnerabilities impose overhead that affect performance, but it's the Spectre Variant 2 patches that incur the largest penalties. We've been waiting for AMD and Intel to release their respective updates, leveling the playing field. AMD gave no indication in its press material that the X470 motherboards used in our Ryzen 7 2700X review already had Spectre 2 mitigations built-in. But hours before launch, we learned that the patches were present. At that point, we didn't have enough time to test with Intel's corresponding microcode.

In our 2700X review, the test rigs included Meltdown And Spectre Variant 1 mitigations. Spectre Variant 2 requires both motherboard firmware/microcode and operating system patches, though. We had already installed the operating system updates for Variant 2 on our Intel-based platforms, so we only lacked the fourth and final piece: new microcode.

There are two options for applying this microcode. One is a Windows KB that allows the operating system to load microcode during boot-up. The other is a motherboard firmware update. We used the Windows KB to install patches on our Z270-based platform, providing a true measure of pre- and post-patch performance.

Faster processors suffer less from the Spectre Variant 2 updates. This creates a conundrum for us and our Z370-based platform. In previous reviews, we noticed that Intel's Core i7-8700 was consistently faster than the more expensive Core i7-8700K on MSI and Gigabyte motherboards. We disclosed this during our launch coverage. The issue becomes relevant today because MSI's latest firmware update, which also includes the Spectre Variant 2 microcode, fixes most of the performance disparities we observed. As we suspected, the problem seems attributable to the Core i7-8700K. It's now faster in several games, performing the way we originally expected. So, on this motherboard, we're using new firmware instead of the Windows-based patch since it fixes our Core i7-8700K issues. That means our Coffee Lake-based CPUs don't correctly reflect pre- and post-patch performance. Instead, allow those results to serve as a general indication of competitive deltas.

In any case, with the exception of our AMD X370-based motherboards, all of the platforms in today's review are fully patched. The company hasn't given us a time frame for securing its previous-gen chipsets, but you can bet that we'll follow up with benchmark results once a Spectre Variant 2 patch becomes available.

Overclocking

We ran our gaming and application tests in the U.S. lab, while power/thermal measurements were collected in our German lab.

In the U.S. lab, we paired our Ryzen 5 2600X with Corsair's H115i cooler for overclocking. This allowed us to maintain a 4.2 GHz all-core frequency at 1.3875V Vcore, 1.2V SoC voltage, and default Load Line Calibration settings. These are the same settings used on the Ryzen 7 2700X, albeit with a slight 0.009V Vcore boost to ensure stability.

First-gen Ryzen processors don't have much memory overclocking headroom, so we're still testing tuned X370 platforms at DDR4-3200. The X470 chipset is remarkably stable at higher data rates with both 2000-series Ryzen CPUs installed. So, we settled on DDR4-3466 with 14-14-14-34 timings. We also ran our overclocked Intel processors at DDR4-3466.

Precision Boost Overdrive

AMD hasn't shared much information on this pending feature, which increases the maximum boost voltage and boost duration. We attempted to disable Precision Boost Overdrive as we tested for our Ryzen 7 2700X review, but didn't observe a performance difference one way or the other. Now we know the feature wasn't toggling correctly due to an issue with the board. Instead, it remained enabled throughout our benchmarking.

Precision Boost Overdrive is an AMD-sanctioned feature, unlike the multi-core enhancements you often find in Intel-based motherboards. Because this is a standard capability for Ryzen 2000-series processors, we leave it enabled.

MSI X470 Gaming M7 AC

The motherboard also provides two M.2 slots with PCIe connectivity. The I/O panel has a USB 3.1 Type C connector. The USB 3.1 and USB 3.1 Gen 2 support fast charging for smartphones and tablets. If RGB is your thing, MSI has you covered. The integrated RGB Mystic lighting allows customizable effects with several software-controlled zones.

Comparison Products

Intel Core i5-8400

Intel Core i5-8600K

Intel Core i7-7700K

Test Systems

Test System & Configuration
Hardware	Germany AMD Socket AM4 (400-Series)AMD Ryzen 7 2700X, Ryzen 5 2600XMSI X470 Gaming M7 AC2x 8GB G.Skill FlareX DDR4-3200 @ DDR4-2933, DDR4-3466Intel LGA 1151 (Z370):Intel Core i5-8600K, i5-8600K, Core i5-8400MSI Z370 Gaming Pro Carbon AC2x 8GB Corsair Vengeance DDR4-3200 @ 2666AMD Socket AM4 Workstation (300-Series)AMD Ryzen 5 1500X, Ryzen 5 1600X, Ryzen 5 1400MSI X370 Tomahawk4x 8GB G.Skill TridentZ DDR4-3200 @ 2667 and 3200 Intel LGA 1151 (Z270)Intel Core i7-7700KMSI Z270 Gaming 72x 8GB Corsair Vengeance DDR4-3200 @ 2400 and 3200All SystemsGeForce GTX 1080 Founders Edition (Gaming)Nvidia Quadro P6000 (Workstation)1x 1TB Toshiba OCZ RD400 (M.2, System)2x 960GB Toshiba OCZ TR150 (Storage, Images)be quiet! Dark Power Pro 11, 850W Power SupplyWindows 10 Pro (Creators Update)U.S.AMD Socket AM4 (400-Series)AMD Ryzen 7 2700XMSI X470 Gaming M7 AC2x 8GB G.Skill FlareX DDR4-3200 @ DDR4-2933, DDR4-3466Intel LGA 1151 (Z370):Intel Core i7-8700K, i5-8600K, Core i5-8400MSI Z370 Gaming Pro Carbon AC2x 8GB G.Skill FlareX DDR4-3200 @ DDR4-2400, DDR4-2667, DDR4-3466AMD Socket AM4 (300-Series)AMD Ryzen 7 1800X, 1700X, 1700, Ryzen 5 1600XMSI X370 Xpower Gaming Titanium2x 8GB G.Skill FlareX DDR4-3200 @ DDR4-2667, DDR4-3200Intel LGA 1151 (Z270)Intel Core i7-7700K MSI Z270 Gaming M72x 8GB G.Skill FlareX DDR4-3200 @ DDR4-2400All EVGA GeForce GTX 1080 FE 1TB Samsung PM863 SilverStone ST1500-TI, 1500W Windows 10 Creators Update Version 1703 - All Spectre and Meltdown mitigations
Cooling	GermanyAlphacool Eiszeit 2000 ChillerAlphacool Eisblock XPXThermal Grizzly Kryonaut (For Cooler Switch)U.S.Corsair H115i
Monitor	Eizo EV3237-BK
PC Case	Lian Li PC-T70 with Extension Kit and Mods Configurations: Open Benchtable, Closed Case
Power Consumption Measurement	Contact-free DC Measurement at PCIe Slot (Using a Riser Card) Contact-free DC Measurement at External Auxiliary Power Supply Cable Direct Voltage Measurement at Power Supply 2x Rohde & Schwarz HMO 3054, 500 MHz Digital Multi-Channel Oscilloscope with Storage Function4x Rohde & Schwarz HZO50 Current Probe (1mA - 30A, 100 kHz, DC) 4x Rohde & Schwarz HZ355 (10:1 Probes, 500 MHz) 1x Rohde & Schwarz HMC 8012 Digital Multimeter with Storage Function
Thermal Measurement	1x Optris PI640 80 Hz Infrared Camera + PI Connect Real-Time Infrared Monitoring and Recording
Acoustic Measurement	NTI Audio M2211 (with Calibration File, Low Cut at 50Hz) Steinberg UR12 (with Phantom Power for Microphones)Creative X7, Smaart v.7 Custom-Made Proprietary Measurement Chamber, 3.5 x 1.8 x 2.2m (L x D x H) Perpendicular to Center of Noise Source(s), Measurement Distance of 50cm Noise Level in dB(A) (Slow), Real-time Frequency Analyzer (RTA) Graphical Frequency Spectrum of Noise

MORE: Best CPUs

MORE: All CPUs Content

VRMark, 3DMark & AotS: Escalation

VRMark & 3DMark

These are busy charts with the addition of our retested Intel platforms, including the Spectre Variant 2 microcode patches. At stock settings, the Ryzen 5 2600X outperforms its overclocked predecessor across the board, which is especially meaningful in the lightly-threaded VRMark workload. Overclocking yields significant gains in synthetic gaming benchmarks, which don't necessarily translate to the rest of our benchmark suite.

Several of the patched Intel processors do lose performance compared to before the updates. This is particularly apparent in VRMark on Intel's Core i7-8700K, while other tests reflect minimal regression. Meanwhile, the Core i5-8400 and -8600K give us mixed results. Core i7-7700K is a more representative measure of pre- and post-patch performance, and it takes a healthy dive in VRMark as well (verified several times by removing and reinstalling the OS-based Spectre patch).

Ashes of the Singularity: Escalation

Ryzen 5 2600X is clearly superior to its predecessor in threaded titles. After all, the stock 2600X beats an overclocked 1600X. Those gains propel the Ryzen 5 up the chart, where it matches a stock Core i7-8700K.

Flip over to the album's next slide, which includes Intel CPUs before and after we patched their platforms. The Core i7-7700K loses a few frames per second in our retest, falling outside of this consistent benchmark's margin of error. Aside from the Core i5-8400's gains, which remind us that firmware updates sometimes fine-tune performance, too, most of the Intel CPUs land within the run-to-run variance we expect to see.

MORE: Best CPUs

MORE: All CPUs Content

Civilization VI Graphics & AI, Dawn of War III

Civilization VI AI Test

The Ryzen 5 2600X represents a nice step forward for AMD, particularly since Intel's processors typically lead in this test due to their per-core performance advantage.

A few of the processors exhibit slight regressions post-patch, but nothing outside of the variances we'd expect.

Civilization VI Graphics Test

An overclocked Ryzen 5 2600X beats the stock Ryzen 7 2700X, which should excite value-seekers.

Then again, Intel's Core i5-8400 slips past the tuned Ryzen 5 in this test, yielding better performance at a lower price (and despite a locked ratio multiplier).

Warhammer 40,000: Dawn of War III

Dawn of War finds the tuned Coffee Lake-based CPUs at the top of our chart. Even the fastest Ryzen (overclocked, no less) lands behind a stock Core i7-7700K.

Interestingly, the Coffee Lake CPUs enjoy slight gains after we patch them, while Core i7-7700K doesn't change much.

MORE: Best CPUs

MORE: All CPUs Content

Far Cry Primal, GTA: V & Hitman

Far Cry Primal

Ryzen 5 2600X nearly matches Intel's Core i7-8700K at stock clock rates. But tuning propels the chip ahead of an overclocked Ryzen 7 2700X and into contention with Core i5-8600K at 4.9 GHz. Clearly, this title responds well to physical core count, favoring platforms without SMT enabled.

Grand Theft Auto V

It's no surprise to see Intel's processors dominate our Grand Theft Auto V charts. An overclocked Ryzen 5 2600X essentially ties the stock Core i7-7700K, while Core i5-8400 stands out again for its higher performance and lower price point.

The post-patch Core i7-8700K averages 9 FPS-higher in GTA V. This is one of the games where -8700K historically lagged the slower Core i7-8700, so it looks like motherboard firmware fixed a few issues. We also see an improvement from the Core i5-8400.

Hitman

A recent update added a frame rate cap to Hitman, causing most of our configurations to reflect a graphics bottleneck. It's no surprise, then, that an overclocked Ryzen 5 2600X nearly matches the fastest Intel CPUs.

MORE: Best CPUs

MORE: All CPUs Content

Shadow Of War & Project CARS 2

Middle-earth: Shadow Of War

Ryzen 5 2600X beats the overclocked Ryzen 7 1800X, and tuning propels it to within 0.5 FPS of the fastest CPU.

Intel's Core i5-8400 is less expensive and tends to outperform the 2600X in many games, but the 2600X does lead in this one. Aside from the Core i5-8400, we observe lower average frame rates from the post-patch Intel processors, though the variances only amount to one or two frames per second.

Project CARS 2

Ryzen 5 2600X delivers a commendable performance in Project CARS 2, but it lags the less expensive Core i5-8400.

Aside from the Core i7-8700K's and i5-8400's performance gains, we don't significant variations related to Spectre mitigations.

MORE: Best CPUs

MORE: All CPUs Content

Office & Productivity

Adobe Creative Cloud

The stock Ryzen 5 2600X trails most of AMD's first-gen Ryzen CPUs, notably lagging behind the previous-gen Ryzen 5 1600X. We reran this benchmark several times to verify its results, and the outcome is repeatable. But given the performance observed in other tests, PCMark's Creative Cloud component may be an outlier.

Although we didn't see much performance variation from the patches in our game testing, that changes drastically in our Adobe Creative Cloud suite. Every Intel processor's overall score takes a significant haircut (the Core i7-8700K drops ~9%, while the Core i5-8400 drops ~10%).

Web Browser

The Krakken suite tests JavaScript performance using several workloads, including audio, imaging, and cryptography.

AMD's processors typically lag Intel's in Web browser benchmarks due to their lower per-core performance. The Ryzen 5 2600X is competitive with Intel's Core i5-8400 in this test. But as we noted in our Ryzen 7 2700X review, overclocking actually results in lower scores during lightly-threaded tasks. That's a bit more surprising in this case because, as we pointed out on the first page, Ryzen 5 2600X sustains up to 4.2 GHz on a single core, which is the same frequency as our all-core overclock. XFR2 contributes an extra boost during sporadic workloads though, and that's likely what we're seeing here.

The MotionMark benchmarks, which emphasize graphics performance (rather than JavaScript), are also sensitive to CPU clock rates. Ryzen 5 2600X isn't as competitive compared to the Intel models, reminding us that AMD still lags what it comes to IPC throughput.

Again, we see performance regressions from Intel's processors in these workloads, which we measured with a Spectre-patched version of Firefox.

Productivity

The application start-up metric measures load time snappiness in word processors, GIMP, and Web browsers under warm- and cold-start conditions. Other platform-level considerations affect this test as well, including the storage subsystem. Initially, we thought that'd be bad news for Intel's CPUs. After all, the security mitigations have an intense impact on I/O operations. Surprisingly, though, we actually recorded higher results from the Intel-based platforms. Meanwhile, the Ryzen 5 2600X beats AMD's first-gen Ryzen CPUs, but trails the Core i5-8400.

Our video conferencing workload measures performance in single- and multi-user applications that utilize the Windows Media Foundation for playback and encoding. It also performs facial detection to model real-world usage. Ryzen processors perform well in this test, joining an overclocked Core i7-8700K at the top of the chart. At stock settings, the Ryzen 5 2600X handily dispatches Intel's Core i5-8400.

The photo editing benchmark measures performance with Futuremark's binaries using the ImageMagick library. Common photo processing workloads also tend to be parallelized, so Ryzen 7 2700X naturally takes a lead. The Ryzen 5 2600X performs well given its price point. And the Intel CPUs all take a hit after we get them patched.

MORE: Best CPUs

MORE: All CPUs Content

Rendering, Encoding & Compression

Rendering

Threaded rendering workloads favor Ryzen's SMT-enabled cores. Ryzen 5 2600X slots in right where we'd expect it to land, while Intel's processors suffer slight performance hits after installing the Spectre mitigation patches.

Encoding & Compression

LAME is the quintessential example of a single-threaded workload, normally favoring Intel's per-core performance advantage. AMD's 2000-series Ryzen CPUs go a long way in closing the gap by offering better per-core performance than their predecessors.

Our threaded compression and decompression tests adsorb data directly from system memory, removing storage from the equation. The Ryzen 5 2600X fares well during the test, easily beating Intel's Core i5-8400 and -8600K. Given Windows' new dual page table addressing structure that prevents Meltdown-based attacks, we expected more performance overhead after the patches. However, the company's latest processors have a PCID (Post-Context Identifiers) feature that accelerates page table translations. As a result, older Core CPUs without the PCID feature are likely affected more than the ones we're testing.

There's a larger delta between Intel and AMD processors during our HandBrake x265 test compared to the x264 benchmark due to its heavier distribution of AVX instructions. The 2600X slots in where we'd expect given its six cores with SMT technology.

MORE: Best CPUs

MORE: All CPUs Content

Final Analysis

AMD’s Ryzen 5 2600X isn’t quite as impressive as the higher-end Ryzen 7 2700X, but it does offer a solid mixture of performance and value that's well-suited to many different workloads. Whereas professionals might be more interested in the 2700X's eight cores, gamers on a budget will want to check out the 2600X. After all, saving money on other system components is critical at a time when you're certain to pay a premium for discrete graphics.

In the chart below, we plot gaming performance with both average frame rates and a geometric mean of the 99^th percentile frame times (a good indicator of smoothness), which we then convert into a frame-per-second measurement. We also have price-to-performance charts that get split up to include CPUs-only and extra platform costs. For the models that don't come with a bundled cooler, we add $25 for a basic heat sink. We also add $20 if overclocking requires a more expensive motherboard (as is the case for Z370). The Intel test results reflect our patched configurations.

Games show the Ryzen 5 2600X offering a universal improvement over AMD's previous-gen Ryzen 5 1600X. Unfortunately, we see limited gains from overclocking, though that's just as well given this family's meager headroom. More important is that Ryzen 5 2600X beats the Ryzen 7 1800X throughout our suite.

Intel's Core i5-8600K is also in the 2600X’s crosshairs; AMD takes aim with a significantly lower price, a bundled thermal solution, and compatibility with less expensive motherboards. If you're not worried about overclocking, though, the Core i5-8400 is an even better buy for gaming. It offers nearly the same performance as the 2600X at a ~$50 discount. The i5-8400 drops into value-oriented B-series motherboards and comes with a stock cooler/fan, too.

Although we're big fans of the Core i5-8400 for entertainment, Ryzen 5 2600X is a smarter all-around value when it has the change to stretch its six cores and 12 threads. The processor distances itself from the i5-8400 in our rendering, encoding, compression, and decompression apps. It even challenges the eight-core Ryzen 7 1700X in several tests, particularly after tuning. That highlights the improvements borne of the Ryzen 2000-series’ enhanced multi-core boost algorithms and lower memory/cache latency.

Like all of AMD’s processors, the Ryzen 5 2600X comes with an unlocked ratio multiplier. AMD is pushing the frequency/voltage curve to its limits, so we didn’t experience massive gains in some mundane workloads. However, we did see more of a benefit with the 2600X in heavily-threaded tasks compared to the Ryzen 7 2700X. That’s largely due to the 2600X’s lower multi-core boost frequencies.

We wish AMD was ready with its B450-series motherboards at launch time. But you can still pair the Ryzen 5 2600X with a capable 300-series model.

The Spectre patches did take some wind out of Intel’s sails in many of our application tests, but the impact varies by application. In most cases, the regressions aren’t severe enough to change our recommendations. Still, it's always disappointing to observe performance stepping backward. Luckily for Intel, gaming wasn't affected much.

Intel beefed up its Coffee Lake-based Core i5s by adding 50% more cores. Up against the Ryzen 5 1600X, we couldn't help but acknowledge Intel's great performance and generally better compatibility with existing games. This time, however, AMD brings the heat in our benchmarks, while most of its optimization-oriented issues are ancient history. If gaming is your only concern, save some cash and pick up a Core i5-8400. But we think you’ll be happier with the Ryzen 5 2600X, which has more resources to handle general desktop workloads with ease.

MORE: Best CPUs

MORE: All CPUs Content

AMD Ryzen 7 2700X Review: Redefining Ryzen

palcorn@outlook.com (Paul Alcorn) — Thu, 19 Apr 2018 13:00:00 +0000

Redefining Ryzen

AMD's return to prominence last year found it rolling out a long stream of CPUs that pressured Intel in almost every segment of the desktop PC market. Even after Intel countered with dramatic adjustments to its processor portfolio, AMD continues gobbling up market share. Even in the face of stiff competition, AMD says it enjoys as much as 50% of CPU sales to DIYers on sites like Newegg and Amazon.

And a slow transition to 10nm manufacturing continues to leave Intel vulnerable. AMD is now ready to evolve its Zen architecture with a round of new processors. To be sure, the improvements they offer are iterative. The low-hanging fruit that made it possible for first-gen Ryzen to compete are already baked in. These chips do incorporate some notable advantages, though.

To begin, second-generation Ryzen processors are manufactured using an optimized 12nm LP node that promises performance and efficiency gains compared to the original Ryzen's 14nm LPP process. AMD also tweaked the Zen architecture, now dubbed Zen+, to support higher frequencies, more sophisticated multi-core boost rates, and faster memory/caches. Overall, the company claims that its 2000-series facilitates nearly equivalent gaming performance compared to similarly-priced Core CPUs, plus a 20% advantage in threaded workloads.

AMD certainly hasn't forgotten its core message: more cores and features for less money. The second-gen Ryzen processors are priced competitively, all models come with beefy stock coolers, and they are backward compatible with older Socket AM4 motherboards. AMD even throws in free caching software to sweeten the deal. It all starts with silicon though, so let's take a look.

Ryzen 7 2700X

Ryzen 2000-series processors, otherwise known by their "Pinnacle Ridge" code name, are based on the same basic Zen core design as previous-gen models. But they benefit from 12nm manufacturing, along with targeted tweaks to improve cache and memory latency. The company says its resulting Zen+ architecture delivers up to a 3% boost in IPC (instructions per cycle) throughput.

The CPUs still utilize a dual-CCX configuration tied, together with Infinity Fabric. Not surprisingly, then, they're divided into eight-core, 16-thread Ryzen 7 and six-core, 12-thread Ryzen 5 families, both with 16MB of L3 cache. Like the Ryzens that came before, all 2000-series models boast unlocked ratio multipliers for easy overclocking. Intel, in comparison, still charges a premium for its overclockable K-series SKUs.

AMD separates its the 2000-series stack into high-performance X-series models and their non-X counterparts. But it shrinks the Ryzen 7 family from three models to two. Ryzen 7 2700X would seem to suggest a Ryzen 7 1700X replacement. However, it actually replaces the flagship Ryzen 7 1800X. AMD claims that its 2700X offers up to 12% more performance than Ryzen 7 1800X in threaded applications. Much of that improvement comes from a 100 MHz-higher base clock and 200 MHz of additional boost frequency (though multiple other refinements also contribute).

While the Pinnacle Ridge processors drop into 400-series motherboards, AMD is only releasing its X470 chipset at launch time. We still don't have a release date for the less expensive B450- and A420-based motherboards. As the company originally promised, it continues supporting Socket AM4 (and purportedly will until 2020), so the new Ryzen CPUs also work with 300-series motherboards after a BIOS update. First-gen Ryzens do work with 400-series platforms as well, allowing you to drop an older CPU into a brand-new board, if desired.

	AMD Ryzen 7 2700X	AMD Ryzen 7 1800X	AMD Ryzen 7 2700	AMD Ryzen 5 1600X	AMD Ryzen 5 2600X	AMD Ryzen 5 2600	Intel Core i7-8700K	Intel Core i7-8700	Intel Core i5-8600K	Intel Core i5-8400
MSRP	$329	$349	$299	$219	$229	$199	$359	$303	$257	$182
Cores/Threads	8/16	8/16	8/16	6/12	6/12	6/12	6/12	6/12	6/6	6/6
TDP	105W	95W	65W	95W	95W	65W	95W	65W	95W	65W
Base Freq. (GHz)	3.7	3.6	3.2	3.6	3.6	3.4	3.7	3.2	3.6	2.8
Precision Boost Freq. (GHz)	4.3	4.1	4.1	4.0	4.2	3.9	4.7	4.6	4.3	4.0
Cache (L3)	16MB	16MB	16MB	16MB	16MB	16MB	12MB	12MB	9MB	9MB
Unlocked Multiplier	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Yes	No
Cooler	105W Wraith Prism (LED)	-	95W Wraith Spire (LED)	-	95W Wraith Spire	65W Wraith Stealth	-	Intel	-	Intel

The $329 Ryzen 7 2700X should sell for $20 less than an 1800X (though prices of previous-generation chips will likely fall as long as stock is robust), while the $299 Ryzen 7 2700 lands right where you formerly found the 1700. AMD's non-X models were apparently more popular with enthusiasts since they also had unlocked multipliers, enabling similar performance as the pricier models (after some tuning) for less money. You could save $50 stepping down from Ryzen 7 1800X to the 1700, for example. But that gap shrinks to $30 this time around.

AMD's Ryzen 7 2700X grapples with Intel's $359 flagship Core i7-8700K. Though that seems like an uncomfortably close comparison, AMD allows you to overclock with one of its value-oriented B-series motherboards (B350-based, for now), whereas Intel compels users splurge on a Z-series platform for overclocking. Adding the CPU and motherboard together, Intel's premium ends up being quite a bit higher.

Second-gen Ryzens now support up to DDR4-2933 RAM as well, trumping the Coffee Lake architecture's official DDR4-2666 ceiling (with a few caveats that we'll cover shortly). More bandwidth should help latency-sensitive apps, such as games. Also, X470 motherboards pave the way for better memory overclocking than previous-gen platforms.

AMD's first-gen X-series processors, which topped out at 95W, came without a bundled thermal solution. This time around, all 2000-series CPUs include a cooler. The 105W Ryzen 7 2700X includes a "Wraith Prism" LED cooler that features four direct-contact copper heat pipes, three independent RGB zones, switchable fan profiles, and a 39 dB(A) noise rating. The cooler is rated to dissipate 116W of waste heat in "L" mode (2800 RPM) and 124W in "H" mode (3600 RPM). Cooler Master manufactures the heat sink/fan, while AMD provides software for controlling the lighting and fan profiles. Company representatives claim the cooler represents a roughly $43 value, and that it also allows for some overclocking headroom.

As with the generation before, AMD employs Indium solder between its die and heat spreader to improve thermal transfer. In contrast, Intel uses standard thermal interface material on its Core i7-8700K. Also, that Intel chip doesn't come with a cooler, widening the price gap between a Ryzen 7 2700X-based configuration and a current-generation unlocked Intel Core i7.

According to AMD, its 2000-series CPUs benefit from an improved SensMI suite that also includes its new StorMI Technology. The latter is a software-based tiering solution that melds the low price and high capacity of a hard drive with the speed of an SSD, 3D XPoint (including Intel's Optane parts), or even up to 2GB of RAM. AMD sold this software as a $20 add-on in the past, but now it comes free as part of the 2000-series package. As with any tiering utility, you assume the same risks of data loss inherent to a RAID 0 array. For more details about this software, read our feature: AMD and Enmotus Expand FuzeDrive Offerings.

The GlobalFoundries 12nm LP Process

As mentioned, AMD's 2000-series CPUs are not manufactured on GlobalFoundries' 14nm GPP node, but rather its 12nm LP process technology. The ported-over design helps boost transistor performance, but does not affect die area or transistor density. As a result, Pinnacle Ridge's ~4.8 billion transistors and 213mm² area remain the same as first-gen Ryzen.

Lower leakage current does enable roughly 300 MHz-higher clock rates or a 50mV core voltage reduction at any given frequency compared to 14nm manufacturing. The company also refined some of the architecture's critical pathways with higher-performance transistors. All told, AMD claims the 12nm design enables up to 11% less power consumption than 14nm-based Ryzen CPUs at the same clock rates, or up to 16% more performance at the same thermal design power. All-core overclocks are expected to land in the 4.2 GHz range moving forward.

AMD also adds other nuanced refinements to the performance story, reportedly improving L1, L2, and L3 cache latencies, while also reducing memory latency by 11%.

Ryzen 7 2700X's 105W TDP represents a 10.5% increase compared to the 1800X for a 4.65% increase in boost frequencies. That seems like a simple trade-off of power consumption for higher clock rates. But the TDP rating also takes the multi-core Precision Boost 2 and XFR2 algorithms' higher power draw into account, allowing access to Socket AM4's full 95-amp current ceiling even during stock operation.

Precision Boost 2 And XFR2

AMD's previous-gen Ryzen processors have Precision Boost (a Dynamic Voltage Frequency Scaling implementation similar to Intel's Turbo Boost), and eXtended Frequency Range, which provides additional frequency uplift if your cooling solution has thermal headroom to spare. Those 1000-series CPUs only offer dual-core or all-core Precision Boost and XFR clock rates. But lightly-threaded applications (like games) often offload less-critical tasks to other threads. Unfortunately, light helper threads can apply enough of a load to trigger the lower all-core frequency, limiting performance potential even when the CPU could be operating at higher clock rates.

The new Precision Boost 2 (which debuted on the desktop with AMD's Raven Ridge processors) and XFR2 algorithms improve performance in threaded workloads by raising the frequency of any number of cores. Precision Boost 2 delivers up to 500 MHz-higher clocks during multi-core workloads, while XFR2 adds an additional 7% boost if your cooler is beefy enough. This extends Ryzen's already-strong threaded performance to a wider variety of tasks, though it levels off when the processor reaches 60°C (tCase) or 95 amps of current. Precision Boost 2 and XFR2 also work on 300-series motherboards.

AMD doesn't share a list of specific multi-core Precision Boost 2 and XFR2 bins, because its opportunistic algorithms achieve different frequencies based on temperature, current, and load.

MORE: Best CPUs

MORE: Intel & AMD Processor Comparison Hierarchy

MORE: All CPUs Content

AMD Ryzen 7 2700X

AMD Ryzen 5 2600X

X470 And Ryzen Master 1.3

AMD's Socket AM4 was designed with robust power delivery capabilities that aren't entirely used by first-gen Ryzen processors. The 2000-series chips are much better at leveraging the platform's current headroom through their improved boost algorithms. Some value-oriented motherboards employ scaled-back power delivery capabilities, so AMD's second-gen Ryzen CPUs communicate with the platform to modulate performance based on what the motherboard can do. That's a necessary addition to accommodate Ryzen 7 2700X's 105W TDP, which didn't exist before this new chip line. As a result, less-capable motherboards may not expose the full performance potential of higher-TDP processors like the Ryzen 7 2700X.

The processor monitors Package Power Tracking (PPT) and Thermal Design Current (TDC) variables, measuring available margin to the motherboard's maximum power output and current, respectively. Electrical Design Current (EDC) also indicates the maximum current possible from the VRMs during peak/transient conditions. A control loop feeds the real-time telemetry data back to the Infinity Fabric, which then allows the processor to dynamically affect performance based on thermal and power conditions.

If the motherboard BIOS supports it, AMD exposes some of these monitoring features with its updated Ryzen Master 1.3 overclocking software. The fastest cores are identified during the binning process and flagged by Ryzen Master with gold stars on a per-CCX basis. The third- and fourth-fastest cores are marked with a circle.

AMD's software now supports per-CCX overclocking as well, and includes a built-in stress test. The warranty does not cover damage caused by overclocking, so exercise caution.

Because there are still plenty of 300-series motherboards available for sale, AMD designed a badge to let you know that a firmware update may be necessary before dropping a 2000-series CPU into one of those older platforms. Unless your 300-series motherboard has an out-of-band update mechanism like BIOS Flashback, you need a previous-gen Ryzen processor to update it. AMD also offers its Boot Kit solution, which is a loaner processor you can use to update the motherboard firmware.

Eventually, all 300-series motherboards will support 2000-series processors right out of the box. AMD expects X470 and X370 boards to coexist for the foreseeable future, so it may be possible to find excellent deals on those previous-gen motherboards.

DIMM Slots Filled	Memory Ranks	Supported Speed
2 of 2	Single	2933*
2 of 2	Dual	2677
2 of 4	Single	2933*
2 of 4	Dual	2400
4 of 4	Single	2133
4 of 4	Dual	1866

*Note: requires a motherboard with at least six PCB layers. DDR4-2667 is supported on four-layer PCBs.

AMD's 2000-series processors support up to DDR4-2933 with a pair of single-rank DIMMs, though you need a six-layer motherboard to unlock that capability. Support drops back to DDR4-2667 for four-layer motherboards. Fortunately for enthusiasts, most mainstream platforms utilize six or eight layers.

From what we've seen thus far, X470 motherboards have an improved layout to facilitate aggressive memory overclocking. As you might expect, X470 boards in our labs are much more mature at launch than the 300-series platforms we battled last year. Thanks to this, we're easily running memory at DDR4-3466 with tight timings. Our motherboard team also noticed vastly improved overclocking with all memory slots populated, which was an issue on some X370 motherboards.

X470-based motherboards feature lower power consumption, higher multi-hub USB throughput, and improved power delivery. But they still have the same connectivity options as 300-series motherboards.

I/O Source	USB 3.1 Gen2	USB 3.1 Gen1	USB 2.0	PCIe Gen3	GPP PCIe Gen2	SATA
AMD Ryzen SoC (1000- and 2000-series)	0	4	0	20x	0	2
X470/370	2	6	6	0	8	8
B350	2	2	6	0	6	6
A320	1	2	6	0	4	6

The first line in our chart covers Ryzen's I/O capabilities, which you then combine with one of the chipsets underneath to determine platform connectivity. A Ryzen CPU sports 20 PCIe 3.0 lanes. Sixteen are dedicated to the PCIe slots, while four lanes are dedicated to SATA ports or a 4x link for NVMe SSDs. Four of the SATA ports can also be assigned to SATA Express interfaces at a 2:1 ratio, yielding a maximum of two SATA Express connections.

As you can see, the X470 chipset offers the same connectivity options as its predecessor, with two USB 3.1 Gen2 ports, four USB 3.1 Gen1 ports, six USB 2.0 ports, and eight general-purpose PCIe 2.0 lanes that vendors can carve up for additional functionality (like hanging M.2 slots off of the chipset or enhanced 5/10GbE support).

MORE: Best CPUs

MORE: All CPUs Content

Cache And Memory Performance, IPC

Memory Latency

AMD Measurements	L1 Cache Latency	L2 Cache Latency	L3 Cache Latency	Memory Latency
Latency Improvements	13%	34%	16%	11%

AMD's first-gen processors demonstrated higher memory latency than we expected, affecting the performance of memory-sensitive applications. The company claims it reduced memory latency by 11% this time around, as well as cutting cache latencies by double-digit percentages. We'll start by measuring the memory and Infinity Fabric subsystems, and then move on to IPC tests.

SiSoftware's Sandra is used to measure cache and memory latency with three different access patterns, giving us more granularity than a single test. Sequential access patterns are almost entirely prefetched into the TLB, so that one's a good measure of prefetcher performance. The in-page random test measures random accesses within the same memory page. It also measures TLB performance and represents best-case random performance. The full random test features a mix of TLB hits and misses, with a strong likelihood of misses, so it quantifies worst-case latency.

We tested both the Ryzen 7 1800X and Ryzen 7 2700X on the same X470 motherboard. We include results with the Ryzen 7 2700X at DDR4-2933 for the stock configuration, DDR4-3466 for the overclocked configuration, and DDR4-2666 to normalize it with AMD's Ryzen 7 1800X.

With normalized DDR4-2667 data rates and timings, the Ryzen 7 2700X posts impressive gains over Ryzen 7 1800X, regardless of the data access pattern. As percentages, the 2700X's improvements weigh in at 11.49% for full random, 6.64% for in-page, and 9.35% for the sequential access pattern.

The Infinity Fabric speeds up as we increase memory frequency to the 2700X's default DDR4-2933. This fabric ties the IMC and cores together, so we record even larger improvements of 18% in the full random test, 13.4% with a full random access pattern, and 12.9% with the sequential metric.

AMD isn't fully disclosing the steps it took to improve memory latency, but we suspect the company worked on the Infinity Fabric and integrated memory controller to realize these gains.

Cache Latency And Bandwidth

Regardless of the memory access pattern, the smallest data chunks fit into L1 cache. As the data gets larger, it populates the 2700X's higher tiers of cache, which we outlined in the following table:

	L1	L2	L3	Main Memory
Range	2KB - 32KB	64KB - 512KB	1MB - 4MB	8MB - 1GB

% Improvement Over 1800X	L1	L2	L3
In-Page	11.11%	51.72%	26.38%
Full-Random	11.11%	53.5%	25.64%
Sequential	11.11%	13.3%	13.3%

The cache latency reductions that we measured are even better than what AMD suggested we'd see, though its lab might be using different access patterns. Regardless, the apples-to-apples results in our table are downright impressive.

We also see a notable increase in cache bandwidth. Feeding the cores with lower latency and higher throughput is a win-win on the performance front. Intel's S-series processors still have a big single-core L1 bandwidth advantage, but AMD's updated L2 cache is measurably faster than the 1800X in both single- and multi-threaded tests. AMD even enjoys better L2 cache latency than Intel in the sequential test and better L3 cache latency with several data patterns.

To Infinity, And Beyond

The updated Zen+ design fuses two four-core CCXs together with the Infinity Fabric, which is a crossbar that also handles IMC, northbridge, and PCIe traffic. As such, fabric latency is a critical variable that ensures the memory latency gains we observe can actually be delivered to the cores.

SiSoftware Sandra's Processor Multi-Core Efficiency metric helps illustrate the Infinity Fabric's performance. We use the Multi-Threaded test with the "best pair match" setting (lowest latency). The utility measures ping times between threads to quantify fabric latency in every possible configuration. We boil those benchmarks down to latency averages for the different pathways, but head here for a more detailed explanation of the various components.

AMD reduced Ryzen 7 2700X's intra-core latency by 11.8% and the critical cross-CCX latency by 8.3%. We also notice that Ryzen 7 2700X offers significantly improved fabric bandwidth.

Instructions Per Clock

It's important to remember that IPC can vary by workload, so dissimilar tasks may yield different outcomes. We set a static 3 GHz clock rate for the following tests:

Our single-core Cinebench benchmark suggests a 1.6% IPC improvement favoring Ryzen 7 2700X. But while AMD does improve, Intel still holds onto a distinct IPC throughput advantage. Switching to the Multi-Threaded Cinbench test gives Ryzen 7 2700X a 2.7% improvement over its predecessor.

Core i9-7820X employs two 256-bit AVX FMA units per core that operate in parallel, whereas Ryzen's Zen architecture divides 256-bit AVX operations across two FMA units per core. That difference hands the Skylake-X processor a commanding lead in y-cruncher. We do see a 3.9% increase in the 2700X's Multi-Threaded y-cruncher result compared to Ryzen 7 1800X. But the gains in single-threaded AVX performance are marginal.

We see similar results in our single-core cryptographic tests, though Ryzen 7 2700X takes an 8.7% lead over the 1800X in the Multi-Threaded AES-256-ECB encryption workload. AMD's Zen architecture includes two AES cryptographic accelerators for each core, so it isn't surprising to see Ryzen dominate over Intel's S-series CPUs in the AES-256-ECB tests.

MORE: Best CPUs

MORE: All CPUs Content

Overclocking, Spectre, And Test Setup

Overclocking

We ran our gaming and application tests in the U.S. lab, while power/thermal measurements were collected in our German lab.

In the U.S. lab, we paired our Ryzen 7 2700X with Corsair's H115i cooler for overclocking. This allowed us to maintain a 4.2 GHz all-core frequency at 1.3785V Vcore, 1.2V SoC voltage, and the default Load Line Calibration settings. Since we couldn't smash through to 4.3 GHz without exceeding AMD's 1.40V maximum recommended Vcore setting, we stopped at 4.2 GHz.

We did encounter temperatures as high as 90°C during extended AVX testing, so we recommend a capable closed-loop or custom water cooler for overclocking. AMD would really benefit from an AVX-offset feature as well, which could cut clock rates during power-hungry AVX workloads. Should you choose to go the more extreme route, there have been reports of 5.8 GHz with Ryzen 7 2700X under LN2.

First-gen Ryzen processors don't have much memory overclocking headroom, so we're still testing tuned X370 platforms at DDR4-3200. However, the X470 platform was remarkably stable at higher data rates with Ryzen 7 2700X. So, we settled on DDR4-3466 with 14-14-14-34 timings (though we're confident that more time to tune would yield even higher overclocks). We also ran our overclocked Intel processors at DDR4-3466.

Spectre And Meltdown

Our test rigs now include Meltdown And Spectre Variant 1 mitigations. Spectre Variant 2 requires both motherboard firmware/microcode and operating system patches. We have installed the operating system patches for Variant 2.

Today's performance measurements do not include Intel's motherboard firmware mitigations for Spectre Variant 2 though, as we've been waiting for AMD patches to level the playing field. Last week, AMD announced that it’s making the mitigations available to motherboard vendors and OEMs, which the company says should take time to appear in the wild. We checked MSI's website for firmware updates applicable to our X370 platforms when AMD made its announcement, but no new BIOSes were available (and still aren't).

Unfortunately, we were only made aware that Variant 2 mitigations are present in our X470 board's firmware just before launch, precluding us from re-testing the Intel platforms with patches applied. We're working on this now, and plan to post updated results in future reviews.

The lack of Spectre Variant 2 patches in our Intel results likely give the Core CPUs a slight advantage over AMD's patched platforms. But the performance difference should be minimal with modern processors.

Test Setup

AMD is working on a Precision Boost Overdrive feature, which seems similar to the Multi-Core Enhanced Turbo (MCE) feature that allows Intel's K-series processors to run at their maximum Turbo Boost bin across all cores at all times. The setting on Intel platforms modifies the CPU's clock rate and voltage to deliver higher performance, basically amounting to factory-sanctioned overclocking.

AMD's Ryzen Master 1.3 software doesn't currently let you activate this feature from within Windows. But as we often find with MCE, AMD's Precision Boost Overdrive is enabled by default in many BIOSes. After extensive experimentation, we can conclude that the option doesn't deliver an appreciable performance gain in its current form. Thus, we ran our tests with Precision Boost Overdrive disabled.

Comparison Products

Intel Core i5-8400

Intel Core i5-8600K

Intel Core i7-7700K

Test Systems

Test System & Configuration
Hardware	Germany AMD Socket AM4 (400-Series)AMD Ryzen 7 2700XMSI X470 Gaming M7 AC2x 8GB G.Skill FlareX DDR4-3200 @ DDR4-2933, DDR4-3466Intel LGA 1151 (Z370):Intel Core i5-8600K, i5-8600K, Core i5-8400MSI Z370 Gaming Pro Carbon AC2x 8GB Corsair Vengeance DDR4-3200 @ 2666AMD Socket AM4 Workstation (300-Series)AMD Ryzen 5 1500X, Ryzen 5 1600X, Ryzen 5 1400MSI X370 Tomahawk4x 8GB G.Skill TridentZ DDR4-3200 @ 2667 and 3200 Intel LGA 1151 (Z270)Intel Core i7-7700KMSI Z270 Gaming 72x 8GB Corsair Vengeance DDR4-3200 @ 2400 and 3200All SystemsGeForce GTX 1080 Founders Edition (Gaming)Nvidia Quadro P6000 (Workstation)1x 1TB Toshiba OCZ RD400 (M.2, System)2x 960GB Toshiba OCZ TR150 (Storage, Images)be quiet! Dark Power Pro 11, 850W Power SupplyWindows 10 Pro (Creators Update)U.S.AMD Socket AM4 (400-Series)AMD Ryzen 7 2700XMSI X470 Gaming M7 AC2x 8GB G.Skill FlareX DDR4-3200 @ DDR4-2933, DDR4-3466Intel LGA 1151 (Z370):Intel Core i7-8700K, i5-8600K, Core i5-8400MSI Z370 Gaming Pro Carbon AC2x 8GB G.Skill FlareX DDR4-3200 @ DDR4-2400, DDR4-2667, DDR4-3466AMD Socket AM4 (300-Series)AMD Ryzen 7 1800X, 1700X, 1700, Ryzen 5 1600XMSI X370 Xpower Gaming Titanium2x 8GB G.Skill FlareX DDR4-3200 @ DDR4-2667, DDR4-3200Intel LGA 1151 (Z270)Intel Core i7-7700K MSI Z270 Gaming M72x 8GB G.Skill FlareX DDR4-3200 @ DDR4-2400Intel LGA 2066Intel Core i7-7820XMSI X299 Gaming Pro Carbon AC4x 8GB G.Skill FlareX DDR4-3200 @ DDR4-2666All EVGA GeForce GTX 1080 FE 1TB Samsung PM863 SilverStone ST1500-TI, 1500W Windows 10 Creators Update Version 1703
Cooling	GermanyAlphacool Eiszeit 2000 ChillerAlphacool Eisblock XPXThermal Grizzly Kryonaut (For Cooler Switch)U.S.Corsair H115i
Monitor	Eizo EV3237-BK
PC Case	Lian Li PC-T70 with Extension Kit and Mods Configurations: Open Benchtable, Closed Case
Power Consumption Measurement	Contact-free DC Measurement at PCIe Slot (Using a Riser Card) Contact-free DC Measurement at External Auxiliary Power Supply Cable Direct Voltage Measurement at Power Supply 2x Rohde & Schwarz HMO 3054, 500 MHz Digital Multi-Channel Oscilloscope with Storage Function4x Rohde & Schwarz HZO50 Current Probe (1mA - 30A, 100 kHz, DC) 4x Rohde & Schwarz HZ355 (10:1 Probes, 500 MHz) 1x Rohde & Schwarz HMC 8012 Digital Multimeter with Storage Function
Thermal Measurement	1x Optris PI640 80 Hz Infrared Camera + PI Connect Real-Time Infrared Monitoring and Recording
Acoustic Measurement	NTI Audio M2211 (with Calibration File, Low Cut at 50Hz) Steinberg UR12 (with Phantom Power for Microphones)Creative X7, Smaart v.7 Custom-Made Proprietary Measurement Chamber, 3.5 x 1.8 x 2.2m (L x D x H) Perpendicular to Center of Noise Source(s), Measurement Distance of 50cm Noise Level in dB(A) (Slow), Real-time Frequency Analyzer (RTA) Graphical Frequency Spectrum of Noise

MORE: Best CPUs

MORE: All CPUs Content

VRMark, 3DMark And AotS: Escalation

VRMark & 3DMark

Gaming is where rubber meets the road for most enthusiasts. AMD tells us that its 2000-series processors should be nearly equivalent to Intel's comparable models, at least at stock settings. And there's no doubt that Ryzen 7 2700X will excel in heavily-threaded titles. But tests that are sensitive to clock rate and IPC throughput, such as VRMark, have traditionally been a challenge for Ryzen.

The 2700X bridges the gap between Intel's processors and first-gen Ryzen. AMD's stock 2700X outstrips the Ryzen 7 1800X by 11.8%. More impressively, it also beats the overclocked 1800X by 2%. The 2700X's lead over its predecessor extends further after tuning its cores and memory subsystem.

Synthetic benchmarks are great because they tend to scale more clearly than real-world applications. 3DMark's real usefulness lies in measuring the amount of performance available to game engines, giving us a peek at what highly-optimized games could be capable of.

Ryzen 7 2700X's 16 threads beat Core i7-8700K's 12 threads in our DX11 and DX12 CPU tests, even after overclocking. The 2700X also bests Intel's $589 Core i7-7820X during both tests.

Ashes of the Singularity: Escalation

Ashes of the Singularity: Escalation evokes memories of AMD's early struggles with the Zen architecture. This was one of the first games to receive an update optimized for AMD's processor design.

Although the patch improved performance, Ryzen 7 1800X still fails to beat a stock Core i7-8700K. But Ryzen 7 2700X and its Precision Boost 2 algorithm turn the tables, giving AMD an advantage in stock and overclocked trim.

MORE: Best CPUs

MORE: All CPUs Content

Civilization VI Graphics & AI, Dawn of War III

Civilization VI AI Test

Civilization's AI test measures performance in a turn-based strategy game and tends to favor per-core performance. Ryzen 7 2700X almost ties the Coffee Lake-based Core i7-8700K at its stock settings. However, Intel gains more from overclocking, pulling away after our tuning efforts.

Again, notice that the stock Ryzen 7 2700X is fast enough to beat the overclocked 1800X.

Civilization VI Graphics Test

The stock Ryzen 7 2700X beats every other Ryzen CPU by ~10 FPS (or more). But Intel's stock CPUs have no problem maintaining their advantage.

Tuning the 2700X yields a 6.3% speed-up on average. But that doesn't help it catch the Core i5-8600K, which takes the top two spots in our chart.

Warhammer 40,000: Dawn of War III

The overclocked Ryzen 7 2700X lands just shy of the Core i7-7700K and -8700K. Tuning those chips allows them to walk away from AMD's flagship, though.

Meanwhile, Ryzen 7 2700X beats AMD's Ryzen 7 1800X by 14% right out of the box.

MORE: Best CPUs

MORE: All CPUs Content

Far Cry Primal, GTA: V, Hitman

Far Cry Primal

Far Cry Primal responds well to Intel's Core i7-7700K. It's also interesting that a stock Core i5-8600K beats the overclocked 6C/12T Core i7-8700K. Then again, we've seen this tendency before.

The stock Ryzen 7 2700X lags behind Intel's newest K-series CPUs, though tuning does help AMD's case.

Grand Theft Auto V

Grand Theft Auto V favors Intel architectures and, more generally, multi-core designs with high clock rates. The overclocked Ryzen 7 2700X vies with Intel's chips at stock frequencies. But again, giving Coffee Lake the same treatment propels those chips to the top of our chart.

Ryzen 7 1700 suffers from a low base clock rate and languishes at the bottom of our chart as a result. This processor often provides similar performance as AMD's Ryzen 7 1800X after some tuning. However, it's clear that the Ryzen 7 2700X sets a new high water mark for AMD CPUs in games.

Hitman

Hitman's GOTY update imposed a hard 90 FPS cap on performance, so this title no longer scales well with high-end PCs. Unfortunately, some popular AAA games employ similar frame rate limits, so we leave this result in place to show that not all titles respond to faster components.

We do see slight scaling from Ryzen 7 1700 up to Intel's overclocked models. However, these differences would be hard to spot during a gaming session. Ryzen 7 2700X lands in a familiar position ahead of AMD's previous-gen Ryzen CPUs.

MORE: Best CPUs

MORE: All CPUs Content

Shadow Of War, Project CARS 2

Middle-earth: Shadow Of War

Middle-earth: Shadow of War doesn't scale as dramatically as some of our other benchmarks, and it certainly isn't as sensitive to IPC throughput and frequency as Shadow of Mordor. While CPU reviews tend to focus on games that scale well with certain host processing specifications, some games just can't get enough graphics performance.

Ryzen 7 2700X comes tantalizingly close to matching the Coffee Lake-based processors at stock and overclocked settings.

Project CARS 2

Project CARS 2 is purportedly optimized for threading. A 6C/6T Core i5-8600K beats the overclocked 8C/16T Ryzen 7 2700X though, so it's clear that parallelism isn't the most influential factor in defining this game's performance.

A stock Ryzen 7 2700X trails all of the Intel CPUs except for Core i7-7820X. But overclocking nudges AMD's flagship closer to the top.

MORE: Best CPUs

MORE: All CPUs Content

Office And Productivity

Adobe Creative Cloud

Adobe's Creative Cloud suite generally favors higher frequencies and IPC throughput, giving Intel an advantage. Still, Ryzen 7 2700X provides a nice balance of high clock rates and core count, yielding an impressive 13.8% speed-up in the overall score compared to AMD's tuned Ryzen 7 1800X.

Overclocking doesn't deliver the big gains we recorded in our gaming suite. In some of the lightly-threaded application tests (like After Effects), a stock Ryzen 7 2700X is even faster than the overclocked one. This is a result of the 2700X's 4.3 GHz Precision Boost 2 frequency, which outstrips our 4.2 GHz all-core overclock. These tests also aren't as latency-sensitive as gaming workloads, so DDR4-3466 doesn't deliver as much of a performance improvement.

Web Browser

The Krakken suite tests JavaScript performance using several workloads, including audio, imaging, and cryptography. AMD's processors typically lag Intel's in Web browser benchmarks due to their lower per-core performance. However, a stock Ryzen 7 2700X still outpaces its overclocked predecessor.

The MotionMark benchmarks, which emphasize graphics performance (rather than JavaScript), are also sensitive to CPU clock rates. Again, the 2700X's higher stock Precision Boost 2 frequencies allow it to slip past the overclocked configuration. At the same time, we measure a 14.6% gain over the stock 1800X.

Productivity

Ryzen 7 2700X is much more competitive than AMD's previous-generation CPUs. We also observe slim gains from overclocking in many of these workloads.

Our video conferencing workload measures performance in single- and multi-user applications that utilize the Windows Media Foundation for playback and encoding. It also performs facial detection to model real-world usage. Not surprisingly then, a stock Ryzen 7 2700X leads the rest of the field at default clock rates thanks to its 8C/16T configuration and higher frequencies.

MORE: Best CPUs

MORE: All CPUs Content

Rendering, Encoding, And Compression

Rendering

Ryzen 7 2700X takes a commanding lead in the multi-core Cinebench benchmark, which we expected in light of the radical cache latency and bandwidth improvements that AMD made. POV-Ray also shows the 2700X to be a chart-topper, though again it's faster in stock form than overclocked.

AMD's Ryzen 7 2700X leads in many of the threaded workloads, but isn't as impressive in workloads that tax a single core. There, Intel's architectures continue shining.

Core i7-7820X leads in LuxMark. But notice that we don't have OpenCL results for it. This is because the older OpenCL SDK doesn't support AVX-512. Intel updated the SDK fairly recently, and it works correctly with Skylake-X-based processors. We'll have to retest all of these CPUs to reflect the changes, but be assured that AVX-512 is a powerful addition.

Encoding & Compression

LAME is the quintessential example of a single-threaded workload, and the 2700X posts solid gains over Ryzen 7 1800X in its stock configuration.

Our threaded compression and decompression tests adsorb data directly from system memory, thus removing storage from the equation. As per usual, the Ryzen processors dominate the decompression workload while Intel's Skylake-X leads in compression-oriented benchmarks. It's notable that Core i7-8700K needs overclocking in order to beat AMD's flagship.

There's a larger delta between Intel and AMD processors during our HandBrake x265 test compared to the x264 benchmark due to its heavier distribution of AVX instructions. Ryzen 7 2700X is particularly impressive in the x264 metric, where it upsets the capable Core i7-7820X.

We also provide results from y-cruncher, a single- and multi-threaded program that computes Pi using AVX instructions. We tested with version 0.7.3.9474, which includes Ryzen optimizations. The 2700X trails Intel's portfolio in the single-core benchmark. However, parallelization puts it in a more competitive position. Also, we clearly see the benefit of Core i7-7820X's dual 256-bit AVX FMA units (per core) in the AVX workloads.

MORE: Best CPUs

MORE: All CPUs Content

XFR2 vs. Manual Overclocking

XFR2, Cooling, and Clock Rates

In contrast to the first-gen Ryzen models and their strange temperature curves, the tCTL (core temperature) values now correspond with what we'd expect to see. AMD does add a 10-degree offset to the 2700X specifically, which motherboard BIOSes already take into account. We subtract this offset from our own measurements.

Chip quality naturally influences achievable clock rates as well. These effects are seen much more clearly with second-gen Ryzen CPUs supporting XFR2, since they have to be binned precisely.

With as much cooling performance as we could muster, Ryzen 7 2700X still reaches almost 4.2 GHz. A more conventional thermal solution would result in a lower clock rate. However, with a good air cooler, it should be possible to sustain 4 GHz on all cores.

Manual Overclocking

Ryzen 7 2700X can be manually overclocked to 4.3 GHz. But the 1.475V required for this is more aggressive than we want to get long-term. Pushing to 4.35 GHz resulted in a crash no matter how much voltage we applied.

As the following curve shows, power consumption and performance in Cinebench are almost directly proportional, so long as the system runs stably and doesn't crash. It's also worth noting that Cool'n'Quiet is completely disabled on our test platform when we configure the ratio multiplier manually. When that happens, the configured clock rate doesn't drop from its specified maximum, even at idle.

We measure a maximum of 135W in Cinebench and just over 150W in Prime95 with AVX, although this extreme torture test is more of an exhibition.

If you spend some money on good cooling, there's no reason to manually overclock Ryzen 7 2700X. Thanks to XFR2, AMD's flagship should remain stable above 4 GHz, even under full load. Try to go any higher and you'll pay a hefty price in heat, power, and possibly long-term reliability.

MORE: Best CPUs

MORE: All CPUs Content

Power Consumption

At idle, Ryzen 7 2700X lands behind most of the Intel competition, but ahead of previous-gen Ryzen CPUs. First and second place in our chart go to a couple of AMD APUs, perhaps surprisingly.

Under a light CAD workload, Ryzen 7 2700X performs better and uses less power than its predecessor. This shows us that AMD didn't pay for better clocks with a sacrifice to power consumption. Its progress is already apparent at this point in the measurements.

Gaming tells a similar story; the performance increase is again more pronounced than the differences in power consumption.

When it comes to our stress test, AMD's Ryzen 7 2700X is much more reserved than its predecessor. We attribute this to the chip's XFR2 functionality, along with more granular frequency/voltage settings.

Even when we hit it as hard as possible, the new CPU stays stable above 4 GHz.

Performance rises and power consumption falls (if only slightly). There's truth to AMD's marketing material, so says our lab equipment. Ryzen 7 2700X really does deserve attention for these results.

MORE: Best CPUs

MORE: All CPUs Content

Thermals And Noise

The Wraith Prism

Ryzen 7 2700X's Wraith Prism thermal solution is a large, high-finned cooler with four flattened heat pipes and a plate behind them for stabilization. The heat sink's entire contact surface is thus made of copper. Its fins are arranged in such a way that the exhaust air is focused toward the memory and I/O shield.

The fastening clamp is a big disadvantage of this large cooler, which takes us back to the old Athlon XP days. Even at maximum load on all cores in the stress test, the CPU only reaches a maximum temperature of 82.8°C (corrected value), so it remains below the thermal throttle threshold. The cooler handles the 105 watts easily. You can expect peaks up to 70°C and a little above, depending on the motherboard's predefined fan curve.

The cooler is loud and emits 44 dB(A) under load (50 cm distance, 45° diagonal) when the fan is spinning at 2600-2700 RPM. The fan can even be a bit noisy even when the system is idling on the Windows desktop. Unfortunately, the fan adjusts much too rapidly as the cooler reacts to short-term temperature jumps.

We see the result in the narrow-band frequency spectrum of the motor noise, which shifts back and forth between approx. 240 and 300 Hz. The fan generates almost 39 dB(A) at idle, which isn't necessary. It helps if adjust the fan curve to a fixed speed of at least 1400 RPM if the processor is under 60°C. However, you'll have to experiment because each case will require different settings.

AMD has made good progress with XFR2 and the powerful cooling finally pays off in terms of performance. The power consumption remains largely the same and you get a nice clock rate increase, but we don't like the unnecessary noise levels or the fiddly mounting mechanism.

MORE: Best CPUs

MORE: All CPUs Content

Final Analysis

AMD's 2000-series processors aren't revolutionary, but they are far beyond the normal evolutionary updates we've become accustomed to over the last several years.

In the chart below, we plotted gaming performance with both average frame rates and a geometric mean of the 99^th percentile frame times (a good indicator of smoothness), which we then converted into an FPS measurement. We're also presenting price-to-performance charts that get split up to include CPUs-only and extra platform costs. For the models that don't come with a bundled cooler, we add an extra $25 for a basic heat sink. We also add $20 if overclocking requires a more expensive motherboard (as is the case for Z370).

In gaming, AMD's stock Ryzen 7 2700X delivers a great performance boost that rivals its overclocked predecessor in every one of our tests. Tuning the 2700X provides additional performance, though you probably won't notice the difference. Check out our chart: as you can see, the Ryzen 7 2700X effectively ties Core i7-8700K based on the geometric mean. But it sells at a $30 discount, drops into a less expensive motherboard, and comes with a thermal solution that adds even more value.

While the overclocked Core i7-8700K is a fierce competitor, it requires you to buy a Z-series motherboard for overclocking, along with a capable cooler. Core i5-8600K offers most of the -8700K's performance, but you lose Ryzen 7 2700X's sixteen threads and bundled heat sink/fan. We think it's safe to say that AMD is delivering on its pledge to provide a near-equivalent gaming experience in most titles.

If you're searching for a more productivity-oriented processor, Ryzen 7 2700X is incredibly attractive. It offers superior performance compared to the Core i7-8700K in many of our threaded tests, and is much more competitive in lightly threaded applications than previous-gen models.

AMD's Precision Boost 2 and XFR2 algorithms are already pushing the voltage/frequency curve to its limits, so don't expect much in the way of overclocking headroom. We did tune Ryzen 7 2700X up to 4.2 GHz, but a higher dual-core Precision Boost 2 frequency of 4.3 GHz offers better performance than our all-core overclock in certain applications. Significant gains in games were likely a result of heightened sensitivity to our DDR4-3466 memory.

AMD's latest Ryzen 7 delivers a host of features that make enthusiasts swoon, such as an unlocked multiplier, backward compatibility with 300-series motherboards, solder between the heat spreader and die, and an LED-equipped cooler. We only wish that B450-based motherboards were available at launch time. Hopefully we hear more about AMD's lower-cost platform soon.

In a broader sense, AMD is delivering on its first update to the Ryzen processor series, proving that it can execute on its roadmap. It looks like it's going to be another busy year in the CPU space--and that's more good news for enthusiasts and gamers.

MORE: Best CPUs

MORE: All CPUs Content

How to Choose the Right CPU for 1080p Gaming: 14 Processors Compared

palcorn@outlook.com (Paul Alcorn) — Fri, 23 Mar 2018 18:30:00 +0000

Introduction

Last year, we saw innovation from AMD and Intel that we hadn't experienced in a long, long time. The Zen architecture made AMD competitive in segments of the CPU market it previously couldn't touch, and Intel moved as quickly as possible to defend its incumbent position. We thoroughly enjoyed the back-and-forth as both companies jockeyed for enthusiasts' adoration.

But even as new platforms were springing up with more PCIe connectivity than ever before, graphics-card availability dried up. Cryptocurrency miners bought up everything they could find to capitalize on rising valuations. Even today, you can't find modern models anywhere near their suggested retail pricing. We've resorted to buying pre-built systems and scouring the forums for previous-generation cards, trying to score a bargain.

The best spread of CPU technology in ages, paired with sky-high GPU prices, is a recipe for confusion for PC builders. For the same amount of money, enthusiasts can afford less graphics performance than they could not long ago. That makes it easy to overspend on host processing, since balance is thrown out of whack. But you can also get a lot more CPU for your dollar than this time a year ago. How do you make sure you're getting the most for your budget?

Well, we set out on a mission with 14 CPUs and three different GPUs to find the best combination in nine popular games.

Moving The Goalposts

For the last 11 years, Core i7 and Core i5 CPUs featured four cores. Coffee Lake changed this. Now, Core i3s sport four cores, Core i5s include six cores, and Core i7s boast six Hyper-Threaded cores. Intel also gave its low- and high-end models a makeover: Skylake-X stretches up to 18 cores/32 threads for high-end desktops, while Pentium processors have now gained Hyper-Threading technology.

Of course, AMD introduced its line-up of Ryzen 7, 5, and 3 models with copious core counts. Moreover, the Ryzen Threadripper series landed with up to 16 cores/32 threads and such friendly prices that Intel was forced to make its Skylake-X chips more affordable.

The Contenders

As you might imagine, the old rules of picking a CPU family to go with certain graphics cards changed as a result of these new processors. Thus, we decided to investigate using the best performers from each CPU class.

Representing AMD, we have the Ryzen 7, 5, and 3 models. We didn't bother testing last-generation Bulldozer-based CPUs, but we did throw in the Ryzen Threadripper 1950X for good measure.

For Intel, we have K-series Core i7, i5, and i3 CPUs from the Coffee Lake and Kaby Lake generations. We also added the Core i9-7900X and Core i9-7980XE to cover high-end desktops. Out of curiosity (or because we're gluttons for benchmarking punishment?), we couldn't help but include two of the latest Pentium processors, too.

That gives us 14 processors spread across five test platforms. We paired these with the GeForce GTX 1080, GTX 1070, and GTX 1060 (6GB) graphics cards. Although the GTX 1080 is considered an extravagance these days, we have to imagine it'll come down in price someday.

Finally, we selected nine games for testing. Some of the titles are new, while others are older. We did weigh the suite, though, more toward modern games. Some of them are CPU-dependent, others are decidedly graphics-bound, and a few are actually pretty well split down the middle. This allows us to explore bottlenecks from different angles.

Today's tests are all run at 1920x1080. (We have more data coming at 2560x1440 and 3840x2160, too, so expect a follow-up story or stories to present our findings there.) To best represent the experience we'd want to have, all benchmarks were run with the highest graphics settings possible.

Test Systems

To avoid variance from GPU Boost as our GeForce GTX graphics cards heat up, we use multiple runs from each benchmark in quick succession. We select the median value from the last recordings.

Test System & Configuration

Hardware

Intel LGA 1151 (Z370) @ 2666/2400AMD Socket AM4 @ 2667 Intel LGA 1151 (Z270) @ 2400AMD Socket SP3 (TR4) (In Game Mode) @ 2667 Intel LGA 2066 @ 2666 Common To AllGigabyte GeForce GTX 1060 G1 Gaming 6G 1TB 1500W (Creators Update Version 1703)

MORE: Best CPUs

MORE: All CPUs Content

Ashes Of The Singularity: Escalation

Nvidia GeForce GTX 1080

Ashes of the Singularity was one of the first titles to receive a Ryzen-optimized patch. And it appears that host processing is always going to be your bottleneck if you're using a GeForce GTX 1080 at 1920x1080. As we can see, the 18-core/36-thread Core i9-7980XE takes a healthy lead over the rest of the field, while the 10C/20T Core i9-7900X captures second place.

This benchmark scales well with increased core counts, highlighting some of the generational transitions in Intel's line-up. The 6-core/12-thread Core i7-8700K illustrates a big jump over Intel's previous-gen 4-core/8-thread Core i7-7700K, which basically ties the Core i5-8600K.

The Core i3-7350K (and every CPU below it in our chart) struggles mightily during the benchmark.

Nvidia GeForce GTX 1070

Stepping down a notch to the GeForce GTX 1070 finds Intel's Core i9-7980XE and Core i9-7900X providing nearly identical performance, while the rest of the field scales down as we'd expect.

Really, the Core i7-7700K and most processors under it offer nearly the same performance as they did paired up to a GeForce GTX 1080. We do measure a 2.3 FPS delta between the i5-8600K with both graphics cards, but that appears to be an exception. Such small variances hardly make it worth stepping up to the GeForce GTX 1080 in this title at this resolution.

Nvidia GeForce GTX 1060

The limitations of a GeForce GTX 1060 become clear, even in this CPU-bound test, though Intel's Core i9-7980XE still maintains a healthy lead over the rest of the field.

At this point, a graphics bottleneck is emerging, pushing the GTX 1060 to its limit. As a result, the Core i9-7900X and Core i7-8700K fall a few positions. It's possible that a lighter workload isn't pushing the execution cores hard enough for them to maintain their highest Turbo Boost bins. If true, the uncore frequency drops as well, slowing the cache and ring bus frequencies, reducing performance throughout. We'll see this phenomenon in other titles, too.

Compared to our GeForce GTX 1080 results, we see similar performance from every CPU beneath the Core i5-8600K, showing that lower-end processors are still the bottleneck.

MORE: Best CPUs

MORE: All CPUs Content

Civilization VI Graphics

Nvidia GeForce GTX 1080

Intel's Core i9-7900X leads this benchmark with an impressive 106.6 FPS. The remainder of the results are uneventful, generally following trends we've observed through past CPU reviews. The Core i7-7700K averages 102.8 FPS, and the nearest mainstream Ryzen competitor achieves 80.1 FPS.

We tested the Threadripper 1950X in Game Mode simply because there wasn't enough room in our charts for the many combinations that AMD enables. Head over to our Threadripper 1950X Game Mode, Benchmarked article if you're interested in the impact of the various modes on Threadripper's performance. Just be aware that the processor behaves differently in many games, depending on the settings you choose.

Nvidia GeForce GTX 1070

Stepping down to the GeForce GTX 1070 exposes a graphics bottleneck with some of our faster CPUs. Intel's Core i9 models push the limits at around 94 FPS, landing within the range we'd expect for a graphics-bound situation. The rest of the field falls into line based on host processing power, though some CPUs fare worse than others.

Nvidia GeForce GTX 1060

A more definite GPU limitation irons out some of the variance between CPUs. Worthwhile scaling is no longer observable. Case in point: There is only a 6.4% delta between the Ryzen 5 1600X and Threadripper 1950X, showing how a midrange graphics card can neutralize the benefit of a high-end processor if you aren't careful to maintain balance.

Fortunately, this also means you can snag an affordable Ryzen 3 or Core i3 and enjoy almost the same performance as you would with a much more expensive model.

MORE: Best CPUs

MORE: All CPUs Content

Dawn Of War III

Nvidia GeForce GTX 1080

Dawn of War III scales well with increased execution resources. As we can see from the fastest CPUs, though, it also favors high clock rates.

Intel's Core i7-8700K reaches up to 113.6 FPS. However, its 1.5 FPS advantage over the Core i7-7700K isn't very impressive considering the Core i7-8700K boasts 50% more Hyper-Threaded cores.

Meanwhile, the $1000 Core i9-7900X doesn't offer as much performance as a $175 Core i3-8350K. This is why we don't recommend HEDT processors for purpose-built gaming rigs.

Nvidia GeForce GTX 1070

The same general scaling trends persist with the GeForce GTX 1070, albeit at a lower performance level. Intel's Core i3-8350K continues behaving like older Core i5 CPUs, while the Core i5-8600K really hits a price/performance sweet spot with this pairing.

Nvidia GeForce GTX 1060

Again, some of the highest-end processors drop a few places because they aren't utilized fully. There's only a 3.3 FPS (4%) delta between the Core i9-7980XE and the first-place finisher...but that chart-topper is a Core i5!

Again: Striking the right balance is important. Not only is it a waste to buy too much CPU if you don't need the processing power for other workloads, but you can even handicap it with a graphics card incapable of keeping it busy.

MORE: Best CPUs

MORE: All CPUs Content

Far Cry Primal

Nvidia GeForce GTX 1080

Far Cry Primal's Dunia Engine 2 responds well to high clock rates and IPC throughput. However, it also benefits when we turn off simultaneous multi-threading. As a result, you'll notice that the Core i5s are very competitive.

An increasing number of games are being tuned for quad-core CPUs, and this one is no exception. We think it'll take some time before chips with more cores start distinguishing themselves, though Coffee Lake fares well enough.

Although AMD doesn't do as well in our average measurements, its Ryzen 5 1600X and Ryzen 7 1800X processors demonstrate the least amount of frame-time variance. Rest assured that they offer a smooth gaming experience.

Nvidia GeForce GTX 1070

We observe much less variance between CPUs as we step down to a less powerful GeForce GTX 1070. The 6-core/12-thread Ryzen 5 1600X continues outmaneuvering the 8-core/16-thread Ryzen 7 1800X, while the Threadripper 1950X inexplicably jumps higher in the rankings.

Nvidia GeForce GTX 1060

We enter graphics-bound territory with the GeForce GTX 1060, which mixes up the finishing order. Notably, even though they don't lead when it comes to average frame rate, Intel's Core i7-8700K and i5-8600K provide the smoothest performance with the least frame-time variance. The Pentium models also respond well to this pairing.

AMD's Threadripper 1950X is the only processor that lands below our expectations. Then again, we've proven that Game Mode isn't the best option for this title. A bit of tuning might rectify the issue.

MORE: Best CPUs

MORE: All CPUs Content

Grand Theft Auto V

Nvidia GeForce GTX 1080

Grand Theft Auto V favors Intel architectures and, more generally, multi-core designs with high clock rates. It may seem odd that the Core i5-8600K beats Intel's Core i7-8700K, but we've observed this tendency on other occasions.

Did you notice that Core i3-8350K out there in front of the quad-core Core i5-7600K? Clearly, Coffee Lake propels the Core i3 family to new performance heights.

Nvidia GeForce GTX 1070

The finishing order remains intact as we test with Nvidia's GeForce GTX 1070, though average frame rates fall a bit. Still, many of the slower models land within 1 FPS of what we saw from the GTX 1080, suggesting that we're not seeing a graphics-bound workload. Host processing appears to be the limiting factor.

Nvidia GeForce GTX 1060

Plain and simple: Nvidia's GeForce GTX 1060 restricts performance at this point.

MORE: Best CPUs

MORE: All CPUs Content

Hitman

Nvidia GeForce GTX 1080

We began our testing with the Pentium CPUs and worked our way up, benchmarking all nine games for each processor before moving on to the next. So, you can imagine how thrilled we were when we discovered that IO Interactive implemented a 90 FPS frame cap at some point through a patch. Soldiering on anyway, we observed little scaling above the Core i9-7900X.

Intel's Pentium processors average a solid 60+ FPS, though we could see quite a bit of hitching during the test sequence. The Ryzen 3 1300X and the Core i3-7350K fare better, but you want a Ryzen 5 or a Kaby Lake-based Core i5, at the least, for smooth performance in this title.

Nvidia GeForce GTX 1070

The frame cap continues to be a killjoy, but at least a GeForce GTX 1070 and a reasonably-priced Core i5-8600K push Hitman's limits at 1920x1080. There is a bit of unevenness to report from the Ryzen 3 1300X and Pentium models, which manifests in our frame-time variance chart.

Nvidia GeForce GTX 1060

We finally duck under the frame cap, but only because Nvidia's GeForce GTX 1060 isn't fast enough.

There's only a 3% delta between the quad-core Core i3-8350K and first-place finisher. This is definitely a graphics-bound workload that only breaks apart a bit when we drop to the lowest-end CPUs.

MORE: Best CPUs

MORE: All CPUs Content

Shadow Of War

Nvidia GeForce GTX 1080

Middle-earth: Shadow of War doesn't scale as dramatically as some of our other benchmarks, and it certainly isn't as sensitive to IPC throughput and frequency as Shadow of Mordor. CPU reviews tend to focus on games that scale well with certain host processing specifications, such as core count, clock rate, cache size, or memory bandwidth. But some games just can't get enough graphics performance.

This title exhibits a 12.6 FPS (~16%) gap between the fastest and slowest processors in our pool, but every system surpasses the 70 FPS mark easily. There is a difference in smoothness between the Pentium and Ryzen 3 1300X processors, but it isn't extreme.

Nvidia GeForce GTX 1070

Naturally, a GeForce GTX 1070 is even more tied up, tightening the delta from top to bottom to just 3.3 FPS.

The Pentium processors still struggle a bit with hitching.

Nvidia GeForce GTX 1060

A GeForce GTX 1060 puts all of these CPUs on an even footing. Subjectively, there is no difference between the processors, even though our data suggests that the Core i7-8700K lags behind.

MORE: Best CPUs

MORE: All CPUs Content

Project CARS 2

Nvidia GeForce GTX 1080

Project CARS 2 is purportedly optimized for threading, and it seems to scale well with increased host-processing resources. After all, we see a 41 FPS delta between the fastest and slowest processors in our test pool. The Core i5-8600K leads, and it is followed closely by Intel's Core i7-7700K.

Nvidia GeForce GTX 1070

The Core i3-8350K shows that it's a surprisingly capable quad-core CPU, besting AMD's entire line-up. With that said, the Ryzen 5 1600X specifically offers most of what higher-end models can do, but at a much lower price.

Nvidia GeForce GTX 1060

The GeForce GTX 1060 limits performance at 1920x1080. AMD's Ryzen 3 1300X is the only processor to exhibit a marked amount of hitching during our recording.

MORE: Best CPUs

MORE: All CPUs Content

PlayerUnknown’s Battlegrounds (PUBG)

Nvidia GeForce GTX 1080

PlayerUnknown's Battlegrounds can be a frustrating title to benchmark. Its constant cadence of updates seems to introduce more problems than are solved, and we've seen visible hitching from even the fastest CPUs and GPUs.

Aside from the Pentiums, scaling in PUBG is fairly meager. AMD's Ryzen 3 1300X suffers the most jarring stuttering during playback of our recorded game.

Nvidia GeForce GTX 1070

Stepping down to a GeForce GTX 1070 confirms that we're decidedly graphics-bound. Not that this game is particularly detailed; it may just be poorly optimized. Either way, there's no real meaningful conclusion we can draw from such an outcome.

Nvidia GeForce GTX 1060

Our final test leaves little to explain: There are no meaningful differences between CPUs when you're constrained by a GeForce GTX 1060.

MORE: Best CPUs

MORE: All CPUs Content

The Best Combinations of CPU and GPU at 1080p

Boiling our testing down into one general recommendation is almost impossible, especially given the pricing of graphics cards right now. But we can identify some of the more obvious trends.

Fast-paced first-person shooters tend to prefer high clock rates and fast graphics cards. Nimble CPUs like the Intel Core i5 and i7, and AMD's Ryzen 5 and 7, fare best in our performance measurements. Some titles are easily graphics-bound, though that could be due to poor optimizations or minimal effort to parallelize the game engine. PlayerUnknown’s Battlegrounds is the perfect example. It doesn't respond at all to faster processors. It's far better to buy a higher-end graphics card if that game is your main squeeze.

Real-time strategy titles seem to respond more favorably to CPUs with lots of cores; so do many turn-based games. Of course, that genre isn't as sensitive to graphics performance because it doesn't require quick response times. But a fluid experience in RTS games is a must. Often, host processing is the bottleneck, so dialing back your GPU investment and buying a better CPU yields the best performance possible.

Charting Out The Trends: The Best Pairings

We typically plot our test results on a scatter chart with price and performance on two axes, but we aren't even going to take a stab at graphics pricing during these tumultuous times. Instead, we're providing a geometric mean of the 99th percentile frame times, representing smoothness, converted into an FPS measurement for each class of graphics card. Each game factors into this calculation except for the artificially hamstrung Hitman.

Balance is the goal here. Intel's Core i7-8700K is the first chip we think of when it comes to top-of-the-line for gaming PCs. But it doesn't always lead in our benchmarks. It would likely fare better with more heavily threaded titles tested. Otherwise, the Core i5-8600K seems to be a better value pairing for Nvidia's GeForce GTX 1080.

The GeForce GTX 1070 pairs well with either a Core i5-8600K or Ryzen 7 1800X, though the latter notably offers additional performance in other enthusiast-oriented desktop apps. We've also found that the Ryzen 7 1800X is a great chip for game streaming, so give it serious consideration for more taxing workloads. If you're looking to go Intel, aren't interested in overclocking, and want to save a few dollars, the Core i5-8400 complements a GeForce GTX 1070 well, too.

On the value scale, Core i3-8350K and Ryzen 5 1600X both pair well with a GeForce GTX 1060. If you're willing to do a bit of overclocking, though, a Ryzen 5 1600 looks even better.

Speaking of overclocking, most of the CPUs we tested can operate at even higher frequencies. AMD has a much friendlier policy here; all of its chips are unlocked. Meanwhile, Intel charges a premium for its unlocked K-series SKUs. If you want to know more about available headroom, we have plenty of test results with every processor overclocked in their individual reviews, so head to those for an idea of potential gains.

Of course, our findings would be all the more interesting if we could run out and build a well-balanced machine with affordable hardware. Here's hoping the availability of graphics cards improves sometime soon!

MORE: Best CPUs

MORE: All CPUs Content

Do The Meltdown and Spectre Patches Affect PC Gaming Performance? 10 CPUs Tested

palcorn@outlook.com (Paul Alcorn) — Wed, 07 Feb 2018 14:00:01 +0000

Gaming With Meltdown And Spectre

Imagine that nearly every PC, server, and mobile phone on the planet was suddenly vulnerable to data theft at the hands of nefarious actors. Then, imagine the exploit responsible for this vulnerability couldn't be detected by antivirus software because it merely took advantage of normal CPU operations.

Although this sounds like a bad movie script, it unfortunately became a known truth during the first few days of 2018.

On January 2nd, The Register exposed Intel's then-secret Meltdown and Spectre vulnerabilities through investigative journalism. But unbeknown to most of the world, Google’s Project Zero researchers, along with two other independent teams, discovered the vulnerabilities 200 days earlier. As a courtesy, the researchers gave Intel, AMD, IBM, Qualcomm, and ARM a grace period to develop mitigations before making their findings public. Those companies, along with developers contributing to Windows and Linux, worked together behind a veil of secrecy for months.

Perhaps understandably, the initial response to The Register's report was chaotic because it preempted that planned group disclosure. What happened after, though, appeared to be a comedy of errors, especially given the amount of time affected companies had to prepare.

Then again, the firms were trying to plug holes that were baked into hardware and software for more than a decade. In fact, nearly every Intel processor since 1995 was found to be vulnerable, so fixing the issues without breaking compatibility proved to be a mind-boggling challenge.

The patches supposedly have performance implications. So now that the industry is a month into cleaning up its mess, we're ready to start assessing the damage. First up: game performance.

The Land Of Patch Confusion

There are two general vulnerabilities in play here, and they're broken up into three categories. Variants 1 and 2 are what we've come to know as Spectre, while Variant 3 is Meltdown. Intel, ARM, and Qualcomm are susceptible to all three, while AMD is only affected by Spectre.

As we can see, Variant 1 and 3 can be patched in the operating system, while the most nefarious bug, Variant 2, requires both motherboard firmware/microcode and operating system patches.

The initial industry scramble resulted in a flurry of immature and buggy updates. Because the patches weren't distributed as drivers, processor vendors couldn't push them out directly. Instead, they filtered out through Microsoft, Linux-based operating systems, OEMs, and motherboard manufacturers. A rapid sequence of patches, re-patches, and un-patches confused enthusiasts in the know. Everyone else had to have been completely lost.

This is how bad it got: Intel released a motherboard firmware/CPU microcode patch that could cause reboots, system instability, and potential data loss/corruption. Its partners pulled the update. Microsoft published a patch of its own for AMD systems that left some of them unbootable. It, too, had to reverse course, blaming improper documentation from AMD. A fix was released several weeks later.

For now, Intel doesn't have an operating system or microcode patch for Spectre Variant 2. AMD has an OS patch for it, but the company does not have a microcode update to offer. And because microcode patches will have the biggest impact on system performance, today's benchmark results are subject to change.

Intel says it will provide patches for CPUs dating back five years and then move on to older models. Many folks speculate that we may never see patches for those legacy products, though. Both Intel and AMD claim they will have silicon-based mitigations in their next-gen processors. Of course, it remains to be seen how each company works around their security holes without compromising performance.

For now, one thing is for sure: today's patches, particularly those for Spectre Variant 2, affect performance in some workloads. Older CPUs are said to be hit the worst. Microsoft predicts that "some" users with Windows 10 on pre-Broadwell architectures will suffer noticeable slow-downs, while "most" users on Windows 7 and 8.1 on comparable systems will notice a decrease in performance.

Measuring the impact hasn’t been an easy task in our labs. The changing nature of these patches complicates matters: we’ve begun testing several times only to have a patch altered or removed. We’re diligently working on the next round of application benchmarks, and are expanding our scope to include older CPUs. For now, let's focus on gaming with a good selection of recent Ryzen, Kaby Lake-, and Coffee Lake-based processors.

MORE: CPU Security Flaw: All You Need To Know About Spectre

MORE: Best Gaming CPUs

Exploiting The Unexploitable

Meltdown and Spectre are vulnerabilities exploited through side-channel attacks, which are incredibly difficult to block. These occur when an attacker observes the traits of a computer, be it the timing of certain operations or even noise and light patterns, and uses that information to compromise security.

It all starts with data, of course. A CPU loads information from main memory into its registers by requesting the contents of a virtual address, which is in turn mapped to a physical address. While fulfilling this request, the CPU verifies the address' permission bits, indicating whether the process has permission to access the memory address, or if only the kernel can access it. The system approves or denies access accordingly. The industry assumed this technique securely bifurcated memory into protected regions, so operating systems automatically map the entire kernel into the user space memory map's virtual address space. That means the CPU can attempt to access all of the virtual addresses if needed, but it also exposes the entire virtual address map to the user space.

The problem has to do with speculative execution, which is part of out-of-order processing. Pipelined CPU cores process instructions in stages, such as instruction fetch, instruction decode, execute, memory access, and register write-back. Today’s processors break each of these fundamental stages down further, sometimes into 20 or more stages. This helps facilitate higher clock rates.

CPUs employ multiple pipelines to allow for parallelized instruction processing. That's why, in the image below, we see four different colors passing simultaneously through the four stages. Instruction branches can cause the pipeline to switch to another instruction sequence, creating a stall. That means the pipeline doesn’t process data for several clock cycles while it waits for inputs from memory.

To help avoid this, when the processor encounters a branch, its prediction unit tries to guess which instruction sequence the processor will need next. But it makes that determination before processing the instruction. The processor then fetches the predicted branch's instruction and speculatively executes it. This avoids the latency normally incurred if the pipeline stalls (due to the branch), and then fetching the next instruction from memory.

Depending on the address location (L1, L2, L3, RAM), fetching data from memory requires 10s to 100s of nanoseconds. That's slow compared to the sub-nanosecond latency of processor cycles, so having the instruction already in-flight speeds operation tremendously. Most operations are committed because branch predictors often have a 90%+ success rate. The processor simply discards the instruction and flushes its pipeline if the instruction isn't needed (misprediction).

Google's researchers found a sliver of opportunity in how the system handles cached memory access during speculative executions. Normal security checks that keep the user space and kernel memory separate don't happen fast enough during speculative executions that access cached memory. As a result, the processor can momentarily speculatively fetch and execute data from cached memory that it shouldn't be able to access. The system eventually does deny access and the data is discarded, but again, because this doesn't happen fast enough, a window of opportunity for an exploit is opened. Meltdown exploits the branch predictor to run code against the cache, all the while timing this process. That allows it to determine what data is held in the memory, as seen in the short video below.

In the case of Meltdown, an attacker can read passwords, encryption keys, or other data from protected memory. That data could also be used to assume control of the system, rendering all other forms of protection useless. The biggest concern for data centers is that the exploit also allows an application resident in one virtual machine to access the memory of another virtual machine. This means an attacker could rent an instance on a public cloud and collect information from other VMs on the same server.

Attackers can exploit this vulnerability with JavaScript, so you could expose your system simply by visiting a nefarious web site. Browser developers have updated their products to reduce timing granularity. But attackers can also just use normal code delivered via malware or other avenues to execute an attack.

For now, the Meltdown operating system patch involves adding another layer of security checks during memory address accesses. That hampers the latency of system calls, thus slowing performance when applications issue kernel calls. Applications that tend to remain in the user space are less impacted. Intel's post-Broadwell processors have a PCID (Post-Context Identifiers) feature that speeds the process, so they don't experience as much of a slowdown as older models.

The Spectre exploit is much more nefarious because it can exploit a wider range of the speculative execution engine's capabilities to access kernel memory or data from other applications. Some researchers claim that fixing this exploit fully could require a fundamental re-tooling of all processor architectures, so it's possible we'll live with some form of this vulnerability for the foreseeable future. Fortunately, the exploit is extremely hard to pull off, requiring an elevated level of knowledge of the target processor and application. Intel and others have come up with patches for the current Spectre variants, but it is possible that vendors will play whack-a-mole as new Spectre derivatives arrive in the future.

MORE: CPU Security Flaw: All You Need To Know About Spectre

MORE: Best Gaming CPUs

Test Setup

Our methodology for this story is fairly simple in practice. First, we performed a round of benchmarks on our Windows 10-based systems. Then, we applied the relevant patch and repeated the tests. Although the Variant 2 operating system patch we're using was deactivated via Windows, it should represent worst-case performance for now. We did not use a BIOS with Spectre mitigations because none are currently available (they were pulled). We'll follow up with more comparative testing as firmware updates arrive.

We tested each processor at its stock settings, including representatives from the Core i3, Ryzen 3, Core i5, Ryzen 5, Core i7, Ryzen 7, and Pentium line-ups. Naturally, overclocked systems will fare better. And as always, we disable Enhanced Multi-Core Turbo on our test systems.

In order to avoid variance from GPU Boost as our GeForce GTX 1080 heats up, we use multiple runs from each benchmark in quick succession. We select the median value from the last recordings, so, in many cases, the patched/unpatched results could land higher or lower than each other due to the tight variances you'll see today.

Test System & Configuration

Hardware

Intel LGA 1151 (Z370) @ 2666AMD Socket AM4 @ 2667 Intel LGA 1151 (Z270) @ 2666 All 1TB 1500WWindows 10 Creators Update Version 1703, pre- and post-patch

MORE: CPU Security Flaw: All You Need To Know About Spectre

MORE: Best Gaming CPUs

VRMark, 3DMark & AotS: Escalation

Comparison CPUs

AMD Ryzen 7 1800X

AMD Ryzen 5 1600X

AMD Ryzen 3 1300X

VRMark & 3DMark

We aren't big fans of using synthetic benchmarks to represent real-world game performance, but 3DMark's DX11 and DX12 CPU tests provide useful insight into the amount of horsepower available to game engines.

Futuremark's VRMark test lets you gauge your system's suitability for use with the HTC Vive or Oculus Rift, even if you don't currently own an HMD. The Orange Room test is based on the suggested system requirements for current-generation HTC Vive and Oculus Rift HMDs. Futuremark defines a passing score as anything above 109 FPS.

Futuremark specifies a 3% maximum variance between runs if we adhere to normal testing best practices, such as ensuring similar configurations and drivers between systems, allowing the PC to enter an idle state, and conducting tests with the same environmental factors. Of course, there are sub-tests in suites like PCMark 10 that can fall outside the 3% window, at which point we rely upon averages. We didn't encounter those issue with these gaming-focused tests, though.

VRMark reflects a negligible impact from the patches across our entire field of CPUs. At times, the patched operating systems are slightly faster, but these values land within the 3% margin of error. That means those higher results are merely a byproduct of normal run-to-run variability.

The DX11 and DX12 CPU benchmarks respond to core counts and increased parallelism. Frankly, these tests don't show us much of interest. There is some jockeying between patched and unpatched configurations, but the results are uneventful overall.

Ashes of the Singularity: Escalation

Ashes of the Singularity: Escalation scales nicely with the addition of computational horsepower, so graphics aren't limiting us.

We observe minor variations between the patched and unpatched systems. For instance, a patched Core i7-8700K outpaces the vulnerable configuration. But a 1.6% delta is still within our expectations of normal run-to-run variance. We also don't record any extreme changes to the frame time or frame time variance metrics.

Notice that we split our results into two classes due to the sheer number of tests we ran. Be sure to scroll across the album for our comparison of Core i3, Ryzen 3, and Pentium processors.

MORE: CPU Security Flaw: All You Need To Know About Spectre

MORE: Best Gaming CPUs

Civilization VI Graphics & AI, & Dawn of War III

Civilization VI AI Test

The Civilization VI AI test measures CPU performance in a turn-based strategy game, and tends to favor a mixture of physical cores, clock rate, and IPC throughput.

Our results are a mixed bag; half of the processors are slower after the patch, but five turn out to be faster. Our largest variance is 0.15 seconds, though, so its hard to ascribe any significance to the outcome.

Civilization VI Graphics Test

The results of our Civilization VI graphics test aren't any more surprising than what we saw in Ashes of the Singularity. There's little to no observable impact from the patches, and both configurations take turns in the lead. Ryzen 7 1800X encounters the most variance, and that's a mere .8 FPS on average (less than 1%). Even the 99th percentile metrics bounce back and forth between the patched and unpatched configurations.

Warhammer 40,000: Dawn of War III

Dawn of War doesn't seem to be hampered by the patches either, although we do see our first result exceeding a 1 FPS delta. The patched Ryzen 7 1800X falls behind by 1.1 FPS, yielding a mere 1.2% difference. Again, it's a relatively mundane outcome.

MORE: CPU Security Flaw: All You Need To Know About Spectre

MORE: Best Gaming CPUs

Far Cry Primal, Grand Theft Auto V & Hitman

Far Cry Primal

Far Cry Primal's Dunia Engine 2 responds well to high clock rates and IPC throughput. However, it also seems to prefer when we turn off simultaneous multi-threading, so you'll notice the Core i5 processors are very competitive.

We do see a bit more frame time variance between certain models, but the lead again shuffles back and forth between patched and unpatched configurations. That means we can chalk these differences up to expected variances.

Typically, our reviews include similarly-priced models, so the floors and ceilings aren't so pronounced. Today's exploration gives us a good opportunity to see how different CPU classes scale. For instance, the average frame rate delta between Intel's Pentium G4560 and Core i7-7700K is 27 FPS.

Grand Theft Auto V

The Grand Theft Auto V benchmark again reveals little variance between configurations, and a negligible impact on frame times.

We do spot several notable outliers from Intel's Core i3-8350K, but those occur whether the CPU is patched or not. We also record a 42 FPS 99th percentile measurement with the patched Core i3-7350K compared to the unpatched configuration's 47 FPS. That's one of the largest deltas reported in today's data.

Hitman (2016)

We're graphics-bound in our Hitman test, so there is little to no variation between the Core i7 and i5 processors.

MORE: CPU Security Flaw: All You Need To Know About Spectre

MORE: Best Gaming CPUs

Shadow of War, Project CARS 2 & PUBG

Middle-earth: Shadow of War

We swapped out the aging Middle-earth: Shadow of Mordor for Shadow of War during this round of tests. We like that its built-in benchmark is very consistent.

Shadow of War isn't as sensitive to clock rate and IPC throughput as its predecessor. We don't see much variance between the Core i5 and i7 models, which are largely graphics-bound. There is, however, more distance between the lower-end models. But there is only a 13.2 FPS delta between the Core i7-8600K and Pentium G4620.

Project CARS 2

We're also swapping out Project CARS for Project CARS 2.

There's a bit more scaling to observe in this title, with the slowest processor lagging the leader by 41 FPS. Regardless, though, there isn't any significant variance attributable to security patches (aside from a 1.7 FPS delta between the Core i3-8350K results).

PlayerUnknown's Battlegrounds (PUBG)

This is also our first outing with PlayerUnknown's Battlegrounds.

Although we're using a saved game sequence to ensure repeatability, we recently learned that PUBG recordings can lag in-game multiplayer performance by ~4-8 FPS. This is an odd tendency because recorded game sequences usually run faster than actual multiplayer matches. Nevertheless, we recorded a few sessions with our tool and then compared them to in-game results using the same path. Sure enough, the FPS line charts didn't overlap, even though the results indicate similar peaks and valleys. That means the recording is an accurate representation of trends, albeit with slower frame rates.

That doesn't mean the results are any more interesting than they have been, though. The game engine simply doesn't scale well; we only recorded a 5.9 FPS delta between the fastest and slowest processors in our test pool. We probably could amplify this difference by reducing graphics quality. However, we prefer setting the eye candy to maximum for a more enjoyable experience, rather than deliberately creating a synthetic metric.

MORE: CPU Security Flaw: All You Need To Know About Spectre

MORE: Best Gaming CPUs

Game On...For Now

This first round of Meltdown and Spectre patch testing proved fairly uneventful. There really wasn't much to report across our suite of game benchmarks. But that's partly because the status of patches keeps changing. We originally planned to test with both the operating system and microcode updates for Spectre Variant 2, which are expected to impose a significant performance overhead.

Unfortunately, Intel and its partners pulled the microcode patch during our testing, and AMD still doesn't have a fix of its own. We've been told that Intel's updated update is undergoing rigorous validation, but we don't have a time frame for its release. The same goes for AMD's Variant 2 microcode. Of course, we'll add test results with the new patches once they are available.

For newer processors, it looks like the operating system patch won't affect gaming workloads much, if at all. Most games are confined to the user space and don't make frequent kernel calls, so it's possible that the impact on older CPUs could be minor as well (game testing on those is in-progress).

The current patch does have an impact on storage performance, at least when it's measured with synthetic benchmarks. A laggy hard drive would obviously affect level loading times and the storage subsystem's ability to feed the game engine, possibly resulting in choppy scene transitions. We scrutinized our load times and cut scenes closely, and while entry-level CPUs did take longer and were less smooth, it's hard to chalk that up to a security patch because slower processors are, well, slower. We didn't notice any dramatic changes in performance consistency or frame time variance, so any minor impact would likely be limited to storage-imposed symptoms, at least with the patches as they sit currently.

Most of the vulnerability-oriented storage testing we've seen is happening at high queue depths, or using pure read or write workloads that aren't the best indicator of operating system performance. Most real-world accesses occur at lower queue depths, and radical changes in SSD performance, either for better or worse, don't correspond linearly to application performance.

The Spectre Variant 2 patches still loom large for Intel and AMD. Hopefully, both companies can deliver solid updates with minimal impact. We've heard that some applications can be optimized to minimize overhead. And Intel has mentioned that existing patches will mature into more efficient implementations.

For now we remain vulnerable to Spectre Variant 2, and proof-of-concept code is already popping up in the wild, unfortunately. That leaves us exposed while we wait on Intel and AMD, not to mention the rest of the industry, to correctly patch what could be the greatest vulnerability of our time. At least we can enjoy some gaming while we wait.

MORE: CPU Security Flaw: All You Need To Know About Spectre

MORE: Best Gaming CPUs

Gaming And Streaming: Which CPU Is Best For Both?

palcorn@outlook.com (Paul Alcorn) — Wed, 03 Jan 2018 14:00:00 +0000

Introduction

Game streaming is immensely popular right now. And with more than 2 million broadcasters serving approximately 100 million viewers a month, Twitch is unquestionably the leading platform for gamers looking to show off their skills.

Streaming is pretty easy to set up once you decide what you'll use to encode your content. There are several options, each with its own strengths and weaknesses. Hardware-accelerated encoders, powered by Nvidia's NVEnc, AMD's Video Coding Engine, and Intel's QuickSync, all promise to wrap your gaming experience up in a good-looking and easy-to-stream package without hammering your CPU. But it's generally agreed that these encoders sacrifice quality and flexibility compared to software-based alternatives.

Conversely, software encoding with the x264 library is easy alongside utilities like Open Broadcaster Software (OBS) and XSplit. It offers better stream quality than GPU-driven approaches, but at similar bit rates. That's important for services with defined bit rate limitations (like Twitch). Streaming can also chip away at bandwidth caps. Just one hour of streaming at 10 Mb/s equals 4.5GB of data, so you want the most efficient encoder possible.

A host processing-based approach offers more flexible configuration options. Better quality comes at a price, though. Software encoding is a CPU-intensive process, which obviously tugs the other way against our usual aim of running games at the highest frame rates possible.

Hitching, stuttering, lag, and generally bad performance curse gamers who try doing too much with limited software encoding resources. Thus, it's often recommended to use a secondary system for high-quality streaming. In fact, that's the best way to achieve an unencumbered gaming experience while also providing your audience a crisp video stream. That requires attaching your gaming rig to another computer via capture card or local area network to offload the encoding workload, though. Many casual gamers simply don't have the money for such a configuration.

Software encoding can have a huge impact on gaming performance. Until recently, if you wanted to stream and game using the highest-quality solutions on one PC, you almost assuredly had to buy a pricey CPU with lots of threads to throw at the problem. But now those high-end desktop processors are more accessible than ever. AMD's Ryzen 7 chips lowered the bar for getting lots of cores into affordable platforms, and most games can't fully utilize them. That leaves lots of horsepower leftover for streaming at a reasonable price.

Intel's Coffee Lake processors recently debuted with more cores as well, giving those Ryzen CPUs a run for their money in multi-tasked, multi-threaded workloads. Good thing, too. As we'll see, the previous-gen Kaby Lake architecture simply cannot compete against like-priced AMD alternatives. Today's story compares a number of different host processors in combined gaming/streaming workloads. Along the way, we developed test methodology you'll see used in our future CPU reviews.

MORE: Best CPUs

MORE: All CPUs Content

Test Setup

How We Tested

Repeatability is one of the most important components of any useful benchmark methodology. All tests have some degree of uncertainty, but we're looking for a minimal and consistent amount of variability. Results plagued by wild swings in performance from one run to the next aren't usable as accurate benchmarks.

As an example, we've yet to develop any reliable multi-tasking benchmarks. In response to reader requests, we have worked diligently to create a series of tests that measure gaming performance with background applications like Web browsers, email clients, media players, Discord, and Skype open. Windows' prioritization appears to be based on fickle and unexplained factors. The operating system suspends various background processes unpredictably during one scripted sequence, then leaves them fully active during the next (even when the test environment hasn't changed). This unpredictability becomes more, well, unpredictable, as the number of open applications increases. Switching Windows into Game Mode only complicates matters further. So far, we have no solution. Our multi-tasking experiments yield deltas from 5 to 15 FPS between successive runs, which means they land nowhere near our expectations for a reliable benchmark.

Luckily, game streaming is much easier to control. Encoding is a CPU-intensive task that chews up plenty of cycles, so Windows doesn't suspend or otherwise interfere with it. This allows us to create repeatable benchmarks without extreme outliers.

What We're Measuring

Evaluating game streaming performance works across two axes: game quality and stream quality. Of course, we'll measure average, minimum, and 99th percentile frame rates with and without streaming in the background. We'll also include our usual frame time and variance results, which become more important once we start streaming.

We also need to account for stream quality. That means recording the percentage of frames encoded. Each processor pushes different frame rates, so each run correspondingly generates a different number of frames. As such, we measure the percentage of frames successfully encoded as "% of Frames Delivered." In the test below, a Threadripper 1950X CPU encoded 98.9% of the frames generated by our gaming session, meaning it skipped 1.1% of the frames due to encoding lag.

We're streaming at 60 FPS, so we also measure stream quality by listing the percentage of frames encoded within the desirable 16.667ms (60 FPS) threshold. We also include the percentage of frames that land above and below the 60 FPS threshold, which helps quantify the hitching and stuttering a viewer would see on the stream. Subjective visual measurements are still important, so we'll call out tests that generate a bad-looking stream.

Open Broadcaster System

There are several software encoding applications, but we chose Open Broadcasting System (OBS) due to its flexible tuning options, detailed output logs, and broad compatibility with streaming services. We're using the x264 software encoder, along with YouTube Gaming for our streaming service. Any run that reports frames dropped due to networking interference is discarded.

Our ultimate goal is to develop a test that measures CPU performance, so we select parameters that remove the most obvious bottlenecks. Gaming at 1920x1080 with an EVGA GeForce GTX 1080 FE side-steps a GPU limitation (as much as possible). Encoding overhead isn't as high with lesser video cards that generate fewer frames per second. We also test with a 10 Mb/s upload rate, though you can stream at 6 Mb/s or less. Our Internet connection would accommodate up to 35 Mb/s uploads. To vary game selection, we chose Grand Theft Auto V, Middle-earth: Shadow of War, and Battlefield 1 for our tests.

There are several other scenarios we could have added to increase the complexity of our testing, such as a simultaneous video stream from a webcam, recording the game to the host system, or streaming to multiple services at once. We went with just one service to reduce the number of variables...at least for now.

Finding the best streaming options requires some tuning for every game and hardware configuration. There is a delicate balance between game performance on the host system and stream quality for the remote viewer, so fine-tuning is needed to yield the best mix. We picked somewhat general settings that offered a good range of performance by our subjective measure. We also stuck with options that'd establish a level playing field for a wide range of test systems. Just be aware that there are plenty of knobs to turn, some of which could offer better performance than the ones we use (lowering the stream to 30 FPS, for instance, cuts encoding overhead significantly)

Tuning the encoding presets is one of the most direct ways to adjust streaming performance and quality for your system's capabilities. Slower encoding increases compression efficiency, which provides better output quality and reduces compression artifacts. OBS has 10 presets ranging from "ultrafast" (the lowest-quality setting with the least computational overhead) to "placebo" (offering the best streaming quality and consuming the most host processing resources). The placebo setting is aptly named; there is certainly a rapidly diminishing rate of return on stream quality after passing the "slower" preset (two ticks before placebo). More strenuous settings can quickly cripple even powerful processors, particularly if you are streaming from a single host system. Placebo with care.

We split our test groups into three different classes. After evaluating a few Core i3- and Ryzen 3-class processors and determining that they can't stream effectively at our settings, we chose Ryzen 5 and Core i5 models for our entry-level systems. We used the "veryfast" encoding setting for this class of CPU. Naturally, higher-end processors, such as our Ryzen 7/Core i7 and Threadripper/Core i9 chips, offer more performance, so we use the "faster" and "fast" settings, respectively, for brawnier CPUs.

Because we're testing with different encoding presets, you cannot compare test results for the different classes directly.

Test System & Configuration
Hardware	Intel LGA 1151 (Z370)Intel Core i5-8600K, Core i7-8700KMSI Z370 Gaming Pro Carbon AC2x 8GB G.Skill RipJaws V DDR4-3200 @ 2666 and 3200 MT/sAMD Socket AM4AMD Ryzen 5 1600X, Ryzen 7 1800XMSI Z370 Xpower Gaming Titanium2x 8GB G.Skill RipJaws V DDR4-3200 @ 2667 and 3200 MT/sIntel LGA 1151 (Z270)Intel Core i5-7600K, Core i7-7700K MSI Z270 Gaming M72x 8GB G.Skill RipJaws V DDR4-3200 @ 2666 and 3200 MT/sAMD Socket SP3 (TR4)AMD Ryzen Threadripper 1950XAsus X399 ROG Zenith Extreme4x 8GB G.Skill Ripjaws V DDR4-3200 @ 2666 and 3200 MT/sIntel LGA 2066Intel Core i9-7900X, Core i9-7980XEMSI X299 Gaming Pro Carbon AC4x 8GB G.Skill Ripjaws V DDR4-3200 @ 2666 and 3200 MT/sAll EVGA GeForce GTX 1080 FE 1TB Samsung PM863 SilverStone ST1500-TI, 1500W Windows 10 Creators Update Version 1703Corsair H115i

Test System & Configuration

Hardware

Intel LGA 1151 (Z370)Intel Core i5-8600K, Core i7-8700KMSI Z370 Gaming Pro Carbon AC2x 8GB G.Skill RipJaws V DDR4-3200 @ 2666 and 3200 MT/sAMD Socket AM4AMD Ryzen 5 1600X, Ryzen 7 1800XMSI Z370 Xpower Gaming Titanium2x 8GB G.Skill RipJaws V DDR4-3200 @ 2667 and 3200 MT/sIntel LGA 1151 (Z270)Intel Core i5-7600K, Core i7-7700K MSI Z270 Gaming M72x 8GB G.Skill RipJaws V DDR4-3200 @ 2666 and 3200 MT/sAMD Socket SP3 (TR4)AMD Ryzen Threadripper 1950XAsus X399 ROG Zenith Extreme4x 8GB G.Skill Ripjaws V DDR4-3200 @ 2666 and 3200 MT/sIntel LGA 2066Intel Core i9-7900X, Core i9-7980XEMSI X299 Gaming Pro Carbon AC4x 8GB G.Skill Ripjaws V DDR4-3200 @ 2666 and 3200 MT/sAll EVGA GeForce GTX 1080 FE 1TB Samsung PM863 SilverStone ST1500-TI, 1500W Windows 10 Creators Update Version 1703Corsair H115i

MORE: Best CPUs

MORE: All CPUs Content

Intel Core i5 & AMD Ryzen 5

We generated our stream using the "veryfast" encoder preset for this class of processors.

AMD Ryzen 5 1600X

Intel Core i5-8600K

Intel Core i5-7600K

Battlefield 1

First, we run a set of baseline tests to gauge performance without an active YouTube stream. Intel's Core i5-8600K fares best, but the Core i5-7600K and Ryzen 5 1600X aren't far behind.

The story changes when we add streaming to the mix. All three CPUs lose varying amounts of performance, with Ryzen 5 1600X ending up on top (though not by much).

We have two entries in our charts for the Kaby Lake-based Core i5-7600K: normal and high-priority. First, let's talk about the normal results. Intel's 4C/4T architecture hamstrings the Core i5-7600K, and it loses nearly half of its average frame rate. Unsurprisingly, this has an impact on in-game frame time variance and unevenness measurements. Those numbers don't tell the entire story, though. Battlefield 1 is basically unplayable due to hitching and stuttering during our stream.

The stream quality chart shows that Core i5-7600K only delivers 23.6% of its frames during the test, which means it drops an astounding 76.4% of the frames. This also results in a completely unwatchable stream. Flip over to the CPU utilization charts; they tell a profound story. In short, Core i5-7600K doesn't have any headroom available to handle the parallelized encoding workload throughout most of our benchmark.

The OBS software has several different settings to optimize performance, and adjusting process priority is a common tactic to improve streaming on Kaby Lake-based processors. So, we selected the high-priority setting (marked HP Stream in the charts). Effectively, this allows the encoder to steal cycles from the game engine, resulting in lower frame rates. But the encoder successfully processes 100% of the frames. Our Core i5-7600K fell to 51.6 FPS, and its 99th percentile measurements nosedived. On the flip side, less performance means fewer frames to encode per second, which in turn boosts encoding efficiency. Surprisingly, 91.05% of the frames landed within the desirable 16.667ms range.

We could always dial-back the encoder preset to accommodate Intel's lackluster -7600K, but similarly-priced processors in the test pool handle these quality settings adeptly. While streaming, Core i5-8600K provides nearly the same in-game average frame rate as AMD's Ryzen 5 1600X and it encodes 100% of the frames. The -8600K's streaming quality isn't as impressive as Ryzen's: 89.44% of frames land within the desirable range and ~10% of frames fall either above or below the threshold. In our opinion, the stream looked fine, but wasn't as smooth as the Ryzen 5 1600X's output.

A glance at the 99th percentile FPS chart tells us that Ryzen 5 1600X delivers the smoothest gaming experience during our streaming session. It also has plenty of reserves left in the tank, even during the streaming workload. In other words, you could probably increase the encoding preset to a higher level than the competing processors.

AMD's Ryzen CPUs don't overclock as well as Core processors, though. We tuned each model to see if the Intel's available headroom turned the tables. The x264 encoder uses AVX instructions extensively, so it serves as a nice stress test for overclocking stability. No doubt, Coffee Lake's AVX offset could come in handy, though it might also result in lower frequencies during streaming if you activate the feature (or leave it on Auto in the UEFI).

Overclocking nudges the Core i5-7600K ahead of Ryzen 5 1600X during our baseline gaming test, but it does little to rectify the issues encountered during our streaming workload. Four threads just can't hack it. Streaming suffers even after adjusting the priority status. The -7600K does serve up a higher percentage of frames with the overclocked settings and normal prioritization, but it still costs the chip game performance. Priority adjustments have little bearing on gaming smoothness: the -7600K's 99th percentile measurements are terrible with both settings.

A 4.9 GHz overclock does boost the Coffee Lake-based Core i5-8600K's 99th percentile measurements. However, it still trails Ryzen 5 1600X. The tuned -8600K is certainly competitive thanks to a compelling gaming experience, but it can't match the Ryzen 5 1600X's streaming performance. It's apparent the encoder appreciates lots of threads, so Ryzen's 12 are a big advantage.

The frame time chart summarizes in-game performance well. Kaby Lake-based processors scribble their way across with extreme variance. The Core i5-8600K is also easy to spot during our streaming workload. Meanwhile, Ryzen 5 1600X provides a much more consistent in-game experience.

Grand Theft Auto V

Grand Theft Auto V scales well with additional host processing resources and tends to favor Intel architectures. It isn't surprising to see the Core CPUs lead during our baseline tests. But we weren't expecting the Core i5-8600K to maintain its advantage over the competition while streaming.

The Core i5-8600K delivers 100% of its frames during the test, though only 89.71% fall into the perfect 60 FPS bucket. We see a nearly even split of 5% above and below that mark. Visually, the stream looks fine. The -8600K takes more of a hit in the 99th percentile measurements, though it still leads the rest of our contenders.

AMD's Ryzen 5 1600X starts out with the lowest average frame rate during our baseline and falls to 60.7 FPS during the stream. It doesn't suffer too badly in 99th percentile measurements. Meanwhile, the chip provides stellar streaming performance, dropping no frames and almost landing at a smooth 60 FPS.

The Core i5-7600K provides a decent gaming experience during normal streaming, churning out 79.2 FPS. That's because it really isn't encoding frames, though—it drops 84.9% of them. The stream is unwatchable by any measurement. Again, we can boost the -7600K's encoding performance to 100% by specifying high priority, but then the gaming experience is terrible. In fact, large portions of the scene simply do not render correctly.

Overclocking the Core i5-7600K doesn't help; it's still entirely saturated during the streaming test. But tuning does benefit Intel's Core i5-8600K, which leads the pack in both average and 99th percentile metrics. Its streaming quality is decent as well.

The Ryzen 5 1600X's performance improves after overclocking, too. It doesn't provide the absolute best frame rate while streaming, but it is perfectly playable. Also, Ryzen 5 1600X still offers the smoothest stream, if only just barely.

Middle-earth: Shadow of War

Intel's Core i5-7600K continues to disappoint. It serves up an average of 51.4 FPS while streaming (using the high priority setting) and encoding 100% of the frames. Unfortunately, game smoothness goes out the window again. The 15.9 FPS 99th percentile measurement tells us everything we need to know.

The Ryzen 5 1600X offers decent gaming performance while streaming, but that Coffee Lake-based -8600K continues leading.

AMD's Ryzen 5 1600X offers the best of both worlds, though Core i5-8600K is also very capable. Perhaps tuning the encode could get more performance from the Intel chip. Then again, you could say the same thing about Ryzen 5, which offers the best mix of streaming and frame rates in its class. It's obvious from these results that a Kaby Lake-based Core i5 owner needs to dial back streaming settings dramatically or turn to faster hardware.

MORE: Best CPUs

MORE: All CPUs Content

AMD Ryzen 7 & Intel Core i7

Paying more for a processor should also get you more performance, right? With that in mind, we're stepping up to the "faster" encoder preset for this CPU class.

AMD Ryzen 7 1800X

Intel Core i7-8700K

Intel Core i7-7700K (Kaby Lake)

Battlefield 1

The Core i5-7600K disappointed us in the previous round of tests, and Core i7-7700K picks up where the i5 left off. This chip just isn't well-equipped for streaming, despite the additional four threads enabled by Hyper-Threading. The -7700K dropped 94.3% of our frames using the normal settings, which comes as a side effect of excessive CPU utilization. We triggered the high priority setting to try rectifying this situation, but were presented the same bipolar behavior as Intel's Core i5-7600K. There just isn't a straightforward way around saturated CPU cores. You can either stream well or game well, but you can't do both with Kaby Lake-based processors and OBS streaming at 1920x1080/60FPS.

The Ryzen 7 1800X bears down with eight cores and 16 threads to provide solid streaming and gaming performance.

Core i7-8700K, which brings Coffee Lake up to a Hyper-Threaded six-core design, also performs well in this test. Overall, it offers the best gaming performance while streaming, 99th percentiles included. It even ekes past Ryzen 7 1800X with 100% of frames encoded to the stream. AMD's Ryzen 7 delivers 99.9% of the frames, and dropping 0.1% doesn't concern us. The stream is still smooth.

Overclocked, Ryzen 7 1800X delivers an ever-so-slightly better stream. But the Core i7-8700K leads in overall gaming performance. Of course, the 1800X's 117.8 FPS average, while streaming, is plenty impressive. Both processors prove up to this task.

Our encoding workload pegs the -7700K's cores at 100%, while the high priority settings help bring down CPU utilization. Simply, that's because the CPU has fewer frames to encode as a result of lower gaming performance.

Grand Theft Auto V

The Core i7-8700K takes a healthy lead during the baseline and streaming tests. It even provides more in-game performance while streaming than an unencumbered Ryzen 7 1800X.

Ryzen 7 1800X again achieves lower gaming performance, but a slightly superior stream than Intel's -8700K. The differences are pretty much imperceptible, though. Kaby Lake, on the other hand, simply can't compete.

Turning up the clocks benefits Coffee Lake greatly. Core i7-8700K offers the best overall performance, while Ryzen 7 1800X is competitive.

Middle-earth: Shadow of War

Coffee Lake continues to impress with strong in-game and streaming performance. The Ryzen 7 1800X does encode a higher percentage of frames than -8700K within the desirable 16.667ms range, but it also generates fewer frames. AMD's Ryzen does have more headroom to spare during our streaming workload though, so it'd likely accommodate more taxing encode presets.

We find that the stream frame time variance isn't very noticeable unless the 16.667ms percentage drops below a 90% threshold, such as what we see from Intel's Core i7-7700K. The -7700K does serve up better frame rates while it's streaming, even besting AMD's Ryzen 7 1800X. But that goes to show why we need to keep an eye on performance from every angle. Core i7-7700K drops 40% of its frames during this test! That makes any gaming performance advantage meaningless.

Ryzen 7 1800X musters a solid 75.4 FPS during the streaming test and provides a crisp stream to YouTube, as well.

Overclocking amplifies the trends we've been talking about. The same inherent strengths and weaknesses are still apparent. Ryzen 7 1800X and Core i7-8700K are both solid choices for streaming.

The Core i7-7700K demonstrates lower CPU utilization than usual during this test. Surprisingly, that doesn't equate to a usable combination of in-game and streaming performance, though, suggesting other architectural influences may be affecting our numbers.

MORE: Best CPUs

MORE: All CPUs Content

AMD Threadripper & Intel Skylake-X i9

You expect the best when you drop $1000+ on a CPU for gaming and streaming simultaneously. So, we shifted to the "fast" setting for this round of tests.

AMD Ryzen Threadripper 1950X

Intel Core i9-7900X

Intel Core i9-7980XE

Battlefield 1

AMD didn’t design Threadripper for "just" gaming, particularly at lower resolutions. But an intense streaming workload might expose more of the architecture's benefits.

Threadripper does have several configurable modes to tailor to its response to various tasks, as we outlined in our AMD Ryzen Threadripper 1950X Game Mode, Benchmarked article. Game Mode and Creator Mode both impact various titles differently, and we expect those same trends to carry over to our streaming benchmark. As such, we tested the Threadripper 1950X both ways.

The Threadripper 1950X pulls into range of Intel's Skylake-X models during our baseline tests. Creator Mode (CM) exposes all 32 of the 1950X's available threads, yielding more potential horsepower than Game Mode (GM). The 1950X's Creator Mode also makes it possible to encode 100% of the test run's frames, while Game Mode drops 1.6% of them. That isn't a huge sacrifice, but Game Mode also causes Threadripper to average 26 FPS fewer. That means it isn't encoding as many frames, either. We would have expected higher frame rates from a lighter encoding workload, but disabling half of the 1950X's threads in pursuit of higher game performance doesn't always work well when you're streaming, too.

The 18C/36T Core i9-7980XE doesn't perform as well as the 10C/20T Core i9-7900X during the streaming tests. Rather, the -7980XE stumbles in the smoothness department as its 99th percentile frame rates fall below the Threadripper 1950X, despite a lead in average frame rates.

Aside from Threadripper 1950X in Game Mode, all of these processors encode 100% of the frames.

The Core i9-7980XE's game performance is higher than the Threadripper models during our streaming test after bumping up clock rates. It also delivers a much better 99th percentile measurement.

AMD's Threadripper configurations fare better after tuning, too. Both encode 99.9% of the frames we send their way, which is adequate for a quality stream.

Grand Theft Auto V

Grand Theft Auto V's in-game performance goes Intel's way during the baseline benchmarks. Moreover, Core i9-7980XE redeems itself when it comes time to stream. That chip does command quite a premium though, so its advantage doesn't necessarily represent the best value.

Again, Threadripper's Game Mode just doesn't appear to be ideal for streaming. Game Mode provides better baseline frame rates in this title, but it falls to the bottom of our chart once we start encoding video.

Middle-earth: Shadow of War

Once again, Intel's processors offer the best in-game and streaming performance. This isn't entirely surprising; Threadripper may even be overkill for this type of enthusiast workload. We'd expect the architecture to handle workstation-class production workflows more adeptly.

Core i9-7900X does encounter a hiccup as its 99th percentile scores fall below Threadripper 1950X's during the streaming workload. The 1950X in Creator Mode also experiences some variance, with 20.38% of its frames falling below the 16.667ms threshold.

Even after a bit of tuning, Core i9-7900X provides a higher average frame rate than the 1950X in Creator Mode during our streaming benchmark, but can't match AMD's 99th percentile performance.

The Threadripper 1950X's difficulty in Creator Mode while streaming Middle-earth is even more pronounced. Some aspect of OBS doesn't agree with Creator Mode and this one game. We ran the tests several times to ensure it was a repeatable phenomenon.

MORE: Best CPUs

MORE: All CPUs Content

Final Analysis

AMD democratized access to high core counts with an attractively-priced Ryzen portfolio. In a way, we can thank the company for Intel's newfound interest in competing on a $/core basis. Just look at the difference in our results from Kaby Lake to Coffee Lake. As a result, high-quality software encoding on a gaming PC is becoming more realistic to mainstream gamers.

Our testing is indicative of general performance trends. Given enough time and energy, you could almost certainly improve upon our results. Part of that is by design: we're using these settings to compare large groups of processors against one another on a level playing field, as opposed to wringing the most performance out of any one processor. While we can't use our benchmarks to make definitive statements about the possibilities with each chip, we can draw some fair conclusions about how certain architectures behave.

Encoding is a parallelizable workload. If software encoding is your primary goal, you'll definitely want to seek out CPUs with lots of cores and simultaneous multi-threading capabilities. Intel's quad-core Kaby Lake models illustrate how chips that once offered class-leading gaming performance can fall apart during streaming. You can boost their performance by using less intensive quality presets, lowering the streaming frame rate, or sacrificing some quality with GPU acceleration. However, competing processors offer much more performance than Kaby Lake at the same presets and roughly the same price.

Many streamers place video quality over maximizing the frame rate of whatever game they're playing, so your own priorities will largely dictate how you tune your system. In fact, turning on v-sync may be a good way to balance streaming and gaming performance.If you seek the highest in-game performance while you stream, Intel's Coffee Lake-based Core i7-8700K is a good fit. The Ryzen 7 1800X is also competitive and tends to offer better streaming performance. Using our settings, the 1800X also had more CPU headroom leftover for more taxing encode settings, if desired. Granted, some of that extra horsepower is due to the 1800X's lower gaming performance, which means there are fewer frames to encode.

Two extra cores on the Coffee Lake-based Core i5 certainly help its standing, but the lack of Hyper-Threading has a definite impact on streaming performance. In the end, a six-core Core i5-8600K is forced to battle the 12-thread Ryzen 5 1600X, which offers a more balanced profile. Overclocking does help Intel somewhat. It can't overcome the advantage AMD gets from a more thread-heavy architecture, but it shrinks the gap somewhat in streaming workloads.

If you're really serious about streaming and gaming at the same time, the highest-end desktop CPUs are an option. Just expect to pay dearly for them. Most enthusiasts are better served by mainstream processors. Intel's Core i9 models generally provide better performance than the Threadripper 1950X, but they cost more, too. The 1950X is a solid value choice that also offers a diverse range of capabilities.

There are plenty of other solid options for gaming/streaming, and this introductory round of tests only focused on high-end models from each family. We'll expand our testing to locked SKUs as we work through coming CPU reviews.

MORE: Best CPUs

MORE: All CPUs Content

AMD Ryzen Threadripper 1900X CPU Review

palcorn@outlook.com (Paul Alcorn) — Mon, 23 Oct 2017 13:00:00 +0000

Eight Cores For The High-End Desktop

AMD launched its Threadripper CPUs for high-end desktops. But, at the last minute, it also turned heads with a cheaper Ryzen Threadripper 1900X, an 8C/16T model that drops into X399-based motherboards. The company claims its 1900X is an ideal entry point for folks who might want one of the other Threadripper chips at some point down the line.

By now, we're all familiar with the key tenets of AMD's value proposition: you get more cores for less money, more affordable motherboards, and all of its architecture's features. Ryzen Threadripper 1900X carries that philosophy forward, though it sports the same number of cores as its nearest Intel competition (at a similar MSRP, no less), resides on a very pricey motherboard, and it costs more than the "mainstream" Ryzen 7 1800X, also an eight-core contender.

Specifications

The 1900X features a base frequency of 3.8 GHz, the highest base clock rate in AMD's Threadripper family. And it boosts up to 4.2 GHz if your cooler is beefy enough. Two dual-channel memory controllers combine to facilitate plenty of aggregate bandwidth, not to mention support for up to 512GB of DDR4. That's one significant advantage over the 1800X's dual-channel controller. The 1900X also supports ECC UDIMMS. And then there's the extra connectivity: Ryzen Threadripper 1900X exposes up to 64 lanes of PCIe 3.0 and more USB ports than Ryzen 7 1800X, which is limited to 16 PCIe lanes.

Why care about all of that extra I/O? It could be useful for power users with multiple GPUs, video capture cards, or voluminous NVMe-based storage arrays. In yet another example of AMD's broader feature set, the company offers bootable NVMe RAID 0, 1, and 10 for up to 10 attached SSDs, while Intel charges an additional fee for a vROC (Virtual RAID-on-CPU) dongle to unlock its full feature set.

There are vast architectural differences between the 1900X and Ryzen 7 1800X, as well. In a nutshell, the larger Threadripper models distribute active cores across two dies and all four of the quad-core CCXes. But in a bid to mitigate a layer of latency, AMD confines the 1900X's active cores to a single CCX inside each die.

Each CCX features 8MB of shared L3 cache, so disabling two CCXes also removes much of that last-level storage space. This cuts the 1900X's total to 16MB of L3, while the larger Threadripper models have up to 32MB. Fortunately, no changes are made to the entry-level chip's memory or I/O controllers.

	Threadripper 1950X	Threadripper 1920X	Core i7-7820X	Threadripper 1900X	Ryzen 7 1800X
Price	$999	$799	$599	$549	$499
Interface/Chipset	TR4/X399	TR4/X399	LGA 2066/X299	TR4/X399	AM4/X370
Cores/Threads	16/32	12/24	8/16	8/16	8/16
TDP	180W	180W	140W	180W	95W
Base Frequency (GHz)	3.4	3.5	3.6	3.8	3.6
Boost Frequency (GHz)	4.0 (4.2 XFR)	4.0 (4.2 XFR)	4.3 / 4.5 (TB 3.0)	4.0 (4.2 XFR)	4.0 (4.1 XFR)
Cache (L2+L3)	40MB	38MB	19MB	20MB	20MB
Memory Support	DDR4-2666	DDR4-2666	DDR4-2666	DDR4-2666	DDR4-2666
Memory Controller	Quad Channel	Quad Channel	Quad Channel	Quad Channel	Dual Channel
Unlocked Multiplier	Yes	Yes	Yes	Yes	Yes
PCIe Lanes	64	64	28	64	16

Interestingly, AMD still recommends using its Game Mode for the eight-core 1900X. This configuration attempts to confine data processing to the local die and its attached memory controller (NUMA). It also deactivates one die via software to eliminate die-to-die latency. That means AMD's Game Mode switches this potentially mighty CPU into a 4C/8T solution, similar to AMD's Ryzen 5 1500X and 1400.

Despite its cuts, AMD leaves the 1900X's TDP at 180W, similar to the 12-core 1920X and 16-core 1950X. That number is notably higher than Ryzen 7 1800X's 95W TDP and the Core i7-7820X's 140W rating. Like the other Threadripper models, you need to supply your own cooler; AMD doesn't bundle one. At least the company's decision to use Indium solder between its dies and heat spreader goes a long way when we get to overclocking.

Complicating the 1900X's value story is the price of X399 motherboards, which cost significantly more than X370-based platforms. So, if you're strictly a gamer, stay with AMD's Socket AM4-based chips before sinking big-time bucks into Ryzen Threadripper 1900X.

MORE: Best CPUs

MORE: All CPUs Content

Overclocking, Infinity Fabric & Test Setup

The Architectural Bits

The Ryzen 7 series features a single-die design, whereas AMD's Threadripper models employ a quad-die package with just two of its dies active. That means the 8C/16T 1900X effectively wields a dual-die arrangement with four active cores in each die.

Our first slide below shows the alignment of AMD's 16C/32T Ryzen Threadripper 1950X. Each die contains a pair of four-core CPU complexes (CCXes) that incur increased latency when they communicate with the neighboring CCX (denoted as CCX0 and CCX1). Another layer of latency comes into play when they communicate with the CCXes resident on the second die (marked as Die1). Simplified, the greater distance between dies means that die-to-die latency is much higher than the latency between two CCXes resident on the same piece of silicon.

The larger Threadripper models distribute active cores across both CCXes inside each die. The second slide shows how AMD disables one core per CCX (blocked out in blue) to create the smaller 12C/24T 1920X model.

AMD takes an entirely different tack with its 1900X, as seen on the third slide. In a bid to eliminate one layer of latency, AMD confines the 1900X's active cores to a single CCX inside each die. The adjustment makes sense; spreading the cores evenly across all four CCXes increases the chance of incurring latent communication with neighboring CCXes.

Disabling two entire CCXes also has other implications, though. Inactive cores, provided they are near active cores, can absorb excess heat, potentially improving overclockability. Case in point: it's common to achieve lower overclocks on the 16C 1950X than the 12C 1920X. Consequently, the 1900X's clustered cores should reduce latency, eliminating the CCX-to-CCX delay entirely and only leaving us with die-to-die latency. But they could also potentially hamper overclocking.

The Infinity Fabric Breakdown

A few quick tests with SiSoftware's Sandra Multi-Core Efficiency test illustrate the consequence of AMD's design decisions. Flipping the 1900X into Creator Mode, with all eight cores active, results in three distinct layers of latency. In contrast, the 1950X in Creator Mode has four layers.

Switching the 1900X into Game Mode disables one entire die, leaving us with a 4C/8T processor that has only two layers of latency. Again, the 1950X in Game Mode has a third layer that affects performance. But the 1900X in Game Mode also achieves the lowest fabric bandwidth in our line-up of AMD models.

Finally, the two disabled dies remove a total of 16MB of L3 cache from the 1900X. That means it offers half of the multi-threaded cache bandwidth of the 1950X in Game and Creator Mode. Incidentally, the 1900X demonstrates less multi-threaded throughput than Ryzen 7 1800X in Game Mode. But the 1800X also proffers eight cores with simultaneous multi-threading, while the 1900X in Game Mode drops to a 4C/8T processor with only 8MB of L3 cache. Many games are sensitive to memory and cache performance, so it will be interesting to see how that plays out in our game testing.

Overclocking

Overclocking the 1900X was an exercise in simplicity. We merely adjusted the data rate to DDR4-3200 and set timings at 14-14-14-34. We increased Vcore to 1.39V, well below AMD's recommended maximum of 1.45V, and adjusted the SoC voltage to 1.1V. This proved stable up to 4 GHz during extended stress tests. However, even in the face of unsafe voltages, we were unable to attain a stable 4.1 GHz overclock to match our efforts with the 12-core Threadripper 1920X. Dialing back the memory frequency didn't help, either. Considering that AMD supposedly selects the top 5% of its dies for Threadripper CPUs, you might assume that the clustered active core arrangement comes into play. We only see a 100 MHz reduction, so it's more likely that our retail sample is simply on the lower end of the bell curve.

It's notable that a 4 GHz overclock might actually hamper the 1900X's performance in lightly threaded workloads, since we lose the benefit of a quad-core 4.2 GHz XFR boost.

Test Systems

Test System & Configuration
Hardware	AMD Socket SP3 (TR4)AMD Ryzen Threadripper 1950X, 1920X, 1900XAsus X399 ROG Zenith Extreme4x 8GB G.Skill Ripjaws V DDR4-3200 @ 2666 and 3200 MT/sIntel LGA 2066Intel Core i7-7820XMSI X299 Gaming Pro Carbon AC4x 8GB G.Skill Ripjaws V DDR4-3200 @ 2666 and 3200 MT/sAMD Socket AM4 AMD Ryzen 7 1800XMSI X370 Xpower Gaming Titanium2x 8GB G.Skill RipJaws V DDR4-3200 @ 3200 MT/sIntel LGA 1151 Intel Core i7-7700K MSI Z270 Gaming M72x 8GB G.Skill RipJaws V DDR4-3200 @ 2666 and 3200 MT/sAll EVGA GeForce GTX 1080 FE 1TB Samsung PM863 SilverStone ST1500, 1500W Windows 10 Creators Update Version 1703Corsair H115i

Test System & Configuration

Hardware

AMD Socket SP3 (TR4)AMD Ryzen Threadripper 1950X, 1920X, 1900XAsus X399 ROG Zenith Extreme4x 8GB G.Skill Ripjaws V DDR4-3200 @ 2666 and 3200 MT/sIntel LGA 2066Intel Core i7-7820XMSI X299 Gaming Pro Carbon AC4x 8GB G.Skill Ripjaws V DDR4-3200 @ 2666 and 3200 MT/sAMD Socket AM4 AMD Ryzen 7 1800XMSI X370 Xpower Gaming Titanium2x 8GB G.Skill RipJaws V DDR4-3200 @ 3200 MT/sIntel LGA 1151 Intel Core i7-7700K MSI Z270 Gaming M72x 8GB G.Skill RipJaws V DDR4-3200 @ 2666 and 3200 MT/sAll EVGA GeForce GTX 1080 FE 1TB Samsung PM863 SilverStone ST1500, 1500W Windows 10 Creators Update Version 1703Corsair H115i

MORE: Best CPUs

MORE: All CPUs Content

VRMark, 3DMark & AotS: Escalation

Test Notes

The Threadripper processors feature a unique architecture that underperforms AMD's other Ryzen chips in some games. To offset the 1900X's compromises, AMD facilitates its novel Game Mode, which switches the processor into NUMA mode and disables one die. We covered the feature in-depth in our AMD Ryzen Threadripper 1950X Game Mode, Benchmarked article.

AMD aims Threadripper at content creators, heavy multitaskers, and gamers who stream simultaneously. It also says the processors are ideal for gaming at high resolutions. Threadripper CPUs and the GPUs they're likely paired with aren't intended for playing around at low resolutions, particularly in older, lightly-threaded titles. Still, we test at 1920x1080 to emphasize the difference between competing processors, rather than show you results bound by our graphics card.

Intel Core i7-7820X Skylake-X

AMD Ryzen 7 1800X

Intel Core i7-8700K

VRMark & 3DMark

We aren't big fans of using synthetic benchmarks to measure game performance, but 3DMark's DX11 and DX12 CPU tests provide useful insight into the amount of horsepower available to game engines.

As expected in VRMark, Threadripper 1900X offers its best performance in Game Mode, even beating the tuned Creator Mode configuration. VRMark tends to favor high clock rates, so the Ryzen 7 1800X suffers at stock settings due to its lower frequency. However, it nearly catches the 1900X after tuning.

The 1900X in Creator Mode takes a healthy lead over its competition during the DX12 tests, even beating out Intel's tuned Core i7-7820X. It suffers in Game Mode, though. That isn't surprising; we disable half of the chip's compute resources in a test that scales well with core count, after all.

Ashes of the Singularity: Escalation

Ashes of the Singularity is the poster child for threaded game engines. And while the Core i7-7820X at 4.6 GHz leads, the Ryzen 7 1800X at 4 GHz challenges Intel's Skylake-X solution. The 1900X trails mightily in Game Mode, but fares better in Creator Mode with eight threads available.

Even in heavily threaded games, mainstream processors like Ryzen 7 1800X offer the best value.

MORE: Best CPUs

MORE: All CPUs Content

Civilization VI, Battlefield 1 & Dawn of War III

Civilization VI AI Test

The Civilization VI AI test measures CPU performance in a turn-based strategy game. Intel's Core i7-7820X takes the lead after tuning, while the overclocked Threadripper 1920X leads AMD's line-up. Ryzen Threadripper 1900X provides better performance in Game Mode, but we only get a slight speed-up after overclocking.

Civilization VI Graphics Test

As we found during our Game Mode testing, the Threadripper models perform best in this benchmark with Game Mode activated. A tuned Threadripper 1900X takes the lead during our test, even besting Intel's Coffee Lake-based Core i7-8700K.

The Ryzen 7 1800X also proves to be adept, equating to good value for mainstream gamers thanks to a less expensive buy-in. Ryzen 7 1700 offers similar performance after overclocking, but includes a nice discount.

Battlefield 1 (DX11)

As usual, we bump up against a graphics bottleneck at the top end of our Battlefield 1 results. An overclocked Ryzen 7 1800X offers the best performance from AMD's portfolio, although the Threadripper 1900X is right behind in Creator Mode. The 1900X loses quite a bit of steam when we toggle Game Mode on; even our overclocking efforts aren't enough to overtake the Ryzen 7 1800X.

Warhammer 40,000: Dawn of War III

At stock settings with Game Mode enabled, Ryzen Threadripper 1900X outperforms AMD's stock Ryzen 7 1800X. The overclocked 1800X nudges past 1900X at the same 4 GHz, while Threadripper 1920X offers the best performance from the Ryzen portfolio.

While Threadripper 1900X beats Core i7-7820X at its stock settings, tuning improves the Intel model's performance significantly, and it ends up ahead.

MORE: Best CPUs

MORE: All CPUs Content

Grand Theft Auto V, Hitman & Shadow of Mordor

Grand Theft Auto V

Grand Theft Auto finds the 1900X and 1800X offering similar performance levels in both stock and overclocked configurations. Running the 1900X in Creator Mode provides better performance than a stock Ryzen 7 1800X.

Meanwhile, the overclocked 1920X beats Intel's stock Core i7-7820X, though tuning propels Intel into a commanding lead.

A stock Coffee Lake-based Core i7-8700K nearly catches the tuned -7820X, so you can imagine it would take the lead after some overclocking of its own.

Hitman (2016)

The 1900X doesn't respond well to our Hitman benchmark, demonstrating lower average and minimum frame rates compared to the mainstream Ryzen 7 1800X. The gap is much more pronounced between AMD's and Intel's chips, though we imagine this would shrink at higher resolutions.

Middle-earth: Shadow of Mordor

Ryzen Threadripper 1920X and 1900X top the chart in this largely graphics-bound test.

Middle-earth: Shadow of Mordor does tend to favor the highest clock rates, making AMD's slight win over the overclocked Core i7 more impressive.

MORE: Best CPUs

MORE: All CPUs Content

Project CARS, Far Cry Primal & Rise of the Tomb Raider

Project CARS

Intel's Core i7-8700K offers the best performance during our benchmark, while the 1900X lands in the middle of this chart with Game Mode active. Creator Mode makes more threads available to the Madness game engine, though Slightly Mad Studios can't seem to extract maximum performance from AMD's distributed MCM architecture.

Far Cry Primal

Far Cry Primal's Dunia 2 engine refuses to run in Creator Mode on the 32-thread Threadripper 1950X processor. We don't have the same issue with AMD's Ryzen Threadripper 1900X and 1920X, though.

The 1900X gains nothing from Creator Mode. Meanwhile, the Threadripper 1900X's Game Mode fares quite a bit better, gaining even more after overclocking.

A stock Core i7-7820X lags the tuned Threadrippers in Game Mode, but really stretches its legs after we crank it up to 4.6 GHz.

Rise of the Tomb Raider

Rise of the Tomb Raider has long been a thorn in the Ryzen architecture's side. But recent software patches helped rectify AMD's deficits encountered at launch.

The Intel processors still lead during our benchmark, but an overclocked Ryzen 7 1800X ekes out a slight lead over AMD's Threadripper models.

MORE: Best CPUs

MORE: All CPUs Content

Office & Productivity

Test Notes

We toggled Ryzen Threadripper 1900X into Creator Mode for our application testing. This setting exposes the full might of AMD's 8C/16T design, even if some lightly threaded applications fare better under various combinations of the NUMA/UMA and Legacy mode toggles.

Adobe Creative Cloud

Most of the Adobe Creative Cloud test suite favors Intel's processors.

Ryzen Threadripper 1900X serves up competitive performance and benefits from tuning, but the 1920X's overclocked frequency advantage facilitates a slight win in most tests.

After Affects responds well to increased core counts, so the 1900X competes readily. Ryzen 7 1800X isn't as fast in this test, languishing far below the eight-core 1900X. It seems as if the difference comes from Threadripper's quad-channel memory configuration, as repeating the test in dual-channel mode incurs a ~6-second penalty, putting 1900X at the same level as Ryzen 7 1800X. We also observed similar trends with the Photoshop Heavy and InDesign tests.

Web Browser

The web browsing tests align largely based on per-core performance, so frequency and IPC throughput reign supreme. These lightly-threaded metrics highlight the only drawback to overclocking the 1900X.

That is to say our 1900X's 4 GHz ceiling is the same as its maximum quad-core Precision Boost frequency, so overclocking will speed up workloads that use more than four cores. However, if your cooler is beefy enough, a stock Threadripper 1900X hits a 4.2 GHz Extended Frequency Range peak across four cores. So, in some lightly-threaded tasks, you'll see a stock 1900X outperform the overclocked configuration.

Case in point: a stock 1900X outperforms the tuned configuration in our Kraken JavaScript and MotionMark benchmarks, though the deltas are small. A tuned Ryzen 7 1800X also offers better performance than the 1900X. The Intel processors use their frequency and IPC throughput advantage to top the charts.

Cryptography is important for securing online transactions and many other applications. The Threadripper processors enjoy a big lead over competing Intel models in single-core SHA2-256 hashing performance. They are also competitive in the single-core AES-256 tests. But we can clearly see the benefits of the -7820X's AVX2 performance compared to the Core i7-8700K.

Multi-core AES-256 tests align based largely on core count and frequency, but we can spot the advantages of quad-channel memory. The overclocked 1900X doesn't gain much improvement in the SHA2-256 test.

Productivity

We're incorporating portions of the PCMark 10 suite into our test regimen. The application start-up metric measures load-time snappiness for several types of applications, such as word processors, GIMP, and Web browsers, in both warm- and cold-start conditions.

Intel's Core i7-7820X beats the overclocked Threadripper models by a slight margin, but tuning widens the gap considerably. Interestingly, a stock Core i7-8700K almost matches the best effort of our overclocked -7820K. Coffee Lake truly is an impressive performer here.

Video conferencing measures performance in single- and multi-user applications that utilize the Windows Media Foundation for video playback and encoding. It also performs facial detection during the workload to model real-world performance. The test consists of both native and OpenCL acceleration, so we see small performance improvements with OpenCL. The Threadripper 1900X and 1920X fall into the middle of the test results, while the tuned Ryzen 7 1800X provides more performance with native processing.

The photo editing benchmark measures performance with Futuremark's binaries that use the ImageMagick library. We can see the big gains with OpenCL acceleration, and the processors with the highest clock rates generally offer the best results. That means Intel's tuned -7820X and AMD's overclocked 1920X provide similar performance. The 1900X benefits from tuning, but Ryzen 7 1800X at 4 GHz is even faster. Native processing leverages Theadripper 1920X's extra threads to great effect.

We tested the 1900X with both dual- and quad-channel memory to ferret out improvements borne of extra bandwidth. The 1900X with quad-channel memory provided ~6% more performance than the dual-channel setup during the writing test (LibreOffice Writer), helping propel the overclocked 1900X to a lead over the tuned 1800X.

MORE: Best CPUs

MORE: All CPUs Content

Rendering, Encoding, Compression & AVX

Rendering

As expected, the tuned Threadripper 1900X provides similar performance as an overclocked 1800X in our multi-core Cinebench test, though it ekes out a win at stock settings thanks to its higher frequency. Threadripper 1920X's extra cores/threads allow it to take a commanding lead over the rest of the group.

Cinebench's single-core benchmark places Ryzen Threadripper 1900X in the middle of the group. And again, we see the 1900X's stock configuration use its higher XFR boost to beat out the overclocked version. Intel's processors continue to offer higher clock rates and IPC throughput though, so they take the top spots in our chart.

Corona utilizes all available cores and threads, so we see similar performance from the 1900X and Ryzen 7 1800X, though the latter takes a slight lead after tuning.

The single-core POV-Ray testing repeats what we saw under Cinebench: the stock 1900X edges our our overclocked configuration again. Meanwhile, the multi-core POV-Ray and Blender tests benefit from the 1900X's higher frequencies in stock trim. Threadripper 1900X doesn't benefit much from its quad-channel memory during those tests, as evidenced by its similar performance to the Ryzen 7 1800X when they are both locked to the same frequency. PCMark 10's rendering and visualization test tells much the same story.

The 1900X offers solid performance in our OpenCL-accelerated LuxMark test, though. We don't have OpenCL-based results from the Core i7-7820X, but that's not a mistake. We spent considerable time trying to get this test to run correctly on any Skylake-X processor, to no avail. Intel later confirmed our suspicions that OpenCL isn't correctly taking advantage of AVX-512 instructions, so we'll have to wait for a fix before we can generate results. The company has plans to support AVX-512 in a future release of the Intel OpenCL SDK.

Encoding & Compression

LAME finds the 1900X offering similar performance as Ryzen 7 1800X once again, though the eight-core Threadripper is a bit faster in our x264 HandBrake test.

We re-ran that benchmark on the 1900X using dual- and quad-channel memory configurations and recorded a 20% performance advantage with quad-channel. The tuned 1900X only offers a 3% performance lead over our overclocked 1800X, suggesting that some of its advantage may be lost to poor application scaling or code that isn't optimized for the unique Threadripper architecture.

We see a larger delta between the Intel and AMD processors during the HandBrake x265 test than the x264 test, but that is likely due to the former's heavy use of AVX instructions. Again, the tuned 1900X's 10% performance improvement with x265 and quad-channel memory doesn't equate to a large win over a tuned Ryzen 7 1800X. We also provide results from Y-Cruncher, a single- and multi-threaded program that computes Pi using AVX instructions. We tested with version 0.7.3.9474, which includes Ryzen optimizations.

Compression workloads benefit greatly from multi-core architectures, provided the storage subsystem can feed the processor fast enough to utilize its full capabilities. The stock 1900X offers similar performance as an overclocked 1800X during the multi-core compression workload, which is a byproduct of its quad-channel memory advantage. A tuned Core i7-7820X takes the overall lead, but the overclocked 1920X challenges.

The decompression benchmark benefits from integer performance, and the 1920X's ability to work on 24 threads concurrently provides a tremendous advantage over the rest of the pack. AMD's 1920X available resources allow it to post an almost-50% lead in overclocked trim.

MORE: Best CPUs

MORE: All CPUs Content

Final Analysis

Remember a few short months ago when Intel introduced its Kaby Lake-X Core i7-7740X, which sold for an affordable $350, giving the impression of high-end value, but then required a super-expensive X299-based motherboard? Ryzen Threadripper 1900X sort of feels like that to us. But whereas the -7740X totally neutered Intel's platform with just 16 PCIe lanes and a pair of disabled memory channels, at least Threadripper 1900X comes armed with all of its architecture's functionality intact. Sixty-four lanes of PCIe 3.0 and four channels of DDR4 memory with ECC support may make the difference to power users with lots of add-in devices or bandwidth-sensitive workloads. But 1900X just isn't much more compelling than Ryzen 7 1800X, which also supports ECC memory on some motherboards and comes with a more affordable platform.

We plotted the 1900X's gaming performance with both average frame rates and a geometric mean of the 99^th percentile frame times (a good indicator of smoothness), which we convert into an FPS measurement. Our suite includes six games released in 2016 and five older titles that launched in 2014/2015. When we reviewed the higher-end Threadripper models, we hypothesized their extra cores could enable more performance in the future, so we included a chart with newer games. But that's not as big a selling point for 1900X, since its core count matches the 1800X.

If you're a gamer above all else, and semi-professional workloads aren't on your radar, AMD's Socket AM4-based Ryzen 5 and 7 CPUs are a better fit for you than Ryzen Threadripper. You'll see similar frame rates from a $220 Ryzen 5 1600 overclocked moderately. Of course, Intel would counter back that its Coffee Lake-based Core i5s between $200 and $300 are better still. The point is you have multiple options that are great for gaming before ever needing to consider a $500 Threadripper 1900X and a way-expensive motherboard.

The real competition happens in our application workloads. Ryzen Threadripper 1900X can't quite match the $600 Core i7-7820X in most workloads, so professionals on the hunt for overall performance may favor Intel's Skylake-X chip. The Ryzen 7 1800X often serves up similar performance as Threadripper 1900X, and it costs $100 less. Then there's the Core i7-8700K, which also sells for $400, trades blows with AMD's top Ryzen 7 chip, but currently suffers from a bad case of paperlaunchitis.

Consider also that exploiting the 1900X's four memory channels means buying a quad-channel kit of DDR4. And then there are the platforms: right now, the absolute cheapest TR4-equipped motherboard sells for $340. Most models come close to $400.

Of course, AMD says its Ryzen Threadripper 1900X is the lowest-cost way to get into its X399 platform...and it is. However, we can’t ascribe much enthusiast value to this niche option. There are faster choices if you prioritize performance and cheaper alternatives if you're trying to save money. Thus, we aren't particularly attracted to Threadripper 1900X. Please, AMD, don't be upset if we send flowers to this chip's better-looking sibling, Ryzen 7 1800X.

MORE: Best CPUs

MORE: All CPUs Content

Core i7-8700K Review: Coffee Lake Brews A Great Gaming CPU

palcorn@outlook.com (Paul Alcorn) — Thu, 05 Oct 2017 13:01:00 +0000

Intel Brings More Cores

Intel’s Core 2 Quad processors debuted with four cores in 2006. Although six-core models landed four years later in the high-end desktop space, the company’s most accessible chips topped out in quad-core configurations for more than a decade. The Coffee Lake era begins with Intel upending its mainstream line-up by adding two more cores to Core i7, i5, and i3 families. Call this a much-needed improvement, cleverly timed to stave off AMD's core-laden Ryzen assault.

Of course, while Intel's accelerated Coffee Lake-S launch makes it look today's unveiling is a direct response to heated competition, in reality, the long incubation period for new processors means it’s more likely the result of 10nm manufacturing delays.

Just last year, Intel announced a new process-architecture-optimize cadence designed to deliver smaller transistors every third generation. That's a profound departure from the glory days of Intel’s tick-tock model. The latest 14nm++ process is the fourth outing of the 14nm node, which originally debuted with Broadwell back in 2014. So, it appears that PAO is already falling by the wayside. In the days of tick-tock, we'd also be talking about a new architecture right now. But Coffee Lake employs the same Skylake design as Kaby Lake before it. We also get the same fundamental integrated graphics engine found in the previous generation. To be sure, Coffee Lake is another iterative update.

But there’s nothing mundane about adding more cores. Intel claims Coffee Lake offers up to 25% more gaming performance and up to 45% more “mega-tasking” performance. Given similar price points versus Kaby Lake, we're almost certainly looking at a huge step forward in value.

This is obviously a busy year for Intel. But as if the company's product stack wasn't already confusing, its eighth-generation portfolio draws from three separate architectures, including 14nm+ Kaby Lake-R (refresh), 14nm++ Coffee Lake, and 10nm Cannon Lake, which should land next year.

Intel Core i7-8700K

Core i7-8700K serves as this generation's flagship, sporting six Hyper-Threaded cores. Already, that's a big increase from Kaby Lake's 4C/8T maximum. It features the company's highest clock rates, accelerating up to 4.7 GHz via Turbo Boost. The -8700K does sacrifice some base frequency in exchange for a higher core count, though. Its 3.7 GHz specification is 500 MHz lower than the -7700K, offsetting the increased power consumption and heat generated by a 6C/12T configuration.

The -8700K's Coffee Lake design utilizes a 14nm++ process, which Intel claims offers 26% more performance and 52% less leakage power than first-generation 14nm manufacturing. Those advances enable the higher Turbo Boost bins and reduce consumption enough to carve out room for extra cores. A more complex die does necessitate a TDP rating of up to 95W. But that's only 4% higher than Core i7-7700K. And as we've seen before, Turbo Boost allows the CPU to operate beyond its rated TDP as long as current, power, and temperature fall below specified limits. As you might imagine, then, the impact of two additional cores is felt under load.

The top 4.7 GHz Turbo Boost bin should help improve performance in lightly-threaded applications. But Core i7-8700K also includes aggressive multi-core bins to help chew through threaded workloads. Because these CPUs employ Intel's Skylake architecture, we aren't expecting any speed-ups attributable to IPC throughput. All gains come from core count and clock rate alone. Intel isn't officially disclosing a die size or transistor count, but early delidding efforts indicate a ~151mm² area. That's naturally larger than Kaby Lake's ~122mm², reflecting the additional execution and cache resources. Intel confirms that Coffee Lake continues to employ its ring bus, rather than Skylake-X's mesh topology.

Active Cores	1	2	4	6
Intel Core i7-8700K	4.7 GHz	4.6 GHz	4.4 GHz	4.3 GHz
Intel Core i7-7700K	4.5 GHz	4.4 GHz	4.4 GHz	-

Intel also adds 50% more cores to its Core i5 family, and doubles Core i3's core count. But it pulls Hyper-Threading support from Core i3 in the process. Nevertheless, we expect gamers to realize palpable gains moving from dual-core Hyper-Threaded platforms to inexpensive quad-core setups.

Core i5 and i7 also support speedier DDR4-2666 transfer rates, up from Kaby Lake's DDR4-2400 spec. Core i3 remains limited to DDR4-2400, though. This could just be Intel's attempt to segment its line-up, or perhaps the Core i3s are really just quad-core Kaby Lake designs transplanted onto a 14nm++ process.

	Intel Corei7-8700K	Intel Corei7-8700	Intel Core i5-8600K	Intel Core i5-8400	Intel Core i3-8350K	Intel Core i3-8100
Socket	LGA 1151	LGA 1151	LGA 1151	LGA 1151	LGA 1151	LGA 1151
Cores/Threads	6 / 12	6 / 12	6 / 6	6 / 6	4 / 4	4 / 4
Base Frequency	3.7 GHz	3.2 GHz	3.6 GHz	2.8 GHz	4.0 GHz	3.6 GHz
Boost Frequency	4.7 GHz	4.6 GHz	4.3 GHz	4.0 GHz	N/A	N/A
Memory Speed	DDR4-2666	DDR4-2666	DDR4-2666	DDR4-2666	DDR4-2400	DDR4-2400
Memory Controller	Dual-Channel	Dual-Channel	Dual-Channel	Dual-Channel	Dual-Channel	Dual-Channel
Unlocked Multiplier	Yes	No	Yes	No	Yes	No
PCIe Lanes	x16 Gen3	x16 Gen3	x16 Gen3	x16 Gen3	x16 Gen3	x16 Gen3
Integrated Graphics	Intel UHD Graphics 630 (up to 1200 MHz)	Intel UHD Graphics 630 (up to 1200 MHz)	Intel UHD Graphics 630 (up to 1150 MHz)	Intel UHD Graphics 630 (up to 1150 MHz)	Intel UHD Graphics 630 (up to 1150 MHz)	Intel UHD Graphics 630 (up to 1150 MHz)
Cache (L2+L3)	13.5MB	13.5MB	10.5MB	10.5MB	9MB	7MB
Architecture	Coffee Lake	Coffee Lake	Coffee Lake	Coffee Lake	Coffee Lake	Coffee Lake
Process	14nm++	14nm++	14nm++	14nm++	14nm++	14nm++
TDP	95W	65W	95W	65W	91W	65W
Price (per 1K Unit)	$359	$303	$257	$182	$168	$117

Unfortunately, Intel still doesn't enable Turbo Boost on its Core i3 CPUs. So, we could see a performance dip in lightly threaded workloads due to Coffee Lake's lower base frequencies. You do get 100% more cores in exchange, though. Physical cores are always preferable to logical ones, so the new implementation of Core i3 should come out ahead in most of our benchmarks.

As expected, most models continue to wield 2MB of L3 and 256KB of L2 cache per core. As a side effect of its higher core counts, then, Coffee Lake processors enjoy the benefits of more cache. Core i3-8100 is the lone exception with only 6MB of L3 cache.

PCIe connectivity remains unchanged; you get 16 lanes of third-gen PCIe from the CPU's controller. Intel reminds us, though, that it offers up to 40 lanes when we add the platform controller hub's 24.

You'll need a Z370-based motherboard for Coffee Lake processors. The 200-series chipsets are not compatible. And in a clear indication that Intel really hurried its launch schedule, less expensive B- or H-series chipsets won't be ready until next year. Paying a premium for Z-class core logic isn't much of a surprise for enthusiasts, who need the higher-end chipset to support unlocked multipliers. But it's a little bit overkill for everyone else.

Coffee Lake	Intel Corei7-8700K	Intel Corei7-8700	Intel Core i5-8600K	Intel Core i5-8400	Intel Core i3-8350K	Intel Core i3-8100
Cost Per Core/Thread	$59.83 / $29.92	$50.50 / $25.25	$42.83 / $42.83	$30.33 / $30.33	$42 / $42	$29.95 / $29.95
Kaby Lake	Intel Core i7-7700K	Intel Core i7-7700	Intel Core i5-7600K	Intel Core i5-7400	Intel i3-7350K	Intel i3-7100
Cost Per Core/Thread	$84.75 / $42.38	$75.75 / $37.88	$60.50 / $60.50	$45.50 / $45.50	$84 / $42	$58.50 / $29.95
Ryzen	Ryzen 71700X	Ryzen 71700	Ryzen 51600X	Ryzen 51500X	Ryzen 51400	Ryzen 31300X
Cost Per Core/Thread	$49.88 / $24.94	$41.13 / $20.56	$41.50 / $20.75	$47.50 / $23.75	$42.25 / $21.12	$32.50 / $32.50

We're using Intel's 1K unit pricing for comparisons to the Kaby Lake models and AMD's MSRP for price-equivalent Ryzen chips. We may see higher prices on Intel's CPUs at retail, while AMD models routinely sell below MSRP.

Intel adds a ~$20 premium to its K-series SKUs compared to their Kaby Lake equivalents. Overall, though, you pay less per core. Again, Intel removed Hyper-Threading from its Core i3s, so their price per thread remains unchanged. With the exception of Ryzen 3, AMD maintains a price advantage across its portfolio, due in part to SMT on the Ryzen 5 family. The benchmarks will give us a better idea of performance-per-dollar compared to Kaby Lake and Ryzen, though.

Overclocking headroom was one of Kaby Lake's biggest advantages due to Ryzen's limited scaling. Intel adds per-core overclocking support to this generation, but doesn't provide per-core voltage and P-state controls. It also enables live memory timing adjustments (without rebooting), along with memory multipliers up to 8400 MT/s, so you don't have to adjust the BCLK frequency to chase bleeding-edge transfer rates. Finally, enhanced GT and Ring PLL Trim controls add more granular control.

Intel makes some power optimizations to its interface that promise to extend the advantage while overclocking. However, the company continues to insist on using thermal paste between its die and IHS, rather than solder. Like all unlocked Intel models, the Core i7-8700K doesn't include a stock cooler.

Nevertheless, we have to give the big company credit for staying on its toes this year. It already introduced Kaby Lake, Skylake-X, and Kaby Lake-Refresh. Next year, we'll have new Pentium and Celeron line-ups headed our way. But for now, we're looking forward to testing what Intel claims is its best gaming chip yet.

MORE: Best CPUs

MORE: Intel & AMD CPU Comparison Hierarchy

MORE: All CPUs Content

The Z370 Chipset & Graphics

The Z370 Chipset

Much to the dismay of enthusiasts everywhere, you can't buy a Coffee Lake-based CPU and drop it into the fancy Z270 motherboard you may have purchased a few months ago. Instead, you'll have to pony up for a 300-series motherboard. You also can't bring your Skylake or Kaby Lake processor over to a Z370-based platform, even though both sides employ LGA 1151 interfaces.

Intel makes more money selling CPUs than chipsets, so creating an upgrade path would have made sense. But the company tells us that it needed optimized memory trace routing to support DDR4-2666, and it also improved power delivery to support six-core models. It also beefed up package power delivery for overclocking.

According to Intel's specification sheet, the increased package power delivery consists of 18 pins repurposed to VCC (previously designated as reserve pins). There are also changes to the socket that reassign behaviors to some pins, such as swapping Normal Open (NO) and Normal Closed (NC) settings. Intel also improved power delivery to the graphics engine, and although the specific changes there remain undefined, that might improve HD Graphics overclocking.

The changes mean upgrading to Z370 is a technical necessity, rather than planned obsolescence. Regardless, though, AMD's planned support for Socket AM4 through 2020 makes Intel's Z370 chipset requirement appear all the more painful.

CPU-based I/O remains unchanged from Kaby Lake; you still get 16 lanes of PCIe 3.0. Again, though, the whole platform supports up to 40 lanes. The extra 24 originate from the PCH, stuck behind a four-lane DMI 3.0 connection that links the core logic and CPU. As you might imagine, this PCIe-like interconnect can become a bottleneck if enough devices (like M.2 SSDs) are working behind it.

The 22nm Z370 chipset is rated at 6W, and is mostly identical to Z270. It offers up to 10 USB 2.0, 14 USB 3.0, and six SATA 6Gb/s ports. Intel added support for Thunderbolt 3, but hasn't provided much detail on the implementation. Z370 also retains support for Intel Optane Memory. Even the ME version 11 is listed similarly on Intel's ARK, though we do know that Z370's Management Engine doesn't allow Kaby Lake processors to boot on the platform. Two memory channels support up to two DIMMs per channel and a maximum of 64GB. Intel does not support ECC on any of its models (compared to AMD, which does allow motherboard vendors to enable ECC support).

Intel offers a few unique benefits to enthusiasts, such as the XTU (Extreme Tuning Utility) and support for XMP 2.0. The Performance Tuning Protection Plan is also available, which covers damage from overclocking. The price for the warranty plan varies per SKU, and Intel hasn't provided a detailed breakdown for Coffee Lake yet. We aren't aware of any AMD-equivalent coverage.

Intel UHD Graphics 630

Intel will have an optimized graphics driver soon after launch, but we don't see any significant improvements (beyond HDCP 2.2 support for DP 1.2a) to the Gen 9 LP (Low Power) engine. We still get subslices with eight EUs each (totaling 24 EUs in a GT2 configuration), accelerated slightly by a 50 MHz maximum frequency increase across the board.

The biggest change comes from Intel's marketing department, which "upgraded" the High Definition (HD) Graphics 630 brand to Ultra High Definition (UHD) Graphics 630. Apparently, that means the company supports the same legacy VP8 and AVC codecs, HEVC 10-bit decode/encode, VP9 8/10-bit decode, VP9 8-bit encode (no support for VP9 10-bit encode), HDR, and Wide Color Gamut features. The processors now come in 6+2 (i5 and i7) and 4+2 (i3) configurations.

MORE: Best CPUs

MORE: All CPUs Content

How We Test

The MSI Z370 Gaming Pro Carbon AC

MSI continues its Z-series Gaming Pro motherboards, giving the latest implementation similar features as previous models. Pricing should also end up similar. The exact board details and specifications can be found on the manufacturer's website.

Comparison Products

AMD Ryzen 7 1800X

AMD Ryzen 5 1600X

Intel Core i7-7800X

Test Systems

We introduced our new test system and methodology in How We Test Graphics Cards. If you'd like more detail about our general approach, check that piece out.

In this case, only the hardware configuration with CPU, RAM, mainboard, as well as the new cooling system are different, so the summary in table form gives a quick overview of the systems used:

Test System and Configuration
Hardware	Germany Intel Socket 1151 (Z370):Intel Core i7-8700KMSI Z370 Gaming Pro Carbon AC2x 8GB Corsair Vengeance DDR4-3200 @ 2666Intel LGA 2066Intel Core i7-7800XMSI X299 Gaming Pro Carbon AC4x 4GB G.Skill RipJaws IV DDR4-2600AMD Socket AM4 WorkstationAMD Ryzen 7 1800X, Ryzen 5 1600XMSI X370 Tomahawk4x 8GB G.Skill TridentZ DDR4-3200 @ 2667 and 3200 Intel LGA 1151 (Z270)Intel Core i7-7700K, Core i5-7600KMSI Z270 Gaming 72x 8GB Corsair Vengeance DDR4-3200 @ 2400 and 3200All SystemsGeForce GTX 1080 Founders Edition (Gaming)Nvidia Quadro P6000 (Workstation)1x 1TB Toshiba OCZ RD400 (M.2, System)2x 960GB Toshiba OCZ TR150 (Storage, Images)be quiet! Dark Power Pro 11, 850W Power SupplyWindows 10 Pro (Creators Update)USIntel Socket 1151 (Z370):Intel Core i7-8700KMSI Z370 Gaming Pro Carbon AC4x 8GB G.Skill RipJaws V DDR4-3200 @ 2666 and 3200 Intel LGA 2066Intel Core i7-7800XMSI X299 Gaming Pro Carbon AC4x 8GB G.Skill RipJaws V DDR4-3200 @ 2666 and 3200 AMD Socket AM4 AMD Ryzen 7 1800X, Ryzen 5 1600XMSI X370 Xpower Gaming Titanium2x 8GB G.Skill RipJaws V DDR4-3200 @ 2667 and 3200 Intel LGA 1151 (Z270) Intel Core i7-7700K, Core i5-7600K MSI Z270 Gaming M72x 8GB G.Skill RipJaws V DDR4-3200 @ 2666 and 3200 All EVGA GeForce GTX 1080 FE 1TB Samsung PM863 SilverStone ST1500, 1500W Windows 10 Creators Update Version 1703
Cooling	GermanyAlphacool Eiszeit 2000 ChillerAlphacool Eisblock XPXThermal Grizzly Kryonaut (For Cooler Switch)
Monitor	Eizo EV3237-BK
PC Case	Lian Li PC-T70 with Extension Kit and Mods Configurations: Open Benchtable, Closed Case
Power Consumption Measurement	Contact-free DC Measurement at PCIe Slot (Using a Riser Card) Contact-free DC Measurement at External Auxiliary Power Supply Cable Direct Voltage Measurement at Power Supply 2x Rohde & Schwarz HMO 3054, 500MHz Digital Multi-Channel Oscilloscope with Storage Function4x Rohde & Schwarz HZO50 Current Probe (1mA - 30A, 100kHz, DC) 4x Rohde & Schwarz HZ355 (10:1 Probes, 500MHz) 1x Rohde & Schwarz HMC 8012 Digital Multimeter with Storage Function
Thermal Measurement	1x Optris PI640 80Hz Infrared Camera + PI Connect Real-Time Infrared Monitoring and Recording
Acoustic Measurement	NTI Audio M2211 (with Calibration File, Low Cut at 50Hz) Steinberg UR12 (with Phantom Power for Microphones)Creative X7, Smaart v.7 Custom-Made Proprietary Measurement Chamber, 3.5 x 1.8 x 2.2m (L x D x H) Perpendicular to Center of Noise Source(s), Measurement Distance of 50cm Noise Level in dB(A) (Slow), Real-time Frequency Analyzer (RTA) Graphical Frequency Spectrum of Noise

MORE: Best CPUs

MORE: All CPUs Content

VRMark, 3DMark & AotS: Escalation

Test Notes

We generated some odd performance results with Intel's Core i7-8700. Despite more conservative specifications and a lower price, it outperformed the flagship Core i7-8700K in some of our gaming benchmarks. We verified our numbers on motherboards from different vendors, and those vendors confirmed that they're seeing what we see.

Initially, our attention turned to Turbo Boost, based on observations from Core i9-7980XE and its unpredictable behavior. But after careful analysis with multiple utilities, we believe the Turbo Boost algorithms are working correctly for both processors.

After recording higher package and core power consumption on the 65W Core i7-8700 compared to Intel's 95W -8700K in certain situations, we also reached out to Intel for comment. Once we have some answers, you can be sure we'll post our findings.

VRMark & 3DMark

We aren't big fans of using synthetic benchmarks to measure game performance, but 3DMark's DX11 and DX12 CPU tests provide useful insight into the amount of horsepower available to game engines.

VRMark plays well to Intel's frequency and IPC throughput advantages, and despite Core i7-8700K's lower base clock rate, its higher Turbo Boost frequency helps secure a victory over Core i7-7700K. Locking both processors to the same overclocked settings yields a practical tie.

Synthetics that favor more cores, such as the DX12 Time Spy and DX11 Fire Strike benchmarks, highlight the possibly explosive performance gains we could see from Intel's six-core Coffee Lake models in heavily-threaded games. The -8700K handily dispatches its predecessor. It even beats the high-end Core i7-7800X on an expensive X299 motherboard.

The eight-core Ryzen 7 1800X proves to be a formidable match. Still, we record a smaller delta between the tuned Coffee Lake and Ryzen processors than expected, given AMD's two-core advantage. It looks like Intel's higher clock rates and superior per-clock performance help in this case.

The impressive Ryzen 5 1600X easily beats Intel's Core i5-7600K, which languishes at the bottom of the chart due to its now-stingy quad-core design. The new Coffee Lake-based i5s, wielding 50% more cores, should make that price point more interesting.

Ashes of the Singularity: Escalation

Ashes of the Singularity is a heavily-threaded title that favors the Core i7-8700K's extra cores compared to Core i7-7700K.

Intel's latest outstrips the competition at both stock and overclocked settings, beating Core i7-7700K by a considerable margin in both configurations. Unfortunately, performance in most games doesn't scale linearly, so the -8700K's 50%-higher core count only adds up to a 14% lead out of the box. Nevertheless, that jump is enough for Core i7-8700K to snatch the lead away from Ryzen 7 1800X.

MORE: Best CPUs

MORE: All CPUs Content

Civilization VI, Battlefield 1 & Dawn of War III

Civilization VI AI Test

Our Civilization VI AI test measures performance in a turn-based strategy game, and it tends to favor high frequencies and physical cores.

The -8700K's Turbo Boost frequencies yield surprisingly good results. At stock settings, the processor essentially ties Core i7-7700K, but leads a little more definitively after tuning.

Core i5-7600K fares respectably in Civilization VI at its stock settings. But after a 5 GHz overclock, it yields a 15.1-second result (not pictured) that even beats Intel's fastest CPUs.

Civilization VI Graphics Test

The Ryzen processors are much more competitive during Civilization VI's graphics test, but Intel's overclocking headroom proves insurmountable. Compared to Core i7-7700K, the -8700K offers 3.9% more performance in stock form, and a 3% advantage after tuning. Those slim leads mean there is still good value in a Kaby Lake-based platform, particularly in light of the higher price on Coffee Lake K-series SKUs.

Battlefield 1 (DX11)

A graphics bottleneck limits the frame rates of our Battlefield 1 benchmark, so with the exception of Ryzen 5 1600X and Core i5-7600K, the differences between models are slight.

Intel's Core i7 processors lead. More specifically, though, the -8700K does beat its predecessor.

It’s easy to see the workload's more threaded nature during its opening phase, where a quad-core Core i5-7600K trails the test pool.

Warhammer 40,000: Dawn of War III

Dawn of War III scales well on CPUs with more execution resources. Naturally, then, the Core i7-8700K dominates in stock and overclocked trim.

The Core i7-7800X posts another surprisingly low score. Considering the costs associated with Intel's premium X299 platform, Skylake-X generally doesn't make much sense for gaming.

MORE: Best CPUs

MORE: All CPUs Content

Grand Theft Auto V, Hitman & Shadow of Mordor

Grand Theft Auto V

Grand Theft Auto V continues to favor Intel architectures and, more generally, multi-core designs. The Core i7-8700K offers 7.6% more performance than the -7700K at stock settings. This grows to a 11.5% lead after some tuning.

Kaby Lake- and Coffee Lake-based Core i7 processors set themselves apart from the pack during our benchmark, specifically offering a 16-25% performance advantage over the Ryzen models at stock settings.

Hitman (2016)

Core i7-7800X finally shows up in the top half of our chart thanks to its effective architecture and 6C/12T configuration. It's still no match for the -8700K, though, probably due to its more conservative 4 GHz Turbo Boost ceiling.

The new -8700K offers the best performance in stock form, demonstrating a 9 FPS advantage over its nearest competitor. Meanwhile, Core i7-7700K turns the tables with a slight lead after overclocking. Given their similar architectures and manually-defined frequencies, a little variance there is expected. Consider the chips to be tied after tuning.

Middle-earth: Shadow of Mordor

The Core i7-8700K surprisingly falls behind its predecessor at stock settings. We examined each chip's Turbo Boost behavior and discovered that the -8700K ran through at a higher frequency. But, like the -7700K, it doesn’t reach peak frequencies. Rather, it drops to lower clock rates throughout our run.

As mentioned at the outset of our performance testing, we saw higher benchmark numbers from the lower-end Core i7-8700 in several games. Coffee Lake and its Z370 platform are new, so perhaps firmware tuning will iron out some of these issues over time.

MORE: Best CPUs

MORE: All CPUs Content

Project CARS, Far Cry Primal & Rise of the Tomb Raider

Project CARS

There's a solid scaling trend from Intel's disappointing Core i7-7800X up through its impressive -8700K.

Core i7-8700K does offer better performance than -7700K out of the box, but as you get more graphics-bound, the two increasingly converge. That means enthusiasts specifically focused on gaming will see little benefit from upgrading, particularly if they own lower-end graphics cards.

Far Cry Primal

Far Cry Primal is another title that doesn't respond as well to Core i7-8700K as we would expect. Again, lower performance than the -7700K at stock settings implies that we just aren't getting the full benefit of Coffee Lake yet. It does gain some ground when we lock both CPUs at 4.9 GHz, but -8700K fails to completely overtake Kaby Lake.

The Ryzen processors obviously trail Intel's offerings in this game and many others. But they're still solid value plays benefiting from lower prices and more affordable platforms. One of Ryzen's greatest advantages is higher performance in threaded workloads like rendering. Coffee Lake challenges this with more cores, so our content creation benchmarks could upset the narrative we've been telling since Ryzen's launch.

Rise of the Tomb Raider

Core i7-7700K and -8700K offer strong, but similar performance in Rise of the Tomb Raider. The Core i5-7600K bests them both at stock settings, though.

MORE: Best CPUs

MORE: All CPUs Content

DTP, Office, Multimedia & Compression Performance

Although we usually don’t run our application benchmarks on overclocked processors, we're including the two Coffee Lake CPUs at stock and overclocked frequencies this time, since we wanted to know how Intel's architecture scales in different scenarios.

To be fair, we also added a (reasonably) overclocked AMD Ryzen 7 1800X to our results. This makes for an interesting comparison. We also included Intel’s Core i7-7800X, which represents another 6C/12T contender from Intel using its X299 platform. Skylake-X didn't seem to add anything in our gaming benchmarks; can it beat Coffee Lake in productivity applications?

DTP & Presentation

Adobe’s Creative Cloud gives us a look at real-world single- and multi-core performance. As such, it beats synthetic benchmarks as a productivity test.

After Effects CC is a classic when it comes to parallelized tasks, with the number of cores easily being more important than frequency. Core i7-8700K dominates in this one, of course.

Otherwise, in lightly-threaded benchmarks, Coffee Lake only beats its competition thanks to its slightly higher clock rate.

Strangely, Intel’s Core i5-7600K reliably beats the stock Core i7 CPUs in Adobe’s InDesign CC. We have no idea why this might be the case.

Encoding & Multimedia

Intel’s new Core i7-8700K dominates our default HandBrake benchmarks. The Core i7-7800X can’t compete.

The finishing order shifts around a bit under our more demanding high-quality test. Due to its higher core count, AMD’s Ryzen 7 1800X finishes ahead of Core i7-8700K, which, in turn, beats the Skylake-X-based Core i7-7800X.

Compression & Decompression

Core i7-8700K’s six cores, coupled with high clock rates, allow Coffee Lake to leave its competition in the dust. Again, the Core i7-7800X doesn’t even come close.

Things don’t get any better for the competition once it's compared to Core i7-8700K in our decompression metric. Core i7-7800X fares especially badly, demonstrating that the X299 platform isn't universally better just because it's more expensive.

MORE: Best CPUs

MORE: All CPUs Content

2D & 3D Workstation Performance

2D Workstation Performance

Our GDI/GDI+ tests are used to test two different output methods that can be found in older applications and printing tasks. Today, they, or at least a modified version of them, are commonly used to display the graphical user interface (GUI). They are also great benchmarks for direct device write throughput and memory performance when handling gigantic device-independent bitmap (DIB) files.

Synthetic 2D Benchmarks

We take a look at direct device write throughput first. The graphics driver uses the CPU heavily for this task, but doesn’t employ many threads.

There hasn’t been true 2D hardware acceleration since the introduction of the unified shader architecture, after all. Microsoft's Windows driver model provides a huge obstacle for 2D hardware acceleration as well.

We up the ante by introducing memory to the mix. This is done with the help of the only remaining 2D hardware function: generating the graphics output in memory and then copying it to the output device all at once. The benchmark’s the same as before. We just plot a bitmap in memory, as opposed to sending the information directly to the monitor. The bitmap’s copied to it only once it’s complete. This pushes the CPUs, since they’re no longer platform-bound. The results prove interesting: frequency rules, AMD can keep up, and Skylake-X brings up the rear.

AutoCAD 2016 (2D)

Even though AutoCAD does use DirectX, ultimately it just duplicates every single draw function in software. The results are exactly as expected, and IPC throughput is emphasized due to AutoCAD’s limited scaling with additional cores.

3D Workstation Performance

Most professional development applications have been optimized and compiled with Intel CPUs in mind. This is reflected in their performance numbers. Still, we include them in order to motivate developers to focus their efforts on AMD’s Ryzen processors as well. This would give users more than one choice. The same goes for an emphasis on multi-core processors, at least where that’s feasible and makes sense.

AutoCAD 2016 (3D)

Clock rate trumps core count. Intel’s Kaby Lake and Coffee Lake land fairly close to each other, with clock rate determining the winner. AutoCAD’s performance turns out to be close to that of older games, since it uses DirectX and isn’t really optimized to take advantage of multiple cores.

Cinebench R15 OpenGL

Frequency is (almost) everything in the Cinebench R15 OpenGL benchmark. However, Intel’s Core i7-7800X does better than usual.

Our overclocked Ryzen 7 1800X lands behind the stock configuration, regardless of how many times we re-run the benchmark, and we don't have a good explanation as to why.

SolidWorks 2015

SolidWorks 2015 also emphasizes clock rate. At the same frequency, Coffee Lake and its predecessor end up in the same place. This doesn’t come as a surprise though, since SolidWorks 2015 typically doesn’t use more than four cores. The exceptions are a few very specific tasks, which we’ll see tested when we get to the CPU composite score on the next page.

Creo 3.0

Creo 3.0 paints a similar picture; a high core count just doesn’t provide any benefits when it comes to drafting using real-time 3D graphics output.

Blender & 3ds Max (Real-time 3D Preview)

The Blender and 3ds Max real-time 3D previews yield similar results: frequency is everything. Of course, final rendering is a different story, and we'll get to that shortly.

The 3ds Max results aren’t based on time to completion. Rather, this benchmark generates a composite index based on CPU performance during a set time period.

Catia V6 R2012

This is one of the graphics benchmarks that has been optimized time and again (it’s part of the free SPECviewperf 12 suite). However, it still provides a fairly good measure of CPU performance, with an emphasis on clock rate. Based on the previous benchmark results, you can guess how this story ends.

Maya 2013

At the risk of beating a dead horse, our chart paints a picture we've seen several times already. The real-time 3D output numbers don’t tell a complete story, though. As we're about to see, core count reigns supreme when it comes to final rendering.

What we can say is that two additional cores don't hurt Intel's Coffee Lake-based flagship. That's good news for a six-core chip trying to prove itself against a smaller quad-core design. The test results should only get better from here.

MORE: Best CPUs

MORE: All CPUs Content

CPU Computing & Rendering Performance

CPU Workstation Performance

The 3D graphics performance we just measured isn’t all that matters to professional rendering software. Applications run many other tasks (like simulations, compute jobs, preview rendering) on the CPU simultaneously. The full picture’s only achievable by looking at both of them together.

Many modern suites include modules that are based exclusively on computing and simulations. This means we need to go beyond just 3D workstation performance to form our opinion of these high-end CPUs. However, software packages like SolidWorks don’t scale perfectly based on core/thread count. Consequently, even quad-core processors keep up if they run at high-enough frequencies and support SMT. We'll illustrate this by comparing Core i7-8700K to the Core i7-7700K.

Frequency is all that counts in Creo 3.0, so long as you have at least eight threads to throw at the application. As a result, Intel’s new Core i7-8700K doesn't really benefit from its two additional cores.

Clock rate and core count both matter in 3ds Max 2015, allowing Intel’s Core i7-8700K to pull ahead of the -7700K by a larger margin.

Separately, the Core i7-7800X is beaten soundly due to a combination of older architecture and lower clock rate.

The CPU composite score includes rendering, which is broken out into its own section below. There, AMD's Ryzen 7 1800X does really well. It even beats the overclocked Core i7-8700K in stock form.

CPU Performance: Photorealistic Rendering

Final rendering doesn’t require a CPU that's good at everything. Rather, this task wants efficiency and fast parallel computation.

We know big, expensive workstation-class CPUs like Ryzen Threadripper tear these benchmarks to bits. But in the -8700K's more desktop-friendly price range, it goes head-to-head with Ryzen 7 1800X in stock form. Core count matters more than frequency, as you can see comparing -8700K to -7700K.

The console version of LuxRender affirms what we saw in 3ds Max; Core i7-8700K does battle with AMD's Ryzen 7 1800X. Though it trails, remember that we're talking about six cores against eight.

Last, but not least, we take a look at Blender. The usual workload (with a sample size of 200 pixels) confirms what we saw in the preceding benchmarks. Intel’s Core i7-8700K and AMD’s Ryzen 7 1800X finish at the top of our chart. Ultimately, the winner is the CPU with more cores, though Coffee Lake keeps up by virtue of its effective architecture and high clock rate. Meanwhile, our Kaby Lake-based chips trail even the Ryzen 5 sample.

The results obtained from SPECwpc’s Blender loop look similar, even though this benchmark presents a somewhat different task consisting of more than just rendering.

With the workload's rendering portion easing up, a stock Core i7-8700K suddenly becomes more competitive.

This trend intensifies as our benchmark begins incorporating tasks other than photorealistic rendering. Core count isn't the sole determiner of performance; IPC throughput factors in as well. That's why you see older quad-core models with SMT rise through the ranks.

Now Coffee Lake enjoys a definite lead. Core count is a factor, to be sure (Ryzen 7 and Skylake-X beat Kaby Lake), but clock rate is also important.

Intel’s Core i7-8700K offers acceptable performance in semi-professional applications; it's not just a gaming CPU. Depending on the application and how it utilizes host processing resources, you'll either get modest results or a stellar experience. There's no such thing as bad.

We didn't expect such strong numbers from Coffee Lake in these benchmarks, particularly compared to Core i7-7800X. It's hard to imagine why anyone would spend $380 on that CPU and a hefty premium on an X299 motherboard.

MORE: Best CPUs

MORE: All CPUs Content

Scientific & Engineering Computations, & HPC Performance

For these tests, we’re using the SPECwpc benchmark suite for workstations with its wide variety of tasks. It tests a number of very different mathematical computations optimized for parallelization. They typically make heavy use of available memory bandwidth and cache, plus expose issues with latency.

Rodinia

The pre-Euler3D CFD test (Computational Fluid Dynamics benchmark) runs very well on AMD's CPUs, suggesting scaling based predominantly on core count. Intel's overclocked CPUs do pick up some performance compared to the stock configurations, but they can't keep up with Ryzen 7 1800X.

Convolution

In this benchmark, a mathematical operation is performed on two functions (convolution), which results in a third function. Performance scales according to core count. Clock rate has far less impact.

CalculiX

This test is based on the finite element method for three-dimensional structural computations. Intel’s higher frequencies help Core i7-8700K beat out Ryzen 7 1800X, though core count clearly matters as well.

Poisson's Equation

Poisson's Equation is a second-order partial differential equation widely used in physics for boundary value problems.

Intel’s new six-core processors fare well. Higher clock rates don't appear to help much, given our overclocked results.

Sequential Reweighted Message Passing (SRMP)

These are algorithms for discrete energy minimization. The workload benefits from core count, clock rate, and architectural improvements, it appears. For some reason, though, AMD's Ryzen 7 and 5 just don't show well, even though our logs show them to be fully utilized.

Kirchhoff Migration

The earth’s subsurface structure can be determined via seismic processing. One of the four basic steps in this process is the Kirchhoff Migration, which is used to generate an image based on the available data using mathematical operations.

This benchmark and its underlying computations turn out to be a great fit for AMD. Intel's Core i7-8700K only takes the top spot after aggressive overclocking, and that wouldn't be practical for everyday use. Ultimately, AMD’s Ryzen 7 1800X wins in stock trim.

In the end, Core i7-8700K performs well in almost every scenario, especially when the workload scales according to core count and clock rate. Higher frequencies help Intel fend off AMD's eight-core competition, though Coffee Lake represents a compromise in certain tasks.

MORE: Best CPUs

MORE: All CPUs Content

Overclocking, Cooling & Temperature

Overclocking

First, we established the Core i7-8700K’s limits by cooling the chip's IHS to a constant 20°C, side-stepping potential thermal bottlenecks imposed by Intel's unfortunate use of thermal paste between the die and heat spreader. We made it all the way to 5.0 GHz without any problems. This didn't surprise us, given our experiences with Kaby Lake. At 5.1 GHz, we booted into Windows and ran a couple of games, but the processor called it quits when we tried to run Cinebench.

We’d like to point out the possibilities enabled by manual load-line calibration. Depending on the motherboard, different levels and presets are available, or the voltages can be manually adjusted for some experimentation. In this way, the core voltage can be reduced significantly without losing a whole lot of performance (so long as your chip plays along). For the motherboard we tested, the effective voltage was between 1.18V and a maximum of 1.28V when running Prime95 at stock frequencies. This lowered the package’s temperature by almost 8°C.

Unfortunately, many memory kits run into trouble if the load-line calibration is set too low; the result is general instability. CPU quality plays a key role as well. In our particular case, this affected our 5.0 GHz overclock, which didn’t hold up over time. All of the games and some of the workstation applications ran for hours without any problems, but Creo 3.0 and some of the HPC tests crashed after a few minutes.

The 5+ GHz overclocking stories are exciting, to be sure. But remember that most of them aren't validated for long-term reliability. We'd rather drop the clock rate by 100 MHz and not have to deal with intermittent crashes.

Looking at our power consumption and performance graph, we see a bend at ~4.8 GHz. Power use continues increasing with higher clock rates, but the Cinebench score levels off. A failure to continue scaling at 5.0 GHz is a good indicator that our CPU is throttling. It simply cannot dissipate heat quickly enough.

Power Consumption

At idle, the differences in power consumption are fairly marginal. All of the processors end up just about where we'd expect based on previous reviews. AMD's Ryzen processors draw significantly more power than the Intel competition because their idle clock rate is a bit too high.

The new Intel CPU’s average power consumption in applications that combine 2D and 3D loads (like AutoCAD) is in line with the performance we observed.

The processor ends up in almost the same place during our gaming loop.

The finishing order changes dramatically once we fire up an AVX stress test with all cores running at their top Turbo Boost bins. During rendering, we were seeing the -8700K's stock power consumption at 110W, climbing to 133W under a 5 GHz overclock.

AVX without offset pushes the result as high as 170W. The Core i7-8700K at 4.9 GHz even throttles due to its package temperature. And that's in spite of our compressor cooler's efforts! Thermal paste under the IHS does us no favors.

Temperatures

Here’s the good news: unless you render or run Prime95 for hours on end, a good air cooler can theoretically handle 4.8 GHz in a well-ventilated case. Intel’s thermal interface material isn't desirable, but it shouldn't stop you from achieving a decent overclock.

The above graph shows that a closed-loop liquid cooler is able to keep an overclocked Core i7-8700K from throttling after 20 minutes of warming up. A good heat sink and fan combination should perform almost as well, again, given ample airflow.

Under our stress test, the overclocked processor gets uncomfortably hot, even under our compressor cooler.

While we're only measuring an average of 170W, thermal throttling keeps the 180W+ peaks from becoming our average power consumption result. At that point, even the most powerful coolers have to throw in the towel.

To be sure, it's surprising just how much power such a tiny processor can consume once it’s pushed to its limits. Nevertheless, Intel’s Core i7-8700K is relatively easy to cool, even on air. You'll just want to stay away from taxing rendering sessions and AVX-optimized workloads. At that point, you're best off with an all-in-one closed-loop liquid cooler.

MORE: Best CPUs

MORE: All CPUs Content

Conclusion

AMD’s competitive architecture, disruptive pricing, and class-leading core counts continues to turn heads. Intel responded to the Threadripper family by introducing Skylake-X-based CPUs for less money per core than we've ever seen before. Up until now, though, it showed no interest in affecting that metric with its mainstream desktop chips. This changes with Coffee Lake, which employs the same underlying architecture as Kaby Lake, but adds execution cores and cache. Improved Turbo Boost bins also help maintain performance in lightly-threaded tasks.

Intel says that its Core i7-8700K is the company's best gaming processor ever. So, we use a geometric mean of 99th percentile frame times, a good indicator of smoothness, converted into an FPS measurement, to gauge the veracity of this bold claim across our suite. Five of the games we test were released in 2016, and five are older (2014/2015). Extra cores could enable more performance as software evolves, so we also include a chart with newer games that thoroughly utilize available host processing resources. We also have price-to-performance charts that get split up to include both the price of the processor and extra platform costs. For the models that don't come with a bundled cooler, we add an extra $25 for a basic heat sink. We also add $20 if overclocking requires a more expensive motherboard (as is the case for Z370).

Despite a few missteps, Core i7-8700K lives up to Intel’s claims. While it doesn't beat the -7700K by a massive margin, the Coffee Lake flagship does deliver better performance in stock and overclocked form. Of course, adding a Z370 motherboard and competent cooler knocks you over the $400 mark, so be ready to pay for that privilege.

Value-seekers have to be asking if Core i7-8700K's price tag is even worth paying, then. After all, you can get Ryzen 7 1800X-class performance out of an overclocked Ryzen 7 1700 for $300 or less. But based on our matrix, Coffee Lake gives you the best performance (furthest to the right) without getting too crazy on price. We're naturally wondering how Core i5-8600K will fare. For now, though, Core i7-8700K is the gaming CPU to own.

Whereas Kaby Lake tends to fall behind Ryzen once we jump out of the games and start looking at rendering/encoding workloads, Coffee Lake's extra cores, higher clock rates, and overclocking headroom help close that gap. Applications previously notorious for going AMD's way are now more hotly contested.

Unfortunately, you will have to buy a new Z370-based motherboard to support Core i7-8700K. Coffee Lake necessarily breaks compatibility with the not-altogether-old Z270 chipset for higher memory data rates and improved power to the CPU's package. And of course, the Z-series platform controller hub is a requisite if you want access to unlocked ratio multipliers. Lower-end B- and H-series chipsets are coming, but not until next year.

We're also disappointed that Core i7-8700K still utilizes thermal paste between its die and heat spreader. Whereas this was a significant issue during our Skylake-X evaluation, though, it's not as problematic on a 95W CPU. You can easily stave off throttling with a heat sink and fan, or dial in a respectable overclock under a closed-loop liquid cooler.

Intel has its 10nm Cannon Lake processors coming in the second half of 2018, and AMD has a Ryzen refresh cycle coming next year. Knowing this, should you upgrade now or wait for the next wave of hardware? Due to the iterative nature of most updates, we rarely recommend jumping forward one, or even two generations. However, if you routinely find yourself running productivity workloads that might be served well by Core i7-8700K's extra cores, we could see replacing a quad-core chip with six cores. Gamers interested in maximum performance or streaming also stand to benefit, though in a world of single-GPU graphics configurations, you'd be hard-pressed to bottleneck an overclocked -7700K with even a GeForce GTX 1080 Ti (particularly at the high resolutions it's meant to drive).

It's an exciting year to be an enthusiast. Intel obviously planned to beef up its line-up years back, but we can thank AMD for the accelerated timeline and competitive pricing. Competition truly is a wonderful thing.

MORE: Best CPUs

MORE: All CPUs Content

CPU Performance In VR: 11 Games Benchmarked

Chris Angelini — Thu, 07 Sep 2017 13:00:00 +0000

Introduction

When HTC’s Vive launched at $800 and Oculus’ original incarnation of the Rift surfaced for $600, friends and family made it a point to stop by my place for their first tastes of VR. Most of them loved the experience, but nobody ran out and bought an HMD of their own.

More recently, the Rift + Touch kit went on sale for $400, while HTC permanently dropped the Vive’s price to $600. Suddenly, people I know were taking the leap and asking for help building fast-enough PCs. Most of all, I encouraged, buy as much graphics horsepower as possible.

But what about the platform that beefy GPU lives on? How much muscle do you need backing up your favorite GeForce or Radeon card? Oculus sets the bar low, specifying a Core i3-6100, Ryzen 3 1200, or FX-4350 at minimum. However, the company recommends a Core i5-4590, Ryzen 5 1500X or more. HTC suggests a Core i5-4590 or FX-8350 at least. If only there was a way to quantify the benefit of stepping up from entry-level to a more potent host processor...

As it turns out, we’ve already done a fair bit of work to establish a toolset and methodology for benchmarking PC hardware in virtual reality. If you haven’t already read our primer, check out FCAT VR: GPU And CPU Performance in Virtual Reality. That piece introduces the VR rendering pipeline, two approaches to collecting performance data, the ways we can present it, and it introduces our first batch of results. We showed how Oculus’ asynchronous spacewarp technology works, how quality settings affect a game like Chronos, how Nvidia’s Pascal and Maxwell architectures stack up to each other, and how AMD’s Graphics Core Next architecture compared earlier in 2017.

On one page at the very back of our story, we took a peek at host processor performance in Arizona Sunshine, a game purportedly imbued with special CPU extras for owners of Core i7 CPUs (which of course made it controversial). It turned out that a Core i7-6950X and Core i7-6700K did, in fact, enjoy a performance advantage over Core i5-6600K. And all three Intel chips decimated AMD’s FX-8320.

Eager to expand on those initial findings, we put together five distinct platforms, came up with ways to test 11 different Oculus Rift titles, and talked to some of the developers about the ways they utilized host processing resources in their VR games.

What (And How) We Tested 11 Different Games in VR

Compiling all of the necessary hardware was our first challenge to overcome. Again, we’re an international team, and launch-day hardware gets spread all over the world. A few companies stepped in to help fill in the holes, expressing interest in answering the same questions we were asking.

MSI set up all of our host platforms, providing its X299 Gaming Pro Carbon AC (for Skylake-X), Z270 Gaming Pro Carbon (for Kaby Lake and Skylake), X370 Xpower Gaming Titanium (for Summit Ridge), and 990FXA-GD80 (for Vishera).

The company also sent over a Core i9-7900X for us to use as an ultra-high-end contender. We added our own Core i7-7700K to represent the top of Intel’s mainstream Kaby Lake family, and we purchased a Ryzen 7 1800X to compare the performance of AMD’s Zen architecture. Core i3-6320 and FX-8350 serve as floors, upon which the faster CPUs build.

Given Ryzen’s sensitivity to memory performance, we knew our choice in DDR4 would be scrutinized. G.Skill sent its F4-3200C14D-16GFX FlareX kit to complement the Ryzen 7 1800X and its F4-3200C14Q-32GTZ kit for our other DDR4-based configurations. Both were set to 3200 MT/s for testing.

We used a F3-2133C10Q-16GXM Ripjaws X kit at 2133 MT/s to go with AMD’s FX-8350. In this way, we were able to maximize throughput on every platform. The CPUs with dual-channel memory controllers were limited to 16GB (from one DIMM per channel), while the X299 setup featured 32GB (allowing the same one DIMM per channel).

In an effort to give each platform comparable thermal performance, we approached Corsair about a high-end closed-loop solution that we could use on Skylake-X, Socket AM4, LGA 1151, and Socket AM3+. The company sent over its Hydro-series H110i, which not only fit all of our test platforms, but also facilitates the cooling needed to keep our Core i9 from throttling.

Everything else was held constant. We used a GeForce GTX 1080 Ti to alleviate graphics bottlenecks as much as possible, a 500GB Crucial MX200 SSD, and the familiar be quiet! Dark Power Pro 10 850W PSU. Windows 10 was installed fresh and completely updated before we started downloading games from Oculus’ store.

Test Equipment
Cooling	Corsair H110i
CPU	Core i9-7900XCore i7-7700KCore i3 6320Ryzen 7 1800XFX-8350
Graphics	EVGA GTX 1080 Ti
Memory	Flare X 16GB DDR4-3200Trident Z (32GB)
Motherboard	MSI X299 Gaming Pro Carbon ACZ270 Gaming Pro CarbonX370 XPower Gaming TitaniumMSI 990FXA-GD80
PSU	be quiet! Dark Power Pro 10 850W
Storage	MX500 SSD

We still have two PCs sitting side-by-side able to collect data using the hardware- or software-based approaches to FCAT VR. Our primer established the software version’s efficacy, though, so we’re using that utility exclusively to save time and provide insight not otherwise available from video-based analysis (such as unconstrained frame rate, calculated from real frame time measurements).

Again, if you’re interested in learning more about hardware performance in VR and want to get the most out of today’s deep-dive, FCAT VR: GPU And CPU Performance in Virtual Reality is the best place to start.

MORE: Best CPUs

MORE: All CPUs Content

Arizona Sunshine

Arizona Sunshine is a great place to start—not only does it allow us to pick up where we left off in our VR performance primer, but it also comes first alphabetically in today’s 11-game suite.

Our 150-second test starts on the bridge towards the game’s beginning, right as you pick up a second handgun. We run through abandoned cars on the highway, blasting zombies along the way, stopping just before the roadblock leading to the mine.

The FCAT VR output makes two observations obvious on its own. Most clear is the FX-8350’s struggle. It actually hamstrings our GeForce GTX 1080 Ti, keeping the card from rendering 90 real frames per second on its own. Instead, Oculus’ runtime is forced to step in and synthesize frames to prevent stuttering. As a result, 41% of our test sequence’s 13661 frames are the product of asynchronous spacewarp.

Our look at CPU performance in Arizona Sunshine six months ago showed the FX-8320 suffering similarly, though the more taxing test we use today goes a step further to illustrate FX’s shortcomings in games.

Second, the Core i7-7700K’s frame time over time plot (in red) looks better than the rest of the field. Although the Core i9-7900X, Core i7-7700K, Core i3-6320, and Ryzen 7 1800X suitably support our GeForce GTX 1080 Ti in its quest to maintain 90 real frames per second, render times lower than 11.1ms per frame reduce the chance of an errant drop. So, you want as much headroom under there as possible.

Indeed, if we take the frame time results measured by FCAT VR and convert them into a theoretical “unconstrained frame rate”—the performance you’d see if v-sync didn’t force the Rift's output to 90 Hz—Core i7-7700K does emerge a winner.

Vertigo Games didn’t respond to our request for comment, so we can’t tell you just how extensively the developer optimized for threading. But we do know that Arizona Sunshine employs the Unity engine, does not support GPU-accelerated PhysX, but does offer more sophisticated physics effects through the “Advanced CPU Extras” checkbox, which we enable.

Beyond four Hyper-Threaded cores, it looks like the highest clock rate and IPC throughput yields the best unconstrained frame rate. The Core i9’s extra cores don’t help, and we can assume its previous-gen architecture isn’t an asset against Kaby Lake, either.

Ryzen 7 1800X roughly matches the Core i3-6320. And if you’re dabbling with VR on an FX-based platform, you’re doing the experience a serious disservice by limiting your graphics card’s performance.

Core i7 and Core i9 achieve the lowest 99^th percentile frame times, though it’s notable that four of these five platforms land under 11.1ms in that comparison.

Every CPU registers at least some frame time spikes. That’s why you see the “Worst” column jump up quite a bit. Core i9 encounters the fewest dropped frames, while Ryzen 7 fares the worst. Still, 21 dropped frames in a 150-second sequence isn’t subjectively perceptible.

MORE: Virtual Reality Basics

Arizona SunshineView Deal

Chronos

Chronos is one of the Rift’s launch titles, and at the game’s Epic detail settings, it’s one of the most graphically demanding workloads we’ve seen for VR.

Our test sequence lasts 80 seconds, starting from the protagonist’s first moments on the beach and ending before he passes through a specific doorway.

Thanks to the power of our GeForce GTX 1080 Ti, all five platforms avoid dropping into ASW mode through our Chronos benchmark. That doesn’t mean they perform similarly, though—there are clear differences in the frequency of dropped frames, indicated by red spikes in each interval chart.

We again see Intel’s Core i7-7700K achieve the lowest frame times over time. Curiously, it appears the Core i3 lands in second place, followed by the FX we’ve been recommending against. Intel’s Core i9 and AMD’s Ryzen 7 look like the last-place finishers.

Every platform enables a constant 90 FPS delivery to the Rift. But converting frame times to unconstrained FPS illustrates the differences in headroom each host processor offers.

It’s strange to see the Core i3 landing, as predicted, in second place, with AMD’s FX behind it (and ahead of Core i9/Ryzen). In a graph of CPU utilization over 60 seconds, Chronos only appears to use about 8% of our Core i7-7700K, suggesting that the game doesn’t even fully utilize one of the processor’s cores. It’s possible, then, that the fabrics and meshes touted by AMD and Intel serve to hurt the performance of their massively parallel CPUs by adding latency.

Although AMD’s FX-8350 causes the most dropped frames, Core i9-7900X and Ryzen 7 1800X do demonstrate the worst frame times in our 50^th, 90^th, 95^th, and 99^th percentile measurements. They also present some ugly worst-case frame time spikes that are clearly visible toward the end of our frame time over time line graph.

MORE: Best CPUs

MORE: All CPUs Content

ChronosView Deal

Damaged Core

You have to play through quite a bit of Damaged Core before reaching a point that can be accessed for easy and repeated testing. Our benchmark sequence includes 80 seconds of enemy forces assaulting a reactor. The frame times look sinusoidal because the benchmark involves turning from one side to the other in four-second intervals.

A GeForce GTX 1080 Ti is plenty fast for Damaged Core’s most demanding settings. But this Unreal Engine 4-based shooter still scales based on processor performance.

Frame time oscillations make it hard to tell which CPU is fastest. But we definitely see AMD’s FX-8350 struggle the most with high frame times and troublesome variance. In places, insufficient performance causes the FX to trigger ASW. As a result, ~5% of the run’s frames are synthesized by Oculus’ runtime.

This does serve to shelter the FX from dropped frames, though. We observe 35 drops, while Core i3-6320 drops 46 frames and Ryzen 7 1800X drops 53.

Intel’s Core i7-7700K achieves the highest unconstrained frame rate, leaving lots of headroom under the 11.1ms ceiling that translates to 90 FPS. Core i9 lands in second, telling us that there’s either minimal scaling beyond 4C/8T in this game, or that Core i9’s other architectural changes hurt performance more than its extra cores help.

High frame times set FX-8350 apart from the rest of our field in each percentile measurement. And it’s those spikes in the beginning of our benchmark that determine the worst-case peaks.

MORE: Virtual Reality Basics

Damaged CoreView Deal

EVE: Valkyrie

For many early adopters of Oculus’ Rift, EVE: Valkyrie was the experience that got them enthused about VR. CCP Games continues updating EVE, most recently with beta Touch support.

We ran our tests after the developer added Ultra-quality settings, applying volumetric lighting, multi-sample G-buffer anti-aliasing, specular highlights, Lens-Matched Shading, and Multi-Res Shading. These features were added specifically for owners of Nvidia GeForce GTX 1070, 1080, and 1080 Ti cards.

The Station 27 chronicle we used for testing isn’t particularly taxing, but it’s far more consistent than a chaotic dogfight would be. Consequently, we observe fairly even frame times from each platform.

All five CPUs work well enough with the GeForce GTX 1080 Ti to enable >90 FPS throughout our 200-second recording. The Core i9, Core i7, and Ryzen 7 deliver particularly clean-looking runs, while Intel’s Core i3 encounters a couple of additional dropped frames. FX-8350 drops 64 frames during the course of its benchmark. This doesn’t come as much of a surprise, given higher frame time variance and a propensity for breaking above the 11.1ms mark.

That brings up an interesting point: how are we seeing frame times in excess of 11.1ms from the FX-8350, but still experiencing 90 frames per second without dropping into ASW mode? If you look back at our Chronos results, the same thing happened. We did explore this phenomenon in FCAT VR: GPU And CPU Performance in Virtual Reality and determined that “...the envelope can expand/contract due to preemption/parallelization done by the VR runtime. Oculus' adaptive queue ahead feature is designed to facilitate this, so an 11ms cut-off is not absolute, though it’s generally true. In short, optimizations are not enough to keep you from dropping frames if you render at >11ms for an extended period of time.”

All of the calculated unconstrained frame rates exceed 100 FPS. However, Core i7-7700K’s frame times afford it a 36% advantage over FX-8350, a 22% lead over Core i3-6320, an 18% edge on Ryzen 7 1800X, and even a 9%-better finish than Core i9.

The Core i7, Core i9, Ryzen 7, and Core i3 frame times look good all the way through our 99^th percentile measurements. FX-8350 is consistently ~2ms higher than the Core i3. And those two spikes visible in the frame time over time plot show up in this chart when we single-out each CPU’s worst frame time result.

MORE: Best CPUs

MORE: All CPUs Content

EVE: ValkyrieView Deal

Gunjack

This is another CCP Games title based on Unreal Engine 4. Gunjack was introduced for Gear VR back in 2015 and later adapted to the Rift and Vive in 2016. There are no graphics settings, though, so although CCP Games updated Gunjack with higher-resolution graphics, better textures and effects, and higher-quality audio, it’s still meant to be broadly accessible.

Not surprisingly, then, a GeForce GTX 1080 Ti cuts through this workload easily. Dropped frames are sprinkled through our interval charts, but all five platforms render 90 real frames per second through our 80-second sequence.

Unconstrained frame rates in the 200, 300, and 400 FPS range are uncommon, though the deltas between CPUs look pretty familiar. From top to bottom, Core i9-7900X facilitates ~41%-higher results than the FX-8350. Of course, we actually see all of these output at 90 FPS.

Huge frame rates are indicative of tiny frame times. Indeed, four of our five contenders achieve 99^th percentile frame times under 4ms. Isolated spikes across our frame time over time plot correspond to the “Worst” column, though these are outliers at best.

MORE: Virtual Reality Basics

GunJackView Deal

The Mage’s Tale

The Mage’s Tale doesn’t have adjustable quality settings either. However, this UE4-based title, funded by Oculus VR for the Rift, is definitely more demanding than Gunjack. We capture 80 seconds of the introduction as a benchmark.

Big frame time spikes cause a significant number of dropped frames on AMD’s FX-8350. But these don’t last long enough to engage ASW mode.

Dropped frames dot the other four plots as well, corresponding to big frame time spikes.

Interestingly, that green line peeking out below the others hints at a Core i9-7900X win.

Converting measured frame times to unconstrained frame rates confirms the -7900X’s superior performance. Ryzen 7 1800X also nips close at the heels of Core i7-7700K. And FX-8350 isn’t as far behind Core i3-6320 as we’ve seen in other games.

CPU utilization generally doesn’t jump above 10% or so on a Core i7-7700K, so the Core i9’s advantage probably isn’t a result of its 10 Skylake-based cores. We suspect that The Mage’s Tale responds well to larger L2 caches.

Ninety-five percent of FX-8350’s frames land under 10.1ms. Jumping to the 99^th percentile measurement sees this figure rise to 11.2ms. That’s where a lot of the dropped frames come from. Moreover, extreme spikes as high as 346ms are an order of magnitude more severe than the worst frame times encountered on our Core i3, Ryzen 7, and Core i9 platforms.

MORE: Best CPUs

MORE: All CPUs Content

The Mage’s TaleView Deal

Project CARS

A maximum-detail run through a Nürburgring replay in Project CARS yields a fairly consistent benchmark. This is perhaps the most demanding workload in our suite—even a GeForce GTX 1080 Ti is forced to its knees by extreme levels of anti-aliasing.

All five platforms are hammered by Project CARS’ top detail settings, imposing ASW across our 80-second test sequence.

As we’ve become accustomed to seeing, AMD’s FX-8350 incurs the highest frame times. Occasionally, this means ASW has to synthesize two frames for every real one. A series of spikes in the first quarter of our benchmark cause the same issue on Core i7-7700K.

Those anomalous spikes cause Core i7-7700K to trail Intel’s Core i9 in our average unconstrained frame rate graph. Of course, this came as a surprise to us, so we asked the Slightly Mad Studios team to weigh in on our findings.

From our discussion, it sounds like Project CARS doesn’t specify thread affinity on the PC, so all available threads are, in fact, utilized. However, the benefit of this depends on the granularity of tasks being scheduled and dependencies that might result in a stall. There is, then, a saturation point where the advantage of more cores bottoms out.

Cache utilization can become a factor as well. SMS CTO and technical director Ged Keaveney adds, “we are sensitive to [L2 cache usage] on some of our threads, and that can have a big impact depending on the thread topology the OS scheduler chooses. They tend to only care about available time slices and not cache usage, so patterns can occur where multiple cache-heavy threads can end up on the same cluster and the caches get hit harder.” Taking that information into account, in addition to the Core i7’s troubled first quarter, helps explain how Core i9 might establish its lead.

The jump between our 95^th and 99^th percentile frame times on Intel’s Core i7-7700K is wholly attributable to the benchmark's beginning sequence, culminating in a worst-case 43ms frame. Through most of the test, though, Core i7 is well-behaved.

MORE: Virtual Reality Basics

Project CARSView Deal

Robinson: The Journey

Developed on CryEngine V, Robinson: The Journey offers a number of graphics options that we crank up to tax our GeForce GTX 1080 Ti. Collecting performance data involves running a circular path around the perimeter of Robin’s downed craft, including passes by the power-providing creek and lots of foliage.

Monitoring resource utilization during our run shows Robinson using more host processing power than any other game we’ve looked at thus far. Perhaps that’s why Crytek recommends at least a Core i5-4590.

When we break each platform’s performance out into their separate 90 FPS intervals, we see that AMD’s FX-8350 does fall back on ASW once, briefly. Everything else pumps out 90 real frames per second.

More interesting, perhaps, is Intel’s Core i3-6320. Its frame time plot appears to weave in and out with Ryzen 7 1800X (in other words, it serves up stellar frame rates for a dual-core CPU). But the interval chart is dotted with dropped frames. We might guess that, at certain points, two cores can’t feed the GeForce card fast enough, and a frame is dropped. But this deficit is brief. It doesn’t last long enough for the runtime to kick into ASW. After all, the 90^th percentile frame times are under 11ms. That deeper look at performance is illustrative enough that we’d advise against a dual-core CPU—even a Hyper-Threaded one.

Average frame rates (in this case, a look at unconstrained FPS) often miss important subtleties. It looks like Core i9 and Core i7 fare similarly, followed by Core i3 and Ryzen 7. But as we just established, Intel’s Core i3 has issues keeping up.

Even a look at frame times fails to pinpoint Core i3’s hiccups. It still looks comparable with Ryzen 7.

Counting the dropped frames, however, Core i3 is clearly in a league of its own.

MORE: Best CPUs

MORE: All CPUs Content

Robinson: The JourneyView Deal

Robo Recall

If EVE: Valkyrie is the game that hooked a lot of enthusiasts into spending money on VR, then Robo Recall is the title that reminded them they made a good decision. Ironically, Epic Games released it as a free-to-play title (with the purchase of Touch) based on its Unreal Engine 4.

Our benchmark is 150 seconds long, running from the beginning of the first mission. We use High Graphics Quality, 4x on the Antialiasing MSAA scale, a pixel density of 1.0, and no adaptive resolution. Planar Reflections and Indirect Shadows are both enabled.

FX-8350 posts high frame times and lots of up/down variance. This causes AMD’s previous-gen platform to force our GeForce GTX 1080 Ti into ASW mode for about 23% of the test.

Ryzen 7 1800X has some issues of its own, though. A number of frame time spikes cause a disproportionately high dropped frame count. Core i3 isn’t immune, either.

We asked Epic Games founder Tim Sweeny for any information he might be able to provide about how UE4 handles the latest CPU architectures. He responded that the engine generally provides excellent scaling to four cores, and worthwhile scaling up to 8-10 cores in complex scenes.

He continues, “We'll be improving this significantly over time. Sixteen cores in a consumer CPU was a (very pleasant) surprise to everyone, and there's some good headroom for future optimization. Bottom line, if the CPU makers can keep it up, we have solutions that will scale to lots more cores, given sufficient development time.”

The unconstrained frame rate graph shows Intel’s Core i7 and Core i9 on equal footing, while Ryzen 7 and Core i3—the CPUs that dropped the most frames—line up fairly evenly as well. FX-8350 presents the lowest frame rate; however, more than 3000 synthesized frames help smooth the slower platform’s performance, preventing some of the drops that’d otherwise show up without ASW.

Core i7-7700K’s strong performance is evident from its worst-case frame time: a mere 11.8ms.

Ryzen 7’s issues with frame time spikes aren’t prevalent in the raw data. Ninety-nine percent of the CPU’s frames are rendered in less than 11.8ms. But that last percent includes frames rendered in up to 85ms.

MORE: Virtual Reality Basics

Robo RecallView Deal

Serious Sam VR: The Last Hope

Older gamers will remember Serious Sam from its early beginnings in 2001. The fast-paced shooter translates amazingly well to VR, thanks to Oculus Touch. Incredibly enough, Croteam goes so far as to recommend a Core i7-6800 or equivalent CPU, along with a Radeon R9 Fury or GeForce GTX 1070 graphics card! That’s a $400+ host processor and $400+ GPU.

Despite Serious Sam’s hefty recommendations and our merciless selection of Ultra settings across the board, this game runs smoothly on our GeForce GTX 1080 Ti. To be sure, it’s largely GPU-bound. Only AMD’s FX-8350 needs help from Oculus’ runtime to maintain smooth performance towards the end of our 80-second test, which is when on-screen action is most hectic.

Croteam senior programmer Dean Sekulic confirmed for us that many-core CPUs shouldn’t affect graphics performance much. The Serious Engine 4.5’s multi-threaded renderer is API-agnostic, supporting DX11, DX12, Vulkan, OpenGL, OpenGL ES, and even DX9. But it benefits little beyond a core count of three. The physics and collision system is where more cores help out. Sekulic says he believes it uses all detected cores, minus one.

It’s entirely possible that this physics/collision system, or Skylake-X’s quadrupled L2 cache per core, is responsible for Core i9-7900X’s narrow victory over Core i7-7700K in our derived frame rate measurement.

Core i3 succumbs to Ryzen 7 here (although the Core i3 does drop a few more frames), while FX-8350 presents one figure to sum up the weakness we saw in our frame time over time plot.

MORE: Best CPUs

MORE: All CPUs Content

Serious Sam VR: The Last Hope View Deal

The Climb

The Climb is a CryEngine-based title that Crytek suggests pairing with a Core i5-4590, at least. And if Robinson was any indication, this game might be expected to brutalize our lowest-end CPUs.

In fact, the dropped frames we see on our Core i3-6320 are even worse than they were in Robinson, causing the dual-core chip to average 80 delivered frames per second.

FX-8350 fares worse on paper, as it imposes high-enough frame times on our GeForce GTX 1080 Ti that the runtime kicks over into ASW on several occasions. But the resulting experience is smoother—we prefer the FX’s 1800 synthesized frames to the Core i3’s 1500 dropped frames.

Core i9, Core i7, and Ryzen 7 all appear to perform admirably. Let’s convert those frame times to frame rates once more...

Similar to Robinson, this CryEngine-based title favors Core i9 ever-so-slightly more than Core i7. This time, however, Ryzen 7 lands ahead of Core i3, which is of course hampered by dropped frames.

We know there’s an issue here from Core i3’s 50^th percentile frame time of 11.5ms...and it only gets worse from there. By the time we’re looking at the 99^th percentile, frames take as long as 14.5ms to render. And the worst-case frame time is a sluggish 29ms.

Incidentally, that’s right up there with FX-8350, though the FX-based platform’s other results are even worse. ASW cuts in early to smooth out those less-than-ideal render times, though.

MORE: Virtual Reality Basics

Conclusion

VR gaming generates excitement. But even 18 months after we first reviewed the Oculus Rift, significant barriers to widespread adoption remain.

Most material is the price of VR. You have to buy an HMD. And while the Rift and Vive both cost a lot less than they did in 2016, that’s still an outlay of several hundred dollars. There’s also the investment in a high-end gaming PC. Both Oculus and HTC go out of their way to bring minimum requirements down through technologies like asynchronous spacewarp and asynchronous reprojection, respectively. But there’s a big difference between the “barely cutting it” experience and high-quality VR. Our tests today push maximum-quality settings through a GeForce GTX 1080 Ti. Really, that’s not the best way to isolate platform performance, since graphics bottlenecks stand to mask differences between host processors. When we’re immersed in VR, however, we want beautiful visuals.

Premium content is also in short supply. Those AAA titles that do exist knock this medium out of the park. They’re still pretty rare, though. And because the install base of HMDs is relatively small, developers aren’t making much money creating new games. You’re seeing stakeholders like HTC and Oculus subsidize the hefty development costs in order to get good games out of the door. Once those games exist, a greater number of enthusiasts will see the value in spending hundreds of dollars on new hardware. It’s a chicken-and-egg issue that works itself out slowly, and only after significant investment.

The pace at which this industry moves is breathtaking, though. Some of the developers we spoke with had an idea of how previous-gen platforms behaved under their games. Most could only speculate how newer architectures like Core i9 and Ryzen might fare since they hadn’t gone hands-on yet. And they all seemed to have moved on to new projects already. Clearly, the best is yet to come as those talented studios apply what they learned from the games we tested today.

As is usually the case after a cursory look at performance, a bit of data begs for more. We wanted to go wide on game testing, narrowing the comparison platforms to keep the workload manageable. And while it’s great to see how so many different engines handle such a wide range of hardware, a retrospective look back makes us wish we had Core i5 and a lower-end Ryzen chip to fill in holes in the middle.

The good news for enthusiasts building high-end PCs is that Core i9, Core i7, and Ryzen 7 are all capable of backing a GeForce GTX 1080 Ti. Frankly, though, if your primary focus is VR gaming, the Core i7-7700K really can’t be beaten. It’s a top performer and $90 cheaper than the Ryzen. You could pick up the -7700K and a GeForce GTX 1080 for less than a Core i9, leaving money left over for a couple of games.

Slower CPUs and flagship-class graphics cards obviously create balance issues. It’d be far more likely to pair an older FX or Core i3 with a GeForce GTX 1060 or Radeon RX 580, not the GTX 1080 Ti we used. But then you’re talking dialed-back detail settings in order to maintain playable performance. Until we’re able to generate some data with more mainstream configurations, we’d guess that mid-range GPUs are best paired with Core i5 or Ryzen 5 host processors at a minimum, rather than the baselines Oculus/HTC specify.

If there are specific combinations you’d like to see tested, or if you have questions you’d like to have your favorite developers answer, let us know in the comments—we’ll see what we can do!

MORE: Best CPUs

MORE: All CPUs Content

Intel Core i7-7820X Skylake-X Review

palcorn@outlook.com (Paul Alcorn) — Wed, 26 Jul 2017 13:00:00 +0000

Features & Specifications

What we've seen of Intel's newest HEDT platform so far hasn't inspired much excitement. First, there were complaints of high temperatures and limited overclocking, addressed in The Skylake-X Mess Explored: Thermal Paste And Runaway Power. Then we collectively scratched our heads, wondering what the company was thinking in our Intel Core i7-7740X Kaby Lake-X Review.

But there are plenty of options between the quad-core Core i5 and Intel's 36-thread Core i9 flagship. In fact, the X-series includes nine models this time around, more than any other HEDT family to date. And it's the mid-range Core i7s that we expect to be most popular due to their tamer price points.

In yet another sign of a renewed fighting spirit, Intel's $600 Core i7-7820X slots in below the $1000 Core i9-7900X. That big $400 step down from the 10-core model is uncharacteristic for Intel. Its older eight-core Core i7-6900K bore a shocking $1100 price tag. No doubt, Intel is looking to stave off AMD's Ryzen 7 models. While the $500 "savings" versus its previous generation is certainly nice, however, Intel continues to struggle against AMD's disruptive pricing scheme and looser approach to segmentation.

The Core i7-7820X has eight Hyper-Threaded cores, so comparisons to AMD's Ryzen 7 models are inevitable. The $600 -7820X does battle against the $500 Ryzen 7 1800X. And as a result of unlocked multipliers up and down AMD's portfolio, even the $330 Ryzen 7 1700 is a viable competitor. Intel continues to enjoy an advantage in most lightly-threaded workloads, but the company just can't match Ryzen 7's value, particularly in workloads able to exercise all eight cores. It also helps that AM4-based motherboards are a lot less expensive.

Of course, Core i7-7820X drops into an LGA 2066 interface on motherboards with the X299 "Basin Falls" platform controller hub. We already discussed how processor choice can severely limit this chipset's connectivity in our Core i7-7740X review. And fortunately, Core i7-7820X doesn't suffer nearly as much as Kaby Lake-X. However, you do have to tolerate a "mere" 28 lanes of PCIe 3.0. Last generation, Core i7-6850K in roughly the same price range gave you 40 lanes, so we consider the drop to 28 a regression. Granted, AMD only exposes 16 lanes with Ryzen 7, so Intel does end the PCIe comparison ahead.

Core i7-7820X features a 3.6 GHz base clock that boosts up to 4.3 GHz across two cores in lightly threaded workloads. That's a marked increase over what the Broadwell-E-based Core i7-6900K could do. Further, -7820X supports Turbo Boost Max Technology 3.0, which can push the CPU's two "best" cores up to 4.5 GHz using a piece of installed software. In theory, that should allow Skylake-X to dominate single- and multi-threaded benchmarks alike.

Intel also officially supports up to DDR4-2666 across the -7820X's quad-channel memory controller. Compared to Ryzen 7's dual-channel design, Skylake-X can theoretically move a lot more data, which is useful in certain prosumer applications.

Similar to the Core i9-7900X we already reviewed, -7820X is rated for up to 140W. If you're curious about what that number means to power consumption, heat, and overclocking headroom, check out the aforementioned deep-dive (The Skylake-X Mess Explored: Thermal Paste And Runaway Power) for more.

And if you'd like some more background on Intel's 14nm Skylake-X architecture, we'd encourage you to read through Intel Core i9-7900X Review: Meet Skylake-X, where we introduce the new mesh topology, cache hierarchy (-7820X boasts 8MB of L2 and 11MB of L3), and fresh ISA extensions (unfortunately, -7820X loses one AVX-512-capable unit per core compared to -7900X).

Speed Shift, which allows the processor to handle power-state transitions autonomously, also makes its debut on the high-end desktop. The tactic eliminates latent operating system commands and provides faster resumption times from lower power states. That equates to a snappier experience. Intel also includes support for the vROC (Virtual RAID on CPU) feature that allows you to coalesce up to 20 SSDs into a single bootable volume, though you'll have to buy an upgrade key to unlock it. Intel remains curiously silent on pricing, and keys aren't available yet.

MORE: Best CPUs

MORE: Intel & AMD Processor CPU Comparison Hierarchy

MORE: All CPUs Content

Test Setup & Overclocking

Test Systems

We're using MSI's X299 Gaming Pro Carbon AC for testing. We disabled the all-core Turbo Boost feature to ensure a level playing field.

Skylake-X processors can scale beyond DDR-4000, but we settled on DDR4-3200 for our overclocked config to match the transfer rates of AMD's overclocked Ryzen CPUs.

In the U.S. lab, we attained a 4.6 GHz overclock at 1.28V. This proved stable for 12 hours under AIDA's FPU, cache, and CPU tests. Temperatures hovered ~85°C during the non-AVX load. We were also able to run Prime95 for extended periods, but the processor frequently engaged in aggressive throttling due to thermal constraints. As such, we used an AVX offset to lower the clock rate during AVX-enabled workloads.

Trying to push beyond 4.6 GHz required excessive voltage, which in turn generated even more heat. That's no good with Skylake-X, so we stuck with the safer, more stable settings. Budget in a closed-loop liquid cooler at the very least to keep Core i7-7820X from throttling at stock frequencies. More aggressive overclocking begs for a custom open loop.

We introduced our new test system and methodology in How We Test Graphics Cards. If you'd like more detail about our general approach, check that piece out.

Intel Core i9-7900X

Intel Core i7-7700K

Intel Core i7-6900K

The only updated components in our German lab are the CPU, system memory, motherboard, and new cooling solution, so we'll just provide a quick overview in the following table:

Test Equipment & Environment
System	Germany Intel LGA 2066Intel Core i9-7900X, i7-7740XCore i5-7640X, Core i7-7820XMSI X299 Gaming Pro Carbon AC4x 4GB G.Skill Ripjaws 4 DDR4-2600AMD Socket AM4AMD Ryzen 7 1800X, 1700X, 1600X, 1500XMSI X370 Tomahawk2x 8GB G.Skill Ripjaws V DDR4-3200Intel LGA 2011v3Intel Core i7-6900KMSI X99S XPower Gaming Titanium4x 4GB Crucial Ballistix DDR4-2400Intel LGA 1151Intel Core i7-7700K, i5-7600KMSI Z270 Gaming 72x 8GB Corsair Vengeance DDR4-3200 @2400 MT/sAll SystemsGeForce GTX 1080 Founders Edition (Gaming)Nvidia Quadro P6000 (Workstation)1x 1TB Toshiba OCZ RD400 (M.2, System)2x 960GB Toshiba OCZ TR150 (Storage, Images)Be Quiet Dark Power Pro 11, 850W Power Supply Unit (PSU)Windows 10 Pro (All Updates)Be Quiet Dark Power Pro 11, 850W Power Supply Unit (PSU)Windows 10 Pro (Creators Update)U.S.Intel LGA 2066Intel Core i9-7900X, Core i7-7820XMSI X299 Gaming Pro Carbon AC4x 8GB Corsair Vengeance DDR4-2666 @ 2666 and 3200 MT/sIntel LGA 2011v3Intel Core i7-6900KASRock X99 Extreme44x 8GB Corsair Vengeance DDR4-2666 @ 2666 and 3200 MT/sAMD Socket AM4 AMD Ryzen 7 1800X, 5 1600XMSI X370 Xpower Gaming Titanium 2x G.Skill FlareX DDR4-3200 @ 2666 (stock), and 3200 MT/s Intel LGA 1151 Intel Core i7-7700K MSI Z270 Gaming M7 2x G.Skill FlareX DDR4-3200 @ 2400 All EVGA GeForce GTX 1080 FE 1TB Samsung PM863 SilverStone ST1500, 1500W Windows 10 Creators Update Version 1703
Cooling	GermanyAlphacool Eiszeit 2000 ChillerAlphacool Eisblock XPXThermal Grizzly Kryonaut (For Cooler Switch)USCorsair H115iCorsair H100i v2Arctic MX4
Monitor	Eizo EV3237-BK (Workstation, Office, HPC)
Power Consumption Measurement	Contact-free DC Measurement at PCIe Slot (Using a Riser Card) Contact-free DC Measurement at External Auxiliary Power Supply Cable Direct Voltage Measurement at Power Supply 2x Rohde & Schwarz HMO 3054, 500MHz Digital Multi-Channel Oscilloscope with Storage Function4x Rohde & Schwarz HZO50 Current Probe (1mA - 30A, 100kHz, DC) 4x Rohde & Schwarz HZ355 (10:1 Probes, 500MHz) 1x Rohde & Schwarz HMC 8012 Digital Multimeter with Storage Function
Thermal Measurement	1x Optris PI640 80Hz Infrared Camera + PI Connect Real-Time Infrared Monitoring and Recording

MORE: Best CPUs

MORE: All CPUs Content

VRMark, 3DMark & AotS: Escalation

VRMark & 3DMark

We aren't big fans of using synthetic benchmarks to measure game performance, but 3DMark's DX11 and DX12 CPU tests provide useful insight into the amount of raw horsepower available to the game engine.

The VR test rewards a mixture of per-core performance and parallelism, so the tuned -7820X carves out a small lead over the Core i9-7900X due to its better overclock.

Conversely, the heavily-threaded CPU and physics tests score higher on a 10-core -7900X. The Ryzen processors fare well in metrics optimized for parallelism, but Skylake-X's higher performance per clock cycle goes unmatched.

Core i7-7820X has a higher base and overclocked frequency than the -7900X, which helps explain its performance advantage in the API tests. However, a few extra frequency bins shouldn't overshadow the -7900X's extra cores in our threaded DX12 and Vulkan tests. Perhaps fewer cores generating traffic across the mesh is a good thing in latency-sensitive workloads?

Ashes of the Singularity: Escalation

Ashes of the Singularity: Escalation scales exceedingly well with core count, so the Core i7-7820X trails Intel's -7900X in our results.

The Ryzen processors still can't overtake Skylake-X. However, they do dispatch Core i7-7700K with relative ease. This Kaby Lake-based chip is similar to Core i7-7740X, which Intel inexplicably shoehorned into X299 alongside our Skylake-X models. Because the -7740X makes so little sense, we're focusing our comparison on the more popular -7700K.

MORE: Best CPUs

MORE: All CPUs Content

Civilization VI & Battlefield 1

Civilization VI AI Test

Core i9-7900X suffered tremendously during our Civilization VI graphics test, but it clings onto a leading spot in the more CPU-focused AI benchmark. This is surprising considering the metric's preference for quad-core processors.

Even more surprising, Core i7-7820X falls to the bottom of our field. We retested several times with varying parameters, but there is no clear answer to explain the -7820X's lackluster performance. Overclocking helps the -7820X overtake a stock Ryzen 7 1800X, but it embarrassingly trails the tuned six-core Ryzen 5 1600X.

Civilization VI Graphics Test

Core i7-7820X joins Core i9-7900X at the bottom of this chart, while the 8-core Broadwell-E-based -6900K takes pole position. The Ryzen models are very competitive; they outperform both Skylake-X processors.

Intel suggested to us that some programs might require optimizations to accommodate its new mesh topology. Civilization VI appears to be among them.

Battlefield 1 (DX11)

The Core i7-7820X shows up ahead of Core i7-6900K on our chart, though the difference is tiny. Although Ryzen takes the bottom positions, again, we're talking about a slight loss of average frame rate compared to much more expensive CPUs. Without question, our Battlefield 1 benchmark is mostly graphics-bound.

MORE: Best CPUs

MORE: All CPUs Content

Grand Theft Auto V, Hitman & Shadow of Mordor

Grand Theft Auto V

We measure performance during Grand Theft Auto V's F-16 flight sequence with the built-in benchmark.

Core i9-7900X boasts 25% more cores than Core i7-7820X, but due to the vagaries of scaling, it only enjoys a 12% performance advantage at stock clock rates. Still, that's pretty darned good in the world of CPU benchmarking. Both models easily outpace Core i7-6900K and AMD's Ryzen processors. Meanwhile, Core i7-7700K shows why we consider it to be a strong value in gaming PCs.

Hitman (2016)

Intel's Core i7-7700K beats the stock Core i7-7820X, which is an impressive feat considering their similar Turbo Boost frequencies. A bit of tuning propels the -7820X further up the chart, but the Core i9-7900X reigns supreme at both stock and overclocked settings. Tuning also greatly benefits the Ryzen 7 1800X, but it and the overclocked Core i9-7900X also suffer the most frame time outliers.

Middle-earth: Shadow of Mordor

Once again, the Core i7-7820X bests a pricier -7900X. But the nimble Core i7-7700K outpaces both Skylake-X CPUs.

Broadwell-E serves as a baseline to underline Skylake-X's improved performance in applications able to utilize Intel's reworked architecture.

MORE: Best CPUs

MORE: All CPUs Content

Project CARS & Rise of the Tomb Raider

Project CARS

The Core i7-7700K and overclocked Ryzen 5 1600X nudge past Intel's stock Core i7-7820X. The -7820X enjoys a 15% speed-up after we overclock it. However, it's noteworthy that a Core i7-7700K averages 115.7 FPS at 5 GHz. We simply didn't have room for it in our chart. That's a much higher frequency than what you can reasonably attain on Skylake-X.

Also, the Core i7-7820X offers a marked improvement over Intel's 8-core -6900K.

Rise of the Tomb Raider

Performance scaling drops off as we ascend the hierarchy of high-end CPUs. Clearly, there's a graphics bottleneck at play.

The Ryzen CPUs are very competitive in the benchmark's initial stage, but there's a significant performance drop as we move into the test's CPU-intensive section. This penalizes the Ryzen processors with a lower average frame rate.

MORE: Best CPUs

MORE: All CPUs Content

Workstation & HPC Performance

2D Benchmarks: DirectX and GDI/GDI+

If you want to know more about our HPC benchmarks, check out the AMD Ryzen 7 1800X CPU Review. We didn't just copy results from that story, though. Rather, after a number of BIOS updates and software configuration changes, we retested everything. This gives us a more up-to-date picture, reflecting improvements of up to 15% that AMD worked hard to enable.

Intel's Core i7-7820X outstrips the -7900X in our AutoCAD 2D workload due to its frequency and IPC throughput advantage. The Core i9-7900X wins in the GDI/GDI+ benchmarks, though. Both processors provide more performance than a Broadwell-E-based Core i7-6900K.

2D Benchmarks: Adobe Creative Cloud

Per-cycle performance plays a role in these lightly-threaded applications, giving the -7820X an advantage in several tests. Both Skylake-X models suffer lower performance than we'd expect in the Photoshop Heavy and InDesign workloads. Hopefully Adobe is planning an update that'll address this anomaly.

3D Benchmarks: DirectX and OpenGL

The Core i7-7700K vigorously cuts through most of these workloads, indicating that prefer high clock rates, all else being equal.

Both Skylake-X-based chips trade places through several of the tests; the distance between them remains small, though.

CPU Performance: Workstation

Broadwell-E leads the Skylake-X-based processors in a few of these workloads, reminding us that Intel's mesh topology may lead to performance regressions in some cases.

The Ryzen 7 1800X is incredibly competitive during this round of testing.

CPU Performance: Photorealistic Rendering

Rendering benefits from brute-force parallelism, so the 10-core Core i9-7900X naturally provides the best performance.

The workload utilizes all cores fully, so it also provides a good multi-threaded comparison between the eight-core -7820X and Ryzen 7 1800X. Intel's processor takes the lead due to its per-cycle performance advantage, but Ryzen is surprisingly competitive given its lower price and value-oriented platform. It also doesn't require a custom water-cooling loop to reach its potential, whereas Skylake-X does.

CPU Performance: Encoding & Compression/Decompression

The -7820X falls into a predictable place during our threaded encoding workload. Nipping at its heels is AMD's nagging (and much less expensive) Ryzen 7 1800X.

Core i7-7820X struggles mightily with our lightly-threaded decompression workload. Its place in the chart is much lower than we'd expect, given the way Intel implements its Turbo Boost technology.

High Performance Computing (HPC)

Complex HPC applications largely benefit from the -7820X's high clock rate and beefy core count. But aside from the SRMP workload, AMD's Ryzen 7 1800X again proves to be the fly in Intel's high-priced ointment.

MORE: Best CPUs

MORE: All CPUs Content

Power Consumption & Thermals

Intel isn't using a soldered integrated heat spreader (IHS) for Skylake-X or Kaby Lake-X. Instead of the solder most enthusiasts clamor for, heat moves from the die to the IHS through thermal paste. This decision has implications for both our power consumption measurements and overclocking efforts, as you no doubt saw in our Skylake-X Mess Explored: Thermal Paste And Runaway Power article.

We measure power consumption after the voltage converters and CPU, using points on the motherboard.

Power Consumption

These numbers are generated using stock motherboard settings; any significant under-volting has no effect, except to cause stability problems.

Core i7-7820X consumes significantly less power at idle than Broadwell-E, and it also provides a small benefit compared to the ten-core Core i9-7900X.

AMD's Ryzen processors use quite a bit more power during idle, and that doesn't change much between the eight- and six-core models. Its Zeppelin die is a unified four-core building block, so it makes sense that AMD still feeds the disabled silicon with power.

AutoCAD is lightly threaded, but even mundane workloads trigger Skylake-X's aggressive Turbo Boost frequencies. Less than one watt separates the Core i7-7820X and i9-7900X. Meanwhile, the previous-gen Core i7-6900K doesn't hit the same clock rates, conveying a slightly lower power figure.

The Ryzen models are in a class of their own, though we have to point out that they (along with Core i7-6900K) underperform Skylake-X. So, our power measurements aren't indicative of overall efficiency.

The Skylake-X processors draw more power than competing CPUs during our Metro Last Light gaming loop. We do observe a ~6W reduction going from Core i9-7900X to Core i7-7820X, though. Surprisingly, the -6900K is the most frugal Intel processor in the group.

Kicking off a Prime95 run unleashes the Core i9-7900X's hideous power consumption. Core i7-7820X is a little more civil, likely due to its disabled FMA units. But it still draws a lot more power than the Broadwell-E-based -6900K or Ryzen 7s.

Thermal Performance

Thermal output was recorded during AIDA and Prime95 stress tests at stock frequencies. We used the Corsair H100i v2 cooler, which provided enough headroom to keep up with these CPUs.

As expected, we observe a slightly lower average temperature from Core i7-7820X compared to Core i9-7900X.

Speaking of, the Core i9-7900X encountered difficulties during the Prime95 AVX workload. We use AIDA's system stability test window to monitor temperatures during our test run, and it detected thermal throttling. Typically, AIDA provides a percentage to quantify the severity, though we don't know what caused the slow-down.

The system generates bit-coded throttling flags based on three variables: Tcore, Tpackage, and VRM temperatures. AIDA detects these flags, but doesn't specify which one the warning is based on. We suspect the VRMs are to blame in this case.

MORE: Best CPUs

MORE: All CPUs Content

Final Analysis

Intel's designed its X-series line-up to hit (almost) every budget. Prices in the high-end desktop family range from $242 to $2000. Unfortunately, expensive X299-based motherboards and the need for liquid cooling make it difficult to find value in an Intel HEDT package, particularly compared to AMD's Ryzen 7 and the Skylake-S models.

Adopting a mesh topology for Skylake-X was necessary to pave the way for greater scalability in the future. But the move causes some complications. Core i7-7820X offers great frame rates through our gaming suite. However, we did encounter those same performance peculiarities observed during our Core i9-7900X review. We're told that software optimizations could ameliorate this in the future, and hopefully Intel delivers as quickly as AMD did after launching Ryzen.

Regardless, if gaming is your primary use case, we still recommend Core i7-7700K as the best option, even in high-end PCs. Intel's Z270 chipset is more affordable and more mature than the still-evolving X299 platform. It also leaves you more room to budget for a flagship-class graphics card.

In the past, Intel's HEDT portfolio offered the strongest performance in threaded workloads thanks to lots of cores. It couldn't keep up in single-threaded benchmarks, though, due to lower clock rates. Skylake-X changes this with higher base and Turbo Boost frequencies. The addition of Turbo Boost Max Technology 3.0 even introduces a 4.5 GHz ceiling that matches the Core i7-7700K. Of course, when you need greater parallelism, the clocks spin down to allow all eight cores to do their work in game streaming, rendering, and semi-professional apps.

Ryzen 7 1800X typically can't beat the -7820X in those scenarios. But it's certainly competitive. AMD's CPU is also $100 cheaper. On top of that, X370-based motherboards cost a lot less than Intel's X299 platform, and liquid-cooling isn't a requirement in Ryzen's case. Without question, the 1800X on an X370 motherboard presents a compelling alternative to Intel's overpriced combo.

Overclocking is a mixed bag on Skylake-X. The Core i7-7820X appears to enjoy quite a bit of frequency headroom. Thermal throttling jumps up to bite you long before the CPU's true potential is realized, though. Although enthusiasts would like to see solder between Intel's die and heat spreader, at least thermal paste isn't as big of a problem for overclockers on the lower-power Core i7-7700K. It's a much bigger issue with the fire-breathing Skylake-X design.

We don't know what to say about vROC, other than charging a premium for hardware keys to unlock RAID functionality won't go over well on the desktop. Intel remains silent on vROC pricing and availability, so for now, the storage redundancy feature lurks inside, waiting to be enabled.

Intel should probably feel lucky that Core i7-7820X won't be going up against AMD's Threadripper, since the cheapest model will sell for $800. As it stands, this $600 CPU has a hard time justifying its premium over Ryzen 7 1800X, which currently sells for as little as $420. Moving forward, Intel may have to get even more aggressive about winning over enthusiasts. Don't lop off PCIe connectivity. Leave the FMA units intact. Don't nickel-and-dime for storage functionality. Use solder under the IHS. We like the big dies with lots of cores able to clock up when most of them aren't active. But the competition is heating up and Intel is looking a little flat-footed.

MORE: Best CPUs

MORE: All CPUs Content

Extreme Overclocking: 10 Ryzen CPUs Under LN2

Jean Michel "Wizerty" Tisserand — Fri, 14 Jul 2017 13:00:00 +0000

Introduction

AMD's Ryzen processors offer a compelling price/performance ratio right out of the box. But despite their many overclocking-friendly knobs and dials, most enthusiasts struggle to take the CPUs beyond 4 GHz. Given that we know the ins and outs of extreme overclocking, though, we have a solution. It's time to break out the liquid nitrogen!

Allow us to take you on a cryogenic journey, where we'll explore Ryzen's behavior when it's cooled to -196°C. Our experiment will allow us to correlate frequency scaling to temperature, voltage, and core count. We also have some tips on hardcore modding, such as lapping (sanding smooth) the processor.

Note that we previously published an article on overclocking Ryzen using air and water cooling. Check out How To Overclock AMD Ryzen CPUs for more.

If, after reading through everything, you have questions about your own overclocking endeavors, don't hesitate to ask them in our comments section. We'll be keeping an eye out in order to help however possible.

Test Configuration

We don't want to limit our quest to just one sample, or even one model. Thanks to AMD's generosity, we were able to get our hands on every SKU in the Ryzen family. Here's what we have access to for today's overclocking adventure:

2x Ryzen 7 1800X
1x Ryzen 7 1700X
3x Ryzen 7 1700
1x Ryzen 5 1600X
1x Ryzen 5 1600
1x Ryzen 5 1500X
1x Ryzen 5 1400

The processors used for this test are different than the chips used for our previous article on overclocking with air and water cooling. Therefore, we expect to see different results.

In the interest of truly torturing these CPUs, we surrounded ourselves with some of the best hardware available for testing:

The motherboard we're using is Asus' Crosshair VI Hero, equipped to facilitate extreme overclocking. The only feature it's missing is a second BIOS, which could have come in useful for recovering from a corrupted configuration.

This ROG-series motherboard is armed with two sticks of G.Skill Flare X DDR4 memory. These modules were specially developed for Ryzen. Furthermore, they are equipped with Samsung B-die ICs, known for their overclocking headroom.

Last of all, we use a Cooler Master MasterWatt Maker 1200 power supply.

AMD Ryzen 7 1700XView Deal

AMD Ryzen 7 1700View Deal

AMD Ryzen 5 1600XView Deal

AMD Ryzen 5 1600View Deal

AMD Ryzen 5 1500XView Deal

AMD Ryzen 5 1400View Deal

G.Skill Flare X (2x 8GB)View Deal

Cooler Master - MasterWatt Maker 1200View Deal

MORE: Best CPUs

MORE: De-Lidding and Overclocking Core i7-7700K

MORE: CPU Overclocking Guide: How (and Why) to Tweak Your Processor

Preparing The Motherboard

Measurement Points

We begin with a little modification to the motherboard. Test points are available on the PCB, but Asus sadly neglects to include connectors. As a result, it's on you to precisely touch the indicated pads with a multimeter in order to take measurements.

Normally that's not an issue. But when it comes to extreme overclocking, voltages have to be watched more closely and we don't always have our hands free. Right away, we add something to plug our volt meter into for easier measurements.

LN2 mode

Some preparations are easier than others. In this step, we simply move the jumper labeled “LN2 MODE” in order to activate it. Once enabled, the motherboard automatically starts up with higher supply voltages. This will offer better support for the extreme conditions we're planning to apply. Certain hidden profiles in the BIOS are also unlocked.

Don't enable the LN2 MODE jumper if you plan on overclocking at ambient temperatures. Following the activation of this mode, the PLL (Phase Locked Loop) will transition from 1.8V to 2.1V. If you read our previous article, we revealed that raising PLL from 1.8 to 1.9V caused an 8°C temperature increase. Raising the PLL to 2.1V on air cooling should be avoided!

Isolation

When overclocking with liquid nitrogen, special care must be taken to protect your hardware. Ice will form and you don't want to risk water droplets falling onto the electrical components. We presented a number of ways to approach this in De-Lidding and Overclocking Core i7-7700K with Water and LN2.

To start, we remove the heat sink covering the motherboard's VRMs. This step isn't obligatory; it most depends on the overclocker's preference. In our opinion, though, this sink is useful for extreme overclocking, so we prefer to take it off. This facilitates the insulating steps that follow, and even more important gets rid of a big metal mass. That could be a big risk for forming condensation otherwise.

Once we're down to the bare motherboard, we craft a “shield” made from shop towels. This protective layer must be fitted as tight as possible in order to prevent condensation from reaching the PCB.

Unused RAM slots are filled with towels. The space between surface-mounted components gets the same treatment, too. To finish, a section of neoprene is fitted around the socket. This protection serves as a last defense, though we hope that condensation is stopped well before this stage.

If it is properly made and placed, this ultimate protection is sufficient to shield the hardware without damaging it, and can be easily removed. Once the cooling pot (which holds the liquid nitrogen and cools the CPU) is in place, we pack on more shop towels. The pot is wrapped in neoprene, then towels, and any exposed portions of the motherboard are covered with several more layers of towels.

MORE: Best CPUs

MORE: De-Lidding and Overclocking Core i7-7700K

MORE: CPU Overclocking Guide: How (and Why) to Tweak Your Processor

Preparing The RAM

During our first extreme overclocking sessions, we didn't put much effort into preparing the memory. We simply installed the modules into slots with their factory heat sinks in place. This strategy didn't work all that well, as we had to stop our tests multiple times when the hardware refused to start up. Although we can't say for sure, it's likely that humidity around the memory stick closest to the pot was responsible for these failures.

After drying the hardware, everything worked once again. And since this happened more than once, removing the RAM sinks just seemed like a prudent idea.

If you choose to follow suit, be careful: some heat spreaders are so firmly attached to the memory chips that they can pull ICs right off the PCB. To improve our chance of success, we warmed the module up first. Even then, it took a lot of effort to achieve our goal.

Before masking off the chips, we take advantage of their exposure to verify that our sticks are endowed with Samsung B-die memory. Without question, they're the best for overclocking.

The type of the IC is indicated on the chip, just above the red line we drew, “5WB”. If these were E-die chips, we'd see 5WE, or 5WD for D-die. Other manufacturers use different identification marks.

To protect our memory stick, we cover it with adhesive tape. The impermeable nature of this tape should keep condensation from ruining our day. Don't worry, the missing sinks pose no challenge to stability. Even at 1.6V, the chips remain cool.

The small pins at the bottom of the module are too close to the slot to be covered. They're buried under layers of absorbent towels though, and therefore less exposed.

Even without the sinks in place to trap condensation, the surface of the RAM doesn't seem to be much less humid. This picture was taken at the end of an overclocking session, and you can clearly see drops of water on the stick's most exposed side. Nevertheless, our efforts pay off: we didn't have any issues with system cut-outs.

It would have been easier to pile paper towels around the RAM to keep water from collecting on the PCB. But then the modules would have cooled down even more through the motherboard. Given that Samsung's B-die memory dislikes freezing temperatures, they could have then failed to function at high frequencies. For better or worse, nothing is ever simple with extreme overclocking.

MORE: Best CPUs

MORE: De-Lidding and Overclocking Core i7-7700K

MORE: CPU Overclocking Guide: How (and Why) to Tweak Your Processor

1800X: First Test Of Scaling With LN2

By luck, we chose the best Ryzen 7 1800X for our scaling trials right out of the gate.

Ryzen 7 1800X : Frequency As A Function Of Temperature At 1.5V

This first experiment explores how the processor behaves at different temperatures with its core voltage fixed at 1.5V. That seems really high for a test at ambient, and we don't want to damage our CPU just for the sake of generating a chart. On the other hand, once the 1800X is at -196°C, a voltage of 1.5V is actually pretty conservative. In the end, we picked this value as the best compromise between risk at ambient and extreme overclocking performance.

At room temperature (20°C), the processor passes a Cinebench R15 run at 4175 MHz. This is a high frequency for Ryzen, achieved at significant risk. Don't try this at home: warm silicon doesn't like aggressive voltage settings.

By lowering the temperature to 0°C, we're able to dial in a 100 MHz frequency increase. So far, we have an improvement of around 5 MHz/°C.

We continue to lower the temperature by pouring liquid nitrogen in the cooling pot until we arrive at -50°C. The frequency gain is now 250 MHz. Our progression remains constant with the same rate of 5 MHz/°C.

Next we see -100°C, giving us an additional 200 MHz. The trend begins to flatten, indicating that the scaling progression is slowing down slightly (4 MHz/°C).

An additional 50°C drop in temperature shows a gain of only 175 MHz at -150°C. The increase is 3.5 MHz/°C.

For the last step, we reach full pot. At -196°C, with a 46°C drop in temperature, the clock rate stabilizes at 5025 MHz (2.7 MHz/°C).

A full pot signifies that our chamber is filled to the brim. We are at the minimum temperature permissible with liquid nitrogen, which is -196°C. To go any lower, you'd need liquid helium: -269 °C.

Thanks uniquely to the reduced temperature, our sample passes Cinebench R15 with an additional 850 MHz overclock. When you hear that these processors love the cold and are damaged by heat, here is the proof.

Ryzen 7 1800X : Frequency As A Function Of Core Voltage At -196 °C

The next experiment tracks our CPU's behavior at various core voltages with a temperature held constant at -196°C. Only the voltage changes; all other parameters remain unchanged.

At 1.5V, we hit the same clock rate seen in the previous set of tests. This makes sense, of course. However, we'll take the time to mention how well our sample scales. In fact, some of the CPUs we tested couldn't hit 4175 MHz even at a core voltage of more than 1.8V.

With an additional 0.1V, the frequency increases 100 MHz. This is significant, but not exceptional. As a reminder, we saw the same gain under air cooling when transitioning from 1.3 to 1.4V, while the shift from 1.0 to 1.1V offered a superior increase of 250 MHz. Before starting these tests, we would have guessed that the clock rate gained by increasing voltage would be amplified at lower temperatures. That's not the case, though.

The same observation applies when we raise the core voltage an additional 0.1V to 1.7V (+100 MHz).

For the next step, we stabilize 75 MHz higher at 1.8V. This frequency is remarkable: 5300 MHz. Such a clock rate is not common with Ryzen.

We halt the trial at 1.85V. Going any higher yields no frequency increase, and the voltage settings start becoming hazardous to our guinea pig.

The progression we just saw cannot be extrapolated to all Ryzen CPUs at the temperatures and voltages we used for testing. Certain specimens will fare worse when cold, some won't accept more than 1.75V, and others will continue scaling beyond 1.9V. This sample is above average though, even if it's always possible to find something better.

MORE: Best CPUs

MORE: De-Lidding and Overclocking Core i7-7700K

MORE: CPU Overclocking Guide: How (and Why) to Tweak Your Processor

Lapping The CPU

As we ran our tests, we realized that this sample was above average. So, we tried to land a record for eight-core processors, including AMD and Intel CPUs. The combat was relentless. On one hand, there were some very good results obtained by overclockers right around the time Ryzen was released, and it isn't hard to imagine that they had access to a sizeable quantity of hand-picked chips. On the other hand, Intel's Core i7-5960X compensates for its age with frequencies beyond 6 GHz under liquid nitrogen cooling.

We therefore decided to concentrate our efforts on two benchmarks: Cinebench R15 and GPUPI. In both cases, we succeeded in taking second place, in front of the -5960X contenders running around 6200 MHz (in the case of GPUPI).

Lapping In Pursuit Of MHz

At that time, our highest clock speed in GPUPI was 5390 MHz. The leader, Der8auer, was at 5440 MHz. First place, while so close, seemed out of reach. Without a better 1800X at our disposal, we decided to lap our sample in the hopes of better thermal transfer.

During our tests, we saw a gain of 2.7 MHz/°C at -196°C. If lapping helped us gain 15°C, which is not impossible given the high voltages we were using, 5430 MHz should be attainable.

The process proved more laborious than we expected. Within the first few minutes, defects began appearing in the CPU lid's shape. A flat processor should be “worn” completely on the surface in a homogeneous manner. However, we were uniquely attacking the edges of the IHS.

There were two possible scenarios: either the processor was not actually flat, or we were lapping incorrectly. To remove any doubt, we took a brand new razor blade and placed it on the CPU's surface. There it was: the blade only touched the borders, allowing a seam of light to shine through in the center.

Our lapping effort resumed. We used emery cloth (comparable to sand paper and intended for use in sanding metals) attached to a piece of glass to guarantee a uniform surface. We started with a course grit to rapidly remove (relatively, of course) the extra material on the edges.

This is the progress we made in one hour. The nickel coating is removed, revealing copper on the sides. Gradually as we eroded the surplus material, our uniform area grew in size. In the end, almost the entire IHS appears to be copper. Two visible spots persist, but the defect is sufficiently small to be ignored. The processor doesn't need to be polished any further. Having a flat surface is top priority. Fine scratches won't affect performance.

At this point, we resumed our trials with liquid nitrogen, confident in our work and hopeful that we'd realize our estimated gains.

MORE: Best CPUs

MORE: De-Lidding and Overclocking Core i7-7700K

MORE: CPU Overclocking Guide: How (and Why) to Tweak Your Processor

BIOS Settings

New BIOS features allowing the adjustment of RAM timings, which were previously greyed-out, became available during our testing. Unfortunately, those updates started rolling out too late. Given that there is only one BIOS chip on our test platform, and that we had already generated quite a bit of data beforehand, we decided to continue with BIOS version 0083 in order to maintain our methodology.

While some overclockers saw improvements from BIOS version 1201, this wasn't the case for everyone. Memory controller quality played a big role, too.

We already explained the BIOS settings in our previous guide, so we won't dive back into the details. Here's what the voltages we used look like, though:

CPU Core Voltage: Set to 1.5V on our sample. It can be raised to 1.8 or even 1.9V without worry when the cooling pot is at -196°C. During our first tests, we started at 1.5V in the BIOS and then stepped higher through the operating system. After a while, we determined it was safe enough to simply start off with the desired value.
VDDSOC: As with air cooling, avoid setting this any higher than 1.25V. For example, you can set VDDSOC to 1.2V, find the maximum stable frequency for your RAM, and then try to lower the setting to 1.18 or 1.15V. In short, seek the minimum value necessary.
DRAM: For our memory sticks, 1.6V was sufficient for 3200 MT/s at 12-12-12 timings. The most important variable is the quality of processor's memory controller, but aside from switching processors, there is no silver bullet!
1.8V PLL: We didn't see any gains when increasing the PLL voltage. The LN2 mode jumper adjusts this to 2.1V, but you can leave it at 1.8V without risking any problems.
1.05V SB: 1.3V does not seem to pose a problem; however, we didn't realize any performance gains by increasing this parameter.

The Case For LLC

In the “External Digi+ Power Control” sub-menu, you find the CPU Load-line Calibration option. We tried multiple modes available on Asus' Crosshair VI Hero and recorded their voltages with a multimeter. With a voltage setting of 1.8V in our BIOS, we observed the following values under load:

LLC 2: 1.78V
LLC 3: 1.83V
LLC 4: 1.85V

During our test under air cooling, we were surprised to see that even the lowest level of LLC was already too high. With LN2 and higher voltages in play, Level 1 and 2 are no longer sufficient, though. But also be careful not to overdo it, since Level 4 and 5 were too severe. Therefore, we recommend LLC 3 for voltages near 1.8V.

MORE: Best CPUs

MORE: De-Lidding and Overclocking Core i7-7700K

MORE: CPU Overclocking Guide: How (and Why) to Tweak Your Processor

AM4_PCRATIO: A Frequency For Each Core!

Hiding behind this obscure name is a small utility allowing one core to be overclocked, or to set one CCX (CPU Complex) to a different frequency than the other one. If you don't know why you'd bother, allow us to explain in a little more detail.

When overclocking a processor, all of the cores operate at the same clock rate, and this is great for most folks. On the other hand, if you're gunning for a single-threaded benchmark record, performance is held back by the least-scalable silicon.

Take a quad-core processor as an example. Some of its cores are capable of running at 4 GHz, while others hit 4.2 or 4.4 GHz. Without utilizing special tools, you'll lock up as soon as you try pushing past 4 GHz since some of the cores aren't capable of going any higher. Talk about a frustrating situation: the best cores are held back by the worst.

The CCX Version

Using am4_pcratio_ccx (instead of am4_pcratio_focus) allows the frequency of all cores in one CCX to be changed, while the other CPU Complex's cores are automatically adjusted down to a lower clock rate.

Here is an example of three processors: A, B, and C. The value shown in blue indicates the maximum frequency that the cores are capable of operating at.

A: In this example, whether you deactivate cores from the BIOS or from AM4_PCRATIO, you won't see much of a gain because they're all identical.
B: Core 0 is the best one. You can can choose to uniquely activate it from the BIOS, while the others are turned off. With fewer active cores, the processor generates less heat and you can overclock even further. Another method would be to use AM4_PCRATIO with a focus on Core 0. The result would be roughly identical, but you'd end up with other active cores. For instance, Windows would run on some cores at 2 GHz, allowing the benchmark to have unique control of the fastest core.
C: Now the fastest core is Core 3. It isn't possible to reserve only this one using the BIOS. So, using AM4_PCRATIO is your best chance for an improved benchmark score.

Given that each processor is different, it is difficult to quantify the gain from this manipulation. Worst-case, if you spend time trying each core and discover their limits are all the same, you wasted a couple of hours. On the other hand, if your sample is more like our second or third example, the gains can be substantial. Knowing that a competition can be won with a 10 or 20 MHz advantage, this is far from anecdotal.

Put your hard work in at room temperature. That'll allow you to conserve several liters of LN2.

With a bit of experience, it's easy to figure out the limits of individual cores quickly. All you need to do is overclock each one, raising their frequencies individually until your PC crashes. Since we have to run our benchmark multiple times, choose a 10- to 15-second test for brevity.

Our strategy was as follows:

Start by finding the limits of Core 0, then focus on Core 0 by using AM4_PCRATIO and execute GPUPI.
If the benchmark completes, note the stable frequency and proceed to a higher clock rate. Repeat this operation until the system freezes. We now know the maximum frequency for Core 0. If the test completes at 4000 MHz, 4025 MHz, and 4050 MHz, but fails at 4075 MHz, we conclude that 4050 MHz was the limit.
Restart the PC and change focus to Core 1. If 4000 MHz is OK, but 4025 MHz causes a crash, Core 1 isn't as good, so we don't use it.
Repeat this process across the CPU to determine its strongest core.

MORE: Best CPUs

MORE: De-Lidding and Overclocking Core i7-7700K

MORE: CPU Overclocking Guide: How (and Why) to Tweak Your Processor

OC: Ryzen 7 1800X

Since we have multiple Ryzen 7 1800X CPUs, each one gets its own number. Our freshly lapped processor is first.

Ryzen 7 1800X #1: 5411 MHz

The temperature drops, frequencies increase, but our optimism does not last for long. This CPU wouldn't stabilize at 5420 MHz, and it's just barely able to run at 5400 MHz. With a little work, and by tweaking the settings, we manage to pass GPUPI at 5411 MHz, reducing the three-second delta to just 0.9s.

We score a beautiful 2430 points in CineBench R15 at 5323 MHz, again falling short of first place. Our CPU is a good one, but it isn't setting any records.

Before moving on to other samples, we make note that this chip's IMC is quite average. Under LN2 cooling, it's impossible to push our memory beyond 3000 MT/s. Unable to achieve higher data rates, we instead tighten timings to 11-11-11-26 and use a REF_CLOCK setting of 139 MHz to boost our score a bit.

Ryzen 7 1800X #2: 5200 MHz, DDR4 At 3310 MT/s

Our second sample isn't quite as good as the first one. As a result, we didn't spend a lot of time trying to push its performance. After assembling and waterproofing our platform, we spent another two hours toying with the chip's limits. In the end, Cinebench ran successfully at a little higher than 5200 MHz.

Incidentally, this CPU's memory controller is quite good. It allowed us to complete SuperPI 32M with a memory frequency of 1655 MHz (3310 MT/s).

MORE: Best CPUs

MORE: De-Lidding and Overclocking Core i7-7700K

MORE: CPU Overclocking Guide: How (and Why) to Tweak Your Processor

OC: Ryzen 7 1700X & 1700

Ryzen 7 1700X: 5310 MHz

Our overclocking approach didn't change for AMD's Ryzen 7 1700X. The cores in this specimen aren't that great, but they aren't bad either.

Just for giggles, we again tried for some records and ended up taking first place in all of the benchmarks we ran except for SuperPi 32M, which requires a lot of optimization. Sadly, we didn't have enough time to spend hours tuning for each metric. There are a lot more CPUs to get through!

What we have, then, is an average CPU that's easy to benchmark and fairly comparable to our 1800X samples.

Curiously, the scores we obtained in Wprime 32M are systematically very poor on all of our eight-core processors, and we can't explain why. Some overclockers see times between four and five seconds, while others dip under three seconds. While differences from one system to another are normal, the deltas shouldn't be 100%.

Ryzen 7 1700: 5150 MHz And Major Bugs

While we were testing our first Ryzen 7 1700, our friend Niuulh was testing a second sample we lent him for a competition. If you are familiar with the term poisoned chalice, this inadvertently became one in every sense.

We started by overclocking with air cooling, and quickly hit a brick wall. In the first minutes of our effort, the processor was stuck at 1550 MHz. Obviously something was wrong, since the chip's stock clock rate is higher than that.

And yet, the BIOS settings were unchanged from our trials with the 1800X and 1700X. The values indicated in Asus' software even matched those we set in the BIOS. Still, the processor remained fixed at 1550 MHz.

Moving to MSI's X370 Xpower Gaming Titanium didn't solve our problem. Two CPUs, separated by 600km, were suffering the same symptoms. The motherboard didn't seem to be the cause, the operating systems were unique installations, and the overclockers were different. Without a doubt, this was a problem with our processors.

After spending many hours tweaking BIOS parameters, the only solution was to not modify the supplied voltage via the BIOS, modify the OFFSET mode, or raise the REF_CLOCK setting. Whatever was going wrong, this was not business as usual.

The problem became even more severe under LN2 cooling. Once the temperature dropped below -20°C, the processor got stuck at 1550 MHz and nothing would free it.

Since our two processors were affected identically, we came to the conclusion that Ryzen 7 1700 cannot be used with LN2 cooling. We were condemned to powering on at -20°C and dropping the temperature to -196°C after booting to Windows. When the system crashes (every four to five minutes), you have to raise the cooling pot's temperature back to -20°C with a gas heating torch and start over. This eats up a ton of time, nitrogen, and gas, and it sucks the fun right out of overclocking.

Seeing that our Roman and Indonesian friends were able to overclock Ryzen 7 1700 with LN2, we asked them how they did it. Their answer: they didn't do anything special at all.

We decided to test our last 1700, and to our surprise, it had no problems under air or LN2. There went our hypothesis. By comparing processor batches, we determined that the two problematic processors were fabricated on the same date, while the third chip was older. Could this be a manufacturing issue? For now, we don't know.

As far as maximum clock rate goes, our three samples completed Cinebench R15 at anywhere from 5050 and 5150 MHz.

MORE: Best CPUs

MORE: De-Lidding and Overclocking Core i7-7700K

MORE: CPU Overclocking Guide: How (and Why) to Tweak Your Processor

OC: Ryzen 5 1600X & 1600

We have two processors endowed with six cores and two others with four. Which one will overclock the best? Now's the time to place your bets.

Ryzen 5 1600X: 5250 MHz

During our air cooling tests, the 1600X wouldn't cooperate. Bugs kept popping up and holding back the frequency. We assumed things would get worse once we introduced LN2. In the end, while our sample didn't have any issues with liquid nitrogen cooling specifically, it still proved difficult to get running.

At this point, it's pretty safe to conclude that you never know what you're going to get. The behavior of these processors is much too variable from one to the next, even among identical models.

After spending two hours with LN2, we only saw Windows twice. Two times in as many hours is a poor showing, to say the least, which explains the absence of a score.

We hit 5.1 GHz at 1.7V, and even saw 5250 MHz once at 1.85V. The cores are pretty good, but far from Der8auer's diamond in the rough that exceeds 5.4 GHz. Still, the clock rate we recorded was good enough to land second-highest for a Ryzen 5.

For reasons pertaining to our mental health, we didn't bother trying to squeeze any more performance from the processor's memory controller.

At the end of this chip's trials, as we were removing the cooling pot, our Ryzen 5 1600X got stuck to the bottom and was pulled from its socket. This isn't the first time we've seen that. Either the socket doesn't grip tightly enough, or the thermal paste has too much suction. Fortunately, nothing was damaged.

R5 1600: 5075 MHz

Unlike the six-core chips we just finished testing, our 1600 didn't suffer any show-stopping bugs. Whether it be air or LN2 cooling, its behavior is normal. This processor loves high core voltages and exhibits progression up to 1.92V. With that said, it's not a great performer, plateauing at 5075 MHz. The IMC is nothing exceptional; we timidly reached 3000 MT/s at CAS 12, but were stable around 2800 MT/s for our memory-intensive benchmarks.

This doesn't mean that all 1600s will behave this way. We'd guess that fewer active cores would enable higher maximum clock rates. But that doesn't prove to be the case today.

MORE: Best CPUs

MORE: De-Lidding and Overclocking Core i7-7700K

MORE: CPU Overclocking Guide: How (and Why) to Tweak Your Processor

OC: Ryzen 5 1500X & 1400

Ryzen 5 1500X: 5030 MHz

We won't drag things out on you. As we dropped down AMD's Ryzen stack, the maximum frequencies continued to fall. Despite starting off great with a Ryzen 7 1800X that exceeded 5.3 GHz, our 1500X barely crossed the 5 GHz mark. It managed to complete Cinebench successfully at 5030 MHz and a voltage of 1.85V.

When we ran our benchmarks, there were no results on HWBOT using LN2 cooling. So we took advantage of the opportunity to fill the ranks with 13 first-place finishes. To be honest, though, there is no glory in breaking a record when the best competition is only using water cooling. At least future extreme overclockers have a challenge to beat now.

Ryzen 5 1400: 5000 MHz

The last of our 10 processors is the worst of all. It simply doesn't like high voltages, and it stops progressing at 1.8V. Under these conditions, 5 GHz was the limit for Cinebench R15.

This is a disappointment for the enthusiasts who hoped lower core-count Ryzens might be capable of higher clock rates. But it's a logical outcome, given the way AMD bins its dies. The best performers naturally turn into 1800Xes, with all of their cores activated and operating at the highest frequencies.

At least this chip's IMC isn't terrible. It reached 3100 MT/s at CAS 12. The processor didn't have any issues with bugs, and it accepted -196°C without a problem. The test would have been downright enjoyable if we could have squeezed out an extra 300 MHz.

As with the 1500X, we took advantage of the absence of LN2-based scores to place our own.

MORE: Best CPUs

MORE: De-Lidding and Overclocking Core i7-7700K

MORE: CPU Overclocking Guide: How (and Why) to Tweak Your Processor

Overclocking: Air Vs. LN2

Before wrapping up, we wanted to see if there was a strong correlation between a processor's maximum frequency with liquid nitrogen cooling and what it is capable of with air cooling. This has significant appeal: why bother spending precious time and an enormous quantity of LN2? It could be sufficient to pre-select the best processors based on their performance under air cooling, and then test only the best contenders using liquid nitrogen cooling.

So we revisited our 10 Ryzen CPUs and found the maximum frequency of each at 1.35V. The table below allows us to make a comparison between the room temperature results and those with liquid nitrogen.

Sample	Date of Fabrication	Max Freq. @ 1.35V (MHz)	Max Freq. @ -196°C (MHz)	Ideal Voltage @-196°C (V)
Ryzen 7 1800X #1	1703PGT	4050	5320	1.85
Ryzen 7 1800X #2	1711SUT	4025	5220	1.85
Ryzen 7 1700X	1711PGS	3975	5170	1.85
Ryzen 7 1700 #1	1713PGT	3900	5050	1.85
Ryzen 7 1700 #2	1709PGT	3900	5120	1.85
Ryzen 7 1700 #3	1713PGT	3925	5150	1.94
Ryzen 5 1600X	1711SUT	3975	5250	1.85
Ryzen 5 1600	1713SUT	3900	5075	1.92
Ryzen 5 1500X	1712SUT	3850	5030	1.85
Ryzen 5 1400	1714SUT	3900	5000	1.8

At the top of the table, the 1800X CPUs dominate with air, just as they did in our LN2 tests. It seems like a processor capable of passing Cinebench R15 at 4 GHz with air cooling could be able to do 5.2+ GHz under the influence of LN2. The difference between our first and second samples is nevertheless more pronounced with liquid nitrogen than cooling at room temperatures.
The 1700X tends to reinforce our observation; it finishes just below the 4 GHz mark with air, and just below 5.2 GHz with LN2.
The same observation applies to our 1700s. The best with air cooling is also the best with LN2, but an important variable also comes into play: ease of use. Our notes show that these processors gave us a hard time. The samples that struggled under LN2 cooling were problematic under air, too.
Looking back at the Ryzen 5 CPUs, we could say that a processor able to hold 3.9 GHz at ambient should be capable of benching at around 5.1 GHz. Obviously, these numbers aren't always exact, but they do illustrate a trend. This pattern seems to prove correct with the exception of Ryzen 5 1400, which should have been able to reach a higher frequency with LN2. It may have, too, had it not stopped progressing before 1.8V.

MORE: Best CPUs

MORE: De-Lidding and Overclocking Core i7-7700K

MORE: CPU Overclocking Guide: How (and Why) to Tweak Your Processor

Conclusion

While life isn't always black and white, you could say that overclocking Ryzen CPUs using LN2 is definitely cut-and-dried. On one hand, you have some processors that handle cryogenic temperatures and high voltages without a problem. They can reboot without issue at -196°C, allowing you to continue your pursuit of higher clock rates. In other words, they're every overclocker's dream.

But some Ryzen processors are stubborn to the point of being nightmarish. Having to go from -196°C to -20°C after every crash is not pleasurable. The resulting consumption of nitrogen, gas, and patience is downright frustrating.

AMD's Selection Methodology

Enthusiasts in the habit of trying multiple CPUs and keeping only the best won't be bothered. They'll toss stubborn samples aside for someone else to worry about. But if you're on a budget and only have one processor to experiment with, the luck of the draw is particularly unforgiving right now.

Strangely enough, while we expected the processors with the fewest cores to overclock best, the opposite proved true. With only 10 samples on-hand, it's hard to draw statistically reliable conclusions. According to our tests, though, the Ryzen 5 CPUs couldn't stand up to the clock rates achieved by our Ryzen 7s. This is most certainly due to AMD's binning process. Lower-performing chips have some cores deactivated and their clock rates reduced. The best dies go into the Ryzen 7 1800X model.

Thus, your best bet for a good overclock under LN2 is AMD's Ryzen 7 1800X. In our air cooling tests, one Ryzen 7 1700 was able to keep up with the 1800Xes.

Ultra-Solid Processors

To end on a positive note, we want to emphasize that, despite tens of hours testing with liquid nitrogen cooling, using high voltages at cryogenic temperatures, and building/tearing down our test platform (with water everywhere), not a single component was damaged. Ryzen appears solid, despite its relative youth. And when you stumble across a gem of a sample, it's a pleasure to overclock.

MORE: Best CPUs

MORE: De-Lidding and Overclocking Core i7-7700K

MORE: CPU Overclocking Guide: How (and Why) to Tweak Your Processor

Intel Core i7-7740X Kaby Lake-X Review

palcorn@outlook.com (Paul Alcorn) — Wed, 12 Jul 2017 13:00:00 +0000

Kaby Lake Paddles Into HEDT

Last month, Intel's Skylake-X processors debuted at lower price points than we might have expected, given the company's previous high-end desktop CPU launches. The new chips also feature hefty core counts, higher Turbo Boost frequencies, a revamped implementation of Turbo Boost Max 3.0, a realigned cache hierarchy, and a new mesh topology. If you haven't read our review yet, check out Intel Core i9-7900X Review: Meet Skylake-X.

Interestingly, the seven new Skylake-X models are accompanied by two quad-core Kaby Lake-X SKUs. In the past, Intel's entire HEDT portfolio lagged at least a generation behind the mainstream offerings. But Kaby Lake-X puts Intel's most modern architecture in the LGA 2066 interface.

It's only unfortunate that this attempt at rejuvenating the high-end desktop leads to a confusing mix of features on the X299 platform, many of which Kaby Lake-X simply does not support. Fewer PCIe lanes and a dual-channel memory controller are an obvious mismatch with expensive LGA 2066-equipped motherboards, and that's bound to create a dubious value proposition.

Kaby Lake-X Becomes Core i7-7740X

There are slight differences between the familiar Core i7-7700K (Kaby Lake-S), which drops into LGA 1151, and the new Core i7-7740X (Kaby Lake-X) designed to complement the X299 chipset.

Intel increases the -7740X's base frequency by 100 MHz to 4.3 GHz, but leaves the 4.5 GHz Turbo Boost clock rate and 8MB of last-level cache unchanged. The chip's TDP does jump from 91W to 112W, but that's not from any gain in host processing resources: we're still looking at four IA cores and the GT2 configuration of Intel's graphics engine with 24 EUs.

Although the HD Graphics 630 hardware is physically present, Intel fuses it off. This eliminates many of the multimedia features that benefit the mainstream market. Perhaps they won't be as missed in the enthusiast space, though. You do lose Quick Sync, which is useful for accelerating streaming, along with HEVC encode/decode and PlayReady 3.0 "enhanced content protection" support. On the other hand, X299 enables a more robust power delivery system, so cutting out HD Graphics could prove beneficial to our overclocking efforts. After all, the disabled hardware serves as dark silicon, absorbing waste heat from the active on-die components.

Purportedly, a larger heat spreader and package add to Core i7-7740X's capacity for dissipating thermal energy. Since CPUs vary in their ability to sustain Turbo Boost clock rates depending on their thermal headroom, the -7740X might also enjoy a larger window of higher frequencies, improving performance in our benchmarks compared to Core i7-7700K. It's too bad that Kaby Lake-X uses the same thermal interface material as the mainstream models, instead of indium solder.

The -7740X shares a 16-lane PCIe 3.0 controller with Intel's Core i7-7700K, and we'll dive into that liability shortly. DDR4 memory support is officially extended to 2666 MT/s, at least. That's a small improvement over the S-series’ sanctioned 2400 MT/s ceiling. The speed-up isn't related to any architectural enhancement. Rather, Intel tells us it adjusted the dual-channel controller's spec based on “significantly more test and manufacturing data” that suggested an increase was safe.

Surprisingly, the Core i7-7740X bears the same price as Core i7-7700K. So, in short, it looks like Intel took an existing CPU, tweaked a couple of knobs, and adapted it for the larger LGA 2066 interface with different packaging. The end product does overclock a little better than Core i7-7700K, we've noticed.

The Motherboard Land Of Confusion

The new X299 chipset, code-named Basin Falls, goes hand-in-hand with the LGA 2066 interface. Intel moved away from its normal tactic of recycling an enterprise-oriented platform controller hub and instead built X299 to be a beefed-up version of the Z270 chipset.

The Core i7-7740X's 16 lanes of PCIe 3.0 and two memory channels present issues, though. X299 must naturally accommodate Core i9-7900X with its 44 lanes of PCIe 3.0 and quad-channel DDR4 controller, but has to be flexible enough to take the Core i7-7740X, too. That means the lower-end CPU limits you to using four of the eight DIMM slots found on corresponding motherboards. Moreover, some PCIe or M.2 slots might not work when you drop in a -7740X.

The processor still features a DMI 3.0 link to X299, which is similar to a four-lane PCIe 3.0. The chipset supports a hefty 30 HSIO (High Speed I/O) lane allotment that vendors can carve up for expanded functionality. X99 did not support HSIO lanes, so that's a significant addition.

With Core i7-7740X, there isn't a difference between the features you can use on the X299 platform compared to what is already available from the Z270 chipset. X299-based motherboards are undoubtedly high-end parts, and they bear the price premium (~$150) to prove it. That means you're paying for features that you cannot use, which is never a positive selling point.

To reduce costs, Intel tells us that motherboard partners can build Kaby Lake-X-specific platforms with only four DIMM slots to accommodate the new processors. We haven't seen any, though, and we haven't had a single vendor tell us they're working on one.

To make matters worse, none of Skylake-X’s unique features are available from Core i7-7740X, including Turbo Boost Max 3.0, AVX-512 support, the integrated vROC (Virtual RAID on CPU) controller, or Intel's mesh/cache adjustments.

All of these factors combine to create a questionable addition to Intel's high-end desktop portfolio. The company tells us that Kaby Lake-X offers a cheaper way into X299 with an upgrade path for later. We don't think many enthusiasts are interested in paying more now for the potential to upgrade later, though. Can Intel prove us wrong?

MORE: Best CPUs

MORE: Intel & AMD Processor CPU Comparison Hierarchy

MORE: All CPUs Content

Test Setup

Test Systems

We're using the MSI X299 Gaming Pro Carbon AC for testing. We disabled the all-core Turbo Boost feature to ensure a level playing field.

There were several demonstrations of memory overclocking beyond DDR4-4000 with Kaby Lake-X processors at Computex, but we settled on DDR4-3200 for our overclocked config to match the memory data transfer rates of our Ryzen models.

We introduced our new test system and methodology in How We Test Graphics Cards. If you'd like more detail about our general approach, check that piece out.

AMD Ryzen 5 1600X

AMD Ryzen 7 1700X

AMD Ryzen 7 1800X

The only updated components in our German lab are the CPU, system memory, motherboard, and new cooling solution, so we'll just provide a quick overview in the following table:

Test Equipment and Environment
System	Germany Intel LGA 2066Intel Core i9-7900X, i7-7740X Core i7 7740X, Core i5-7640X Core i7 7740X, Core i5-7640XMSI X299 Gaming Pro Carbon AC4x 4GB G.Skill Ripjaws 4 DDR4-2600AMD Socket AM4 WorkstationAMD Ryzen 7 1800X, 1700X, 1600X, 1500XMSI X370 Tomahawk2x 8GB G.Skill Ripjaws V DDR4-3200Intel LGA 2011v3Intel Core i7-6900KMSI X99S XPower Gaming Titanium4x 4GB Crucial Ballistix DDR4-2400Intel LGA 1151Intel Core i7-7700K, i5-7600KMSI Z270 Gaming 72x 8GB Corsair Vengeance DDR4-3200 @2400 MT/sAll SystemsGeForce GTX 1080 Founders Edition (Gaming)Nvidia Quadro P6000 (Workstation)1x 1TB Toshiba OCZ RD400 (M.2, System)2x 960GB Toshiba OCZ TR150 (Storage, Images)Be Quiet Dark Power Pro 11, 850W Power Supply Unit (PSU)Windows 10 Pro (All Updates)Be Quiet Dark Power Pro 11, 850W Power Supply Unit (PSU)Windows 10 Pro (Creators Update)U.S.Intel LGA 2066Intel Core i9-7900XMSI X299 Gaming Pro Carbon AC4x 8GB Corsair Vengeance DDR4-2666 @ 2666 and 3200 MT/sIntel LGA 2011v3Intel Core i7-6900K, Core i7-6950XASRock X99 Extreme44x 8GB Corsair Vengeance DDR4-2666 @ 2666 and 3200 MT/sAMD Socket AM4 Workstation AMD Ryzen 7 1800X, 5 1600XMSI X370 Xpower Gaming Titanium 2x G.Skill FlareX DDR4-3200 @ 2666 (stock), and 3200 MT/s Intel LGA 1151 Intel Core i5-7600K, i7-7500 MSI Z270 Gaming M7 2x G.Skill FlareX DDR4-3200 @ 2400 All EVGA GeForce GTX 1080 FE 1TB Samsung PM863 SilverStone ST1500, 1500W Windows 10 Creators Update Version 1703
Cooling	GermanyAlphacool Eiszeit 2000 ChillerAlphacool Eisblock XPXThermal Grizzly Kryonaut (For Cooler Switch)USCorsair H115iCorsair H100i v2Arctic MX4
Monitor	Eizo EV3237-BK (Workstation, Office, HPC)
Power Consumption Measurement	Contact-free DC Measurement at PCIe Slot (Using a Riser Card) Contact-free DC Measurement at External Auxiliary Power Supply Cable Direct Voltage Measurement at Power Supply 2x Rohde & Schwarz HMO 3054, 500MHz Digital Multi-Channel Oscilloscope with Storage Function4x Rohde & Schwarz HZO50 Current Probe (1mA - 30A, 100kHz, DC) 4x Rohde & Schwarz HZ355 (10:1 Probes, 500MHz) 1x Rohde & Schwarz HMC 8012 Digital Multimeter with Storage Function
Thermal Measurement	1x Optris PI640 80Hz Infrared Camera + PI Connect Real-Time Infrared Monitoring and Recording

MORE: Best CPUs

MORE: All CPUs Content

VRMark, 3DMark & AotS: Escalation

As expected, we notice slight variation between the stock Core i7-7700K and its near-identical -7740X counterpart. Testing at each chip's default memory settings (DDR4-2400 for the -7700K and DDR4-2666 for the -7740X) likely contributes some of the difference. And we notice the same behavior with the stock and overclocked settings.

3DMark's DX11 and DX12 tests follow a similar pattern. The beefier multi-core CPUs leverage their extra threads to take commanding leads in these compute-intensive tests. But tuning brings the -7740X within striking range of the overclocked six-core Ryzen 5 1600X. Intel's Kaby Lake-X model enjoys a greater lead over the -7700K in the DX11 physics test, which we chalk up to an extra 200 MHz of overclocking headroom.

We encountered some strangeness during the 3DMark API overhead test. Core i7-7740X lags its -7700K sibling by a large margin during the DX12 and Vulkan sequences; overclocking does little to correct the disparity. We suspect that this is a 3DMark-specific issue, since most of our DX12-based gaming benchmarks don't concur with these results. The API test is relatively new and has received a few recent updates. It's possible that we see another patch to address this unexpected measurement.

Ashes of the Singularity: Escalation

Ashes of the Singularity: Escalation favors core count over other factors, so we observe the expected variance between 4C/8T Core i7 CPUs and the more well-endowed processors.

Core i7-7740X exhibits a minor gain compared to the -7700K at both stock and overclocked settings. Interestingly, we encounter less variability and frame time outliers with the -7740X than the -7700K, which could be a result of slightly more aggressive memory settings. The slight bump in overclocked frequency only equates to a 2.34% advantage, though.

MORE: Best CPUs

MORE: All CPUs Content

Civilization VI & Battlefield 1

Civilization VI AI Test

Clock rate and IPC throughput tend to win the day in this test, so it isn't surprising to find both Core i7s vying for supremacy.

It was less expected to find the 10-core Core i9-7900X mixing it up. Surely, a tuned Core i7-7700K would be faster at 4.9 GHz, right? The new Skylake-X model likely benefits from to its reworked cache hierarchy.

Again, the differences are slight between Kaby Lake-X and Kaby Lake-S. They're such similar processors, after all.

Civilization VI Graphics Test

Speaking of the -7900X's reworked cache hierarchy, as we noticed in our initial review, the processor does suffer from mesh-imposed performance limitations in some titles. That comes into play during this benchmark.

The 10-core Broadwell-E-based chip leads, but both tuned Core i7s are competitive with the rest of our contenders. The difference between Core i7-7700K and -7440X is negligible at stock settings, and therefore unnoticeable during gameplay. But the -7740X's extra bit of frequency headroom yields a larger delta when we overclock.

Battlefield 1 (DX11)

Our armor-laden walk through the O La Vittoria scene in Battlefield 1 yields large performance variations when we test less-powerful processors. The differences shrink as we benchmark higher-end CPUs, mostly because we're graphics-bound.

We encountered an almost-predictable variance between the two Core i7s in most tests, and that plays out in Battlefield 1 when we measure the stock configurations. A 0.2 FPS delta is simply imperceptible.

The story changes with the application of a robust overclock, though. Try as we might, we simply could not get Intel's -7740X to match the -7700K in an overclocked state. That's particularly baffling due Kaby Lake-X's higher overclocked frequency. Nevertheless, the results for both overclocked Core i7s are repeatable, so it's possible that BIOS updates or chipset drivers could improve performance down the road.

MORE: Best CPUs

MORE: All CPUs Content

GTA V, Hitman & Shadow of Mordor

Grand Theft Auto V

We measure performance during Grand Theft Auto V's F-16 flight sequence with the built-in benchmark.

A now-familiar pattern emerges during the Grand Theft Auto V test sequence. There's a slight difference between the two Core i7s at stock settings, which widens as we dial in the -7740X's higher overclock. We're looking at a 2% advantage favoring Kaby Lake-X; the 2 FPS gain certainly doesn't warrant a step up to the much more expensive X299 platform, though.

The Core i9-7900X is brutally competitive in this test, which runs counter to our expectations of Intel's massively-parallel HEDT CPUs that historically prioritize core count over clock rate. If the -7900X could consistently deliver this level of performance in all titles, it would be even more impressive.

Hitman (2016)

A stock Core i7-7740X trails the -7700K, though tuning changes that dynamic. Both Core i7s stand out from the pack at higher frequencies.

AMD's Ryzen processors lag notably behind their Intel competitors. Still, they average more than 110 FPS.

Middle-earth: Shadow of Mordor

The graphics bottleneck is apparent in Shadow of Mordor. A 1.7 FPS delta between the fastest and slowest processors doesn't give us much to comment on. It's most notable, then, that even an affordable Ryzen 5 1600X can push EVGA's beefy GeForce GTX 1080 FE to its fullest in some games.

MORE: Best CPUs

MORE: All CPUs Content

Project CARS & Rise of the Tomb Raider

Project CARS

We spot a bit more variation between the stock Core i7s, perhaps suggesting some sensitivity to memory frequency. Project CARS scales very well with increased host processing resources, so the brawny Core i9-7900X unperforms our expectations. Again, the -7740X's gain over the -7700K from overclocking is a mere 1.4%.

Rise of the Tomb Raider

Rise of the Tomb Raider recently received a patch that significantly improves Ryzen's DX12 performance, making this one more title optimized for AMD's latest platform. We're still sticking to DX11-based testing, though, to keep the playing field as level as possible.

A stock Core i7-7740X loses by 1.3% to the -7700K, but a bit of overclocking turns the tables by 1%. That certainly isn't worth the X299's price premium.

We clearly see the delta between AMD's Ryzen and Intel's Core processors grow at the test sequence's tail end, where it becomes more CPU-intensive. Core i7-7740X encounters a notable frame time outlier towards the end of the test. Meanwhile, the Ryzen 5 1600X briefly spends time under the 90 FPS mark.

MORE: Best CPUs

MORE: All CPUs Content

Workstation & HPC Performance

Preamble

As we dig through these results, remember that semi-professionals almost never overclock their systems. Also, CPUs like Intel’s Core i9-7900X can be hard to keep cool. Consequently, our CPUs are now running at their stock frequencies.

2D Benchmarks: DirectX & GDI/GDI+

Adding Intel’s new CPU to our AutoCAD 2D and GDI/GDI+ graphics benchmarks doesn’t yield any surprises. The newcomer falls into place exactly where you’d expect based on its clock rate.

2D Benchmarks: Adobe Creative Cloud

The outcome depends on frequency again in this next round of benchmarks.

Interestingly, the Kaby Lake-X CPU lands right behind Core i7-7700K, if only by a hair. This won’t be the last time we see this inexplicable phenomenon.

3D Benchmarks: DirectX & OpenGL

In some instances, Core i7-7740X pulls ahead of the -7700K. In others, it's beaten slightly. Either way, the performance differences are generally very small and within the measurement error range. The only discernible pattern seems to be that Kaby Lake-X leads when compute tasks run in parallel with graphics output. This means the differences could be platform-oriented.

CPU Performance: Workstation

The 3D graphics performance we just measured isn’t all that matters to professional rendering titles. Applications run many other tasks (like simulations, compute jobs, preview rendering) on the CPU simultaneously. The full picture’s only achievable by looking at both of them together.

Intel's Core i7-7740X does well in workloads that emphasize high clock rates over core count.

CPU Performance: Photorealistic Rendering

Final rendering doesn’t need a jack of all trades. Instead, efficiency and fast parallel computation are key. This is why photorealistic rendering gets its own test section.

AMD’s Ryzen CPUs do really well, while Kaby Lake-X doesn't. Hyper-Threading keeps Core i7-7740X from losing sight of the field, but we have to wonder whether it deserves a place in Intel’s high-end desktop line-up.

CPU Performance: Encoding & Compression/Decompression

The Kaby Lake-X-based processor fares reasonably well when asked to perform simple decompression. However, all of the other tasks in this category end in total defeat.

High-Performance Computing (HPC)

This final benchmark discipline spells doom for Intel’s Core i7-7740X. Its designation as a high-end desktop part suggests it should do well, but that's just not the case. With four Hyper-Threaded cores, the -7740X carves out a spot in the back of the field.

Intel’s Core i9 processors might be decent options for the workstation space, but the same cannot be said for Core i7-7740X. This is a consumer CPU that was given the wrong moniker. Despite its lower price point, the expensive X299 platform negates a lot of those savings.

Furthermore, the new Kaby Lake-X-based model doesn't really outperform the mainstream version. Core i7-7700K is positioned correctly, and its successors should have continued along those lines instead of pretending that they are something they're not. Case in point: the performance differences between -7740X and -7700K in the workstation and HPC space are even smaller than in our gaming benchmarks. So, an expensive platform doesn't get you very far.

MORE: Best CPUs

MORE: All CPUs Content

Power Consumption & Overclocking

Intel isn't using a soldered integrated heat spreader (IHS) for Skylake-X and Kaby Lake-X. Instead of the metallic solder most enthusiasts want to see, heat moves from the die to the IHS through inexpensive thermal interface material, which is just a fancy name for common thermal paste. This decision has implications for both our power consumption measurements and overclocking efforts. Fortunately, in light of the lower amount of waste heat produced by Core i7-7740X compared to Core i9-7900X, this isn't as much of a problem.

In order to produce comparable results, we used our industrial-grade Alphacool Eiszeit Chiller 2000 cooler that was needed for the Skylake-X review. A decent all-in-one liquid cooler would have been fine, though.

Power consumption is measured after the voltage converters and CPU, using points on the motherboard. These numbers need to be part of the equation for sizing your PSU. And depending on the motherboard you pick, they can differ by up to 30W. Consequently, our numbers are for the CPU only.

Power Consumption

At idle, Intel’s new Kaby Lake-X model does a lot better than -7700K. It also pulls ahead of AMD's Ryzen processors. Most likely, a lot of this is due to the disabled HD Graphics engine.

AutoCAD 2015 doesn’t stress CPUs all that much, since it can't fully utilize their available resources. Still, Intel’s Core i7-7740X consumes more power than its predecessors. Admittedly, the difference isn't huge. However, it is larger than our measurement error range and completely reproducible.

The situation is similar in our gaming workload. Interestingly, the average frame rates are almost identical as well.

The stress test produces somewhat surprising results. Intel’s Kaby Lake-X model edges out the -7700K. We don't know why the pattern reverses under full load. However, the delta is almost imperceptible in the 2W range.

Overclocking & Stability

Does a larger heat spreader and disabled graphics engine yield a meaningful advantage to Core i7-7740X's overclocking headroom? Before we start tweaking, we’d like to provide some background by showing a graph based on a motherboard manufacturer’s binning program. The voltages needed for Core i7-7740X are very informative:

The Core i5-7640X we'll be reviewing soon made it all of the way to 5.1 GHz, putting it in the middle of the pack. In other words, we have a thoroughly average sample.

We tested for stability at every frequency step using Cinebench R15, The Witcher 3, LuxRender, and Battlefield 1. Notice the deliberate exclusion of the AVX-heavy Prime95.

The Core i7-7740X's curve stays fairly straight, which means that power consumption doesn’t explode as clock rate goes up. Rather, the CPU chugs along until it just quits all of a sudden.

A closer look at the performance and power consumption curve shows that performance scales in an almost completely linear fashion with frequency, whereas power consumption increases faster, but not extremely so. Relating single-core performance to the necessary power yields a very similar picture. The resulting waste heat isn’t as bad as some might have feared. At 5.1 GHz, we’re looking at 90W. A good closed-loop liquid cooler should be able to handle this.

In spite of our sample being average, it yielded a good overclocking experience, especially since performance gains and power consumption increases never drifted too far apart.

MORE: Best CPUs

MORE: All CPUs Content

Temperature Curves & Delta Values

Cooling Nuclear Option: The Chiller

Again, we used our Alphacool Eiszeit Chiller 2000 to produce results that are comparable to those from our Core i9-7900X review. We also used this cooler in all of our AMD Ryzen reviews, which means that those results can be compared as well.

High Temperature Differences Challenge Cooling Performance

Kaby Lake-X suffers from Intel's use of thermal paste between the die and heat spreader. Solder would have been the way to go. The difference between the water block’s temperature, which is held at a constant 20°C, and the CPU temperature (according to its sensors) demonstrate this point nicely.

The curve below shows clearly that waste heat isn't dissipated quickly enough. Just as we did in our AMD Ryzen and Intel Core i9 launch articles, we used a very thin copper plate to measure the heat spreader's temperatures as well.

The next graph represents the temperature difference between the top of the CPU's heat spreader and cores.

Our cooling solution is one of the best you can buy. Nevertheless, we measure a 48°C difference between the top of the Core i7-7740X's heat spreader and its cores. This isn't a horrible result, but it could have been a lot better.

Leakage

Next, we measured power consumption at an identical load using different cooling solutions. The leakage currents were below our measurement equipment’s accuracy range. For this reason, we didn't bother with a graph. Kaby Lake-X doesn't have a problem with leakage currents that could increase power consumption in a meaningful way.

MORE: Best CPUs

MORE: All CPUs Content

Final Analysis

Superficially, Intel’s idea to transplant Kaby Lake onto its high-end desktop platform seems like it'd make sense. But ultimately, X299's excessive cost hamstrings Kaby Lake-X. In many ways, the Core i7-7740X reminds us of Intel’s Core i3-7350K. It's a great chip forced into a platform that doesn't match up with its price point.

Core i7-7740X is nearly identical to the existing -7700K; even its price is similar. At stock settings, the -7740X performs a lot like its mainstream counterpart, despite slightly higher base and Turbo Boost frequencies. We noticed the same performance characteristics as -7700K in lightly threaded workloads, such as gaming and most common desktop apps. But the -7740X pales in comparison to the brawnier Skylake-X models in heavy application tests. Lower idle power consumption numbers are likely a result of the -7740X's disabled graphics engine. However, the overall power and thermal measurements come out to be almost the same. Though we still don’t approve of the thermal paste Intel uses between its die and heat spreader, at least closed-loop liquid coolers are fine for this 112W chip.

The “X” in Kaby Lake-X promises extreme overclocking action, and we did eke out a few hundred megahertz over Core i7-7700K. That extra headroom is nice, but it doesn't translate to large gains in our gaming suite. In fact, at stock settings, the -7700K achieved higher average frame rates, while overclocking gave the -7740X a 1.19% advantage. The 99^th percentile variances between the chips both fell under 1%, with the advantage going to Core i7-7700K in stock and overclocked configurations. In either case, the difference between these two CPUs is imperceptible, even after overclocking them both. Perhaps that'll improve as X299-based motherboards receive more fine-tuned firmware.

Intel's Core i7-7700K is one of the fastest gaming processors we've ever tested, so matching it is a good thing. But the -7740X doesn't make sense as an equal when we factor in the X299 platform. You are going to pay an extra ~$150 for an X299 motherboard compared to a more mature Z270-based setup, and gain exactly zero features. In fact, you lose the integrated GPU that could be useful for other tasks, such as Ultra HD Blu-ray playback. Dropping the -7740X into LGA 2066 actually disables motherboard features you're paying for like extra DIMM slots, Skylake-X's more robust power delivery, and hamstrung chipset functionality.

Our motherboard contacts have expressed a general reluctance to create Kaby Lake-X-specific motherboards, and unless that changes, there is little reason to jump up to the X299 with a Kaby Lake-X chip. Certain enthusiasts will pursue a stouter overclock from the Kaby Lake architecture, and -7740X delivers in that regard. Intel's pitch is that Core i7-7740X serves as an entry point to X299 that paves the way for upgrades to Skylake-X later. We don't believe there's a big market for an expensive path like that, though. For those with the upgrade itch, either stay with the Z270 and spend bigger on a high-end GPU, or splurge on a true HEDT-class processor.

Intel has its Coffee Lake CPUs coming to market later this year, and we've heard rumors of a six-core model landing on the mainstream desktop. Consider waiting for more information on that prior to going the X299/Kaby Lake-X route, if you're still interested in Core i7-7740X. Also, AMD's Ryzen Threadripper is coming soon. There are just so many reasons not to buy this thing...

Core i7-7740X doesn’t give you anything new. Yes, it's one of the fastest gaming chips ever made, but you can get similar performance from the mature Z270 platform for significantly less money. We appreciate the higher overclocking ceiling, but that doesn't translate into enough benefit to justify the extra investment.

MORE: Best CPUs

MORE: All CPUs Content

Intel Core i9-7900X Review: Meet Skylake-X

palcorn@outlook.com (Paul Alcorn) — Mon, 19 Jun 2017 13:00:00 +0000

Introduction

Intel’s new Skylake-X processors span the Core i7 and i9 families, but all drop into the same LGA 2066 interface, enabled by the X299 chipset. They're aimed specifically at high-end desktop users who need anywhere from four to 18 physical cores. Meanwhile, existing Skylake-S processors soldier on, popping into LGA 1151-equipped motherboards for mainstream desktops.

The company claims that several architectural enhancements to Skylake-X provide as much as 15% more performance than Broadwell-E in single-threaded workloads, while heavily-threaded tasks should be up to 10% faster.

Dominance in the high-end desktop segment began a decade ago for Intel. Since then, the CPU market has largely been a lopsided affair. Without much competition challenging its position, Intel never saw fit to cut prices or innovate aggressively. AMD's return to the table with Ryzen changed that with more cores, SMT, and unlocked multipliers, all for less money.

Now Intel is looking to defend its flagship against the recently-announced AMD Threadripper CPU, sporting 16 cores, 32 threads, and 64 lanes of third-gen PCIe connectivity. Of course, the enhancements to Skylake-X aren't any sort of knee-jerk reaction. These were in the works for years. Still, Intel moved on today's announcement with unfamiliar speed (tripping on its laces in the process). It also adjusted pricing in a way we haven't seen before. Enthusiasts, enjoy.

Specifications

The 12+ core Skylake-X models are still on the distant horizon, and our Core i7-7440X was inexplicably delayed in shipping. That leaves us with the 10-core Core i9-7900X for today's review. But first, let's discuss the complete line-up of high-end options:

Kaby Lake-X

We aren't accustomed to seeing current-gen architecture in the HEDT portfolio. Typically, the top models lag the mainstream chips by a generation or two. Dropping a pair of Kaby Lake-based chips into LGA 2066 changes this. Fortunately for fans of familiarity, everything else complementing the X299 platform controller hub is Skylake-based...though that could change soon. Earlier this year, Intel announced its 'Data Center First' strategy, which will see the newest processes coming to Xeon products before the desktop. Considering that the HEDT line-up consists of re-purposed data center dies, HEDT may become leading-edge.

In an unprecedented expansion, Intel grows its HEDT family from four models to nine, including those two Kaby Lake-X models. They're a curious addition, supporting two DDR4 memory channels, whereas Skylake-X exposes four. That means you can only use half of your motherboard's DIMM slots with a Kaby Lake-X CPU installed. Fewer PCIe lanes also result in restricted I/O options. Intel disables the on-die HD Graphics 630 engine, allowing the unused silicon to absorb heat and purportedly improve overclocking headroom. Aside from slightly higher base clock rates and a higher 112W TDP, the Core i5-7640X and i7-7740X are otherwise similar to their Skylake-S counterparts, right down to pricing.

In our opinion, matching "affordable" processors with expensive motherboards evokes images of the Core i3-7350K, which isn't popular due to the same sort of imbalance. Intel tells us that motherboard manufacturers can build low-cost X299 platforms specifically for Kaby Lake-X, but we don't see any of them scrambling to create such a niche product as of yet.

Core i5-7640X

Core i7-7740X

Core i7-7800X

Skylake-X

Intel supports DDR4-2666 on every Skylake-X CPU except Core i7-7800X, up from Broadwell-E's official DDR4-2400 specification. It deliberately disables ECC in order to dissuade the Xeon crowd from adopting more enthusiast-oriented platforms.

Intel hasn't released detailed specifications for the higher-end CPUs, but we expect it'll expose similar memory specifications. We also expect frequencies to decline as core counts increase.

Active Cores	1	2	3	4	Cores 5-10
Intel Core i9-7900X (GHz) Turbo Boost	4.3	4.3	4.1	4.1	4.0

The Core i9-7900X features the same Turbo Boost 2.0 technology enabled on previous-generation processors, except that clock rates are notably higher this time around. Expect 4 GHz with 10 active cores. Intel also arms six of the Skylake-X models with Turbo Boost Max 3.0. The company improved this technology to target the two fastest cores running lightly threaded workloads. In Broadwell-E, Turbo Boost Max 3.0 only accelerated a single core. Both favored cores max out at 4.5 GHz. Naturally, IPC throughput should be up quite a bit, addressing the big disadvantage Intel's big HEDT chips sometimes suffered compared to its nimbler quad-core desktop SKUs. Currently, Turbo Boost Max 3.0 requires a driver on some motherboards. But Intel plans native Windows 10 support to eliminate this in the future.

We also get partial AVX-512 support, meaning the upcoming 18-core flagship should be the first desktop host processor enabling 1 TFLOPS+ of compute performance.

Skylake-X is notably different from Skylake-S in that its cache hierarchy is completely re-worked. Core i9-7900X sports more L2 and less L3, which should improve performance in most applications. A new 2D mesh architecture makes its debut as well. Like AMD's Infinity Fabric, this architectural element isn't a universal win, we're finding (more on this shortly).

Whereas enthusiasts were dismayed to discover that the 10-core Core i7-6950X would sell for $1700+, they should be happy to learn of the 10-core Core i9-7900X's $1000 price tag. Paying a cool grand is the only way to get 44 lanes of PCIe 3.0; stepping down to Core i7-7820X drops you to 28 lanes. As storage migrates to the PCIe bus, those extra lanes could come in useful for SSDs since multi-GPU configurations aren't as popular these days. Intel does expose a new PCIe Virtual RAID on CPU (VROC) feature that allows you to coalesce up to 20 SSDs into a single bootable volume. Notably, you can assemble the RAID array on any available PCIe slot, whereas previous RSTe RAID implementations required a connection to the chipset. Sidestepping the chipset circumvents the bottleneck presented by DMI. This comes at a price, unfortunately. You're forced to purchase an upgrade key that plugs into the motherboard to unlock VROC functionality. Server customers are familiar with this practice, but it's not going to be popular among enthusiasts. We don't even know what the key will cost yet.

Intel does reinstate DMI and PCIe bus overclocking, which should please power users. A new memory controller-PLL trim voltage setting is designed to increase ratio-based memory overclocking capabilities, while a new AVX-512 ratio offset joins the standard AVX offset to control thermals during taxing AVX-enabled workloads.

We encountered a few odd performance anomalies as we tested Core i9-7900X during the past week. This launch certainly feels rushed, and though motherboard firmware updates (from multiple vendors) addressed some of the oddities, others persist. It appears that Intel's Skylake-X models will require a period of optimization, much like AMD's Ryzen processors. Let's take a look at the factors affecting Skylake-X's performance.

MORE: Best CPUs

MORE: Intel & AMD Processor Benchmark Hierarchy

MORE: All CPU Content

Weaving The Fabric

Again, as you'll see in our benchmarks, we ran into some strange performance trends that didn't add up. Given Skylake-X's frequency advantage, reworked cache, and 2D mesh topology, we didn't expect Broadwell-E to stand a chance. But in some cases, the previous-gen flagship outperformed Core i9-7900X. Asked about these anomalies, Intel responded:

...we have noticed that there are a handful of applications where the Broadwell-E part is comparable or faster than the Skylake-X part. These inversions are a result of the “mesh” architecture on Skylake-X vs. the “ring” architecture of Broadwell-E.Every new architecture implementation requires architects to make engineering tradeoffs with the goal of improving the overall performance of the platform. The “mesh” architecture on Skylake-X is no different. While these tradeoffs impact a handful of applications; overall, the new Skylake-X processors offer excellent IPC execution and significant performance gains across a variety of applications.

We covered Skylake-X's mesh architecture in Intel Introduces New Mesh Architecture For Xeon And Skylake-X Processors. Check that piece out for more detail. Of course, there's a lot more to this story, and much of it remains under embargo. But this is a huge change to an already effective design, so it comes as no surprise that the mesh topology doesn't yield extra performance in all of our metrics.

The Background

Interconnects are pathways for moving data between key components inside of a processor, including cores, caches, and PCIe and memory controllers. They affect latency and power consumption, which in turn affects performance and thermal design power.

Intel's ring bus debuted in 2007 with Nehalem, and AMD's HyperTransport was introduced in 2001. Both technologies evolved, but higher processor core counts, more cache, and greater I/O throughput have strained the interconnects. There are a number of ways to improve their performance, though this often requires bumping up data rates, and thus voltage, in order to realize large performance gains.

Intel's bi-directional ring bus, pictured above in red on a Broadwell low core-count die, serves as a good example of the challenge. Data travels a circuitous route to reach the components, and latency amplifies as core count increases. The second image shows the Broadwell high core-count die with 24 cores. Aligning the building blocks into a monolithic bus imposes penalties that make it impractical, so Intel divided the larger die into two separate ring buses. This increases scheduling complexity, and the buffered switches that facilitate communication between the rings add a five-cycle penalty, limiting scalability.

In contrast, AMD introduced its Infinity Fabric with the Zen microarchitecture, currently implemented as two quad-core processor complexes communicating over a 256-bit bi-directional crossbar that also handles northbridge and PCIe traffic. They also share a memory controller. The trip across the Infinity Fabric to the other quad-core CCX and its accompanying cache results in increased communication latency. We detailed the design and measured its latency in our AMD Ryzen 5 1600X Review. We also found that higher memory frequencies can improve the Infinity Fabric's latency characteristics, which is likely one of the key reasons that Ryzen's performance increases with faster memory data transfer rates.

AMD contends that software and platform optimizations can defray some of the performance oddities we've noticed in our testing, and from what we've seen, that is true. AMD's efforts, and an unrelenting string of BIOS, chipset, and software updates, have led to much better performance than we recorded in our inaugural Ryzen 7 review.

AMD's work continues. And now Intel faces the same challenge.

What A Mesh

Intel's 2D mesh architecture made its debut on the company's Knights Landing products. The mesh consists of rows and columns of interconnects between the cores, caches, and I/O controllers. As you can see, the latency-killing buffered switches are absent. The ability to 'stair-step' data through the cores allows for much more complex, and purportedly efficient, routing. Intel claims its 2D mesh features a lower voltage and frequency than the ring bus, yet still provides higher bandwidth and lower latency.

Intel moved the DDR4 controllers to the left and the right sides of the 18-core high core-count die, similar to its Knights Landing design. Previously, they were at the bottom of the ring bus-based designs. The Skylake-X die shot suggests there are six memory controllers (second row down on the right and left columns), so it appears Intel disabled two controllers by default. The company likely uses its smaller LCC die for the Core i9-7900X, though representatives won't say for sure.

Things Get Meshy

Intel designed the mesh to increase scalability. There are trade-offs, however. We turned to SiSoftware Sandra's Processor Multi-Core Efficiency test, which measures inter-core, inter-module, and inter-package latency. The software offers Multi-Threaded, Multi-Core Only, and Single-Threaded metrics. We use the Multi-Threaded test with the "best pair match" setting (lowest latency).

The test measures performance between cores with all possible thread pairs, and for Intel's Core i9-7900X, that results in 189 separate results. We employ a data parser to boil the measurements down into average values.

Processor	Intra-Core Latency	Core-To-Core Latency	Core-To-Core Average Latency	Average Transfer Bandwidth
Core i9-7900X	14.5 - 16ns	69.3 - 82.3ns	75.56ns	83.21 GB/s
Core i9-7900X @ 3200 MT/s	16 - 16.1ns	76.8 - 91.3ns	83.93ns	87.31 GB/s
Core i7-6950X	13.5 - 15.4ns	54.5 - 70.3ns	64.64ns	65.67 GB/s
Core i7-7700K	14.7 - 14.9ns	36.8 - 45.1ns	42.63ns	35.84 GB/s
Core i7-6700K	16 - 16.4ns	41.7 - 51.4ns	46.71ns	32.38 GB/s

The intra-core measurement quantifies latency between threads that are resident on the same physical core, while the core-to-core numbers reflect thread-to-thread latency between two physical cores. Core i9-7900K is most comparable to the 10-core Core i7-6950X, but we included the four-core models as a reference point.

We recorded slightly higher intra-core latency and a larger 10.92ns average latency delta between the Skylake-X and Broadwell-E models. Despite Core i9-7900X's increased latency, we recorded a 17.54 GB/s advantage in average transfer bandwidth. That's a solid 26.7% increase. After generating our first set of -7900X results with DDR4-2666, we followed up with several DDR4-3200 tests and noticed an increase in mesh latency. But we also recorded higher average transfer bandwidth. These results are preliminary, and we are conducting further latency and game testing with different memory transfer rates and timings to provide a more in-depth analysis.

Processor	Intra-Core Latency	Intra-CCX Core-to-Core Latency	Cross-CCX Core-to-Core Latency	Cross-CCX Average Latency	Average Transfer Bandwidth
Ryzen 7 1800X	14.8ns	40.5 - 82.8ns	120.9 - 126.2ns	122.96ns	48.1 GB/s
Ryzen 5 1600X	14.7 - 14.8ns	40.6 - 82.8ns	121.5 - 128.2ns	123.48ns	43.88 GB/s

AMD's Ryzen processors employ a vastly different architecture that yields different measurements. The intra-core latency measurements represent communication between two logical threads resident on the same physical core, and they're unaffected by memory speed. Intra-CCX measurements quantify latency between threads on the same CCX that are not resident on the same core. In the past, we observed slight variances, but intra-CCX latency is also largely unaffected by memory speed. However, we've seen up to a 50% decrease in cross-CCX latency, which denotes latency between threads located on two separate CCXes, by increasing the memory data transfer rate from DDR4-1333 to DDR4-3200.

Fabric Bandwidth

We also plotted the fabric bandwidth results from our tests. Core i9-7900X establishes a large advantage over its Broadwell-E predecessor. The Ryzen processors dwarf Intel's quad-core models, but provide far less average bandwidth than the 10-core Intel CPUs.

MORE: Best CPUs

MORE: All CPU Content

Caching Up & IPC, AVX, Cryptographics

For Skylake-X, Intel shrunk Skylake-S' shared last-level cache and transitioned from an inclusive to a non-inclusive scheme. Efficient caching algorithms that maximize the L2 cache hit rate are a key component of this change. The 'rebalancing' reduces per-core L3 cache to 1.375MB; further, Intel modified it to operate as a victim cache that fills with data evicted from the private L2 cache.

Intel also quadrupled each core's L2 cache from 256KB to 1MB. More lower-latency storage space should have a positive effect on performance, though Intel hasn't said how it implemented the silicon-level changes to its Skylake architecture.

Cache & Memory Latency And Bandwidth

Of course, the mesh's latency and bandwidth have an impact on cache and memory throughput, so we conducted a series of tests to compare our contenders, again using SiSoftware Sandra.

We did spot slightly higher L2 latency from Core i9-7900X compared to Core i7-6950X during the in-page random test, but Skylake-X's L2 latency dropped below Broadwell-E during the sequential access pattern. The multi-threaded cache bandwidth test reported a large performance advantage favoring Core i9-7900X.

IPC, AVX, Cryptographics

Due to limited time with Skylake-X ahead of its launch, we ran a preliminary set of IPC-oriented benchmarks. It's possible that further optimizations, or a more expansive set of workloads, might return different results, but we'll be sussing that out in the days to come.

We set a static 3 GHz clock rate for the following tests. It's important to note that Intel didn't specifically call out increased IPC as the source of its 15% single-threaded and 10% multi-threaded performance gains over Broadwell-E. Instead, the company based its claims on a comparison to Core i7-6950X with pre-silicon Spec*int_rate_base2006 tests (subject to a +/- 5% margin of error).

The single-threaded Cinebench test doesn't show a performance difference between Skylake-X and Skylake-S. However, there is a 1.54% improvement over Broadwell-E. The Ryzen processors clearly don't get as much done per clock cycle, and both trail.

Switching to the multi-threaded Cinebench benchmark exposes a larger difference between Intel's 10-core contenders and the rest of our pool. Core i9-7900X and Core i7-6950X remain the focus, though: we record a 1.93% delta between them.

Core i9-7900X employs two 256-bit AVX FMA units per core that operate in parallel, whereas Ryzen's Zen architecture divides 256-bit AVX operations across two FMA units per core. Intel deactivates one FMA per core on the sub-10-core Skylake-X models. As such, Core i9-7900K has an inherent advantage in the y-cruncher benchmark, a single- and multi-threaded program that computes Pi using AVX instructions. We tested with version 0.7.2.9469, which includes Ryzen optimizations.

The -7900X's single-core SHA2-256 test results are nearly twice that of the two previous-generation models due to Intel's targeted AVX2 optimizations for hashing performance. That same advantage carries over to the threaded test. Intel offers AVX-512 support with the Skylake-X processors but doesn't employ all 11 features in the desktop models. Instead, the company targets specific feature sets at different market segments.

The Zen architecture includes two AES cryptographic accelerators for each core, so it isn't surprising to see Ryzen dominate the single-core AES-256-ECB tests. Intel's processors leverage their core count advantage to turn the tables in the threaded AES workload.

MORE: Best CPUs

MORE: All CPU Content

The X299 Basin Falls Chipset & Test Setup

The Basin Falls X299 Chipset

The Kaby Lake-X and Skylake-X processors drop into an LGA 2066 interface, supported by Intel's X299 chipset. Curiously, the platform controller hub isn't repurposed from Intel's server portfolio this time around. The 14nm, 6W piece of core logic features a four-lane DMI 3.0 connection, which is similar to a PCIe link between the processor and chipset. That's a nice doubling of throughput compared to X99's 2 GB/s DMI 2.0.

Basin Falls, Intel's code-name for X299, supports 30 HSIO (High Speed I/O) lanes that vendors can carve up into eight SATA 3.0 and 10 USB 3.0 ports, but lacks native Thunderbolt 3 and USB 3.1 Gen 2 support. Intel plans to add support in future chipsets, naturally.

X299 supports up to three RST PCIe 3.0 x4 storage devices, though the reduced PCIe lane allocations on many Skylake-X SKUs will limit connectivity options. Intel's I218 Jacksonville LAN PHY and Optane Memory support also make an appearance, but most HEDT systems will likely have an SSD. That’s a better option than Optane Memory, which merely serves as a cache for an HDD.

Although LGA 2066 incorporates more pins than LGA 2011v3, it shares many of the same dimensions. As a result, it's physically compatible with existing LGA 2011v3 coolers. However, Intel recommends water cooling as a minimum requirement due to Skylake-X's TDP rating. Skylake-X processors also have an integrated voltage regulator (IVR) that is similar to the FIVR implementation on Broadwell-E parts.

Like all of Intel's unlocked SKUs, the Core i9-7900X doesn't come with a bundled cooler. The company does sell its $85-$100 Liquid Cooling TS13X as an option, but as we'll discuss shortly, invest in a beefy custom loop if your plans for Skylake-X include overclocking.

Test Systems

X299-based motherboard firmware is evolving rapidly, right up through Skylake-X's launch. Reports indicate spotty Turbo Boost Max 3.0 support, which could lead to dissimilar test results from one review to the next. We tested our MSI X299 Gaming Pro Carbon AC extensively and found that it implements Turbo Boost correctly. We also disabled the all-core Turbo Boost feature to ensure a level playing field.

There were several demonstrations of memory overclocking beyond DDR4-4000 with Skylake-X processors at Computex, but we settled on DDR4-3200 to match the memory data transfer rates of our Ryzen models.

We introduced our new test system and methodology in How We Test Graphics Cards. If you'd like more detail about our general approach, check that piece out.

The only updated components in our German lab are the CPU, system memory, motherboard, and new cooling solution, so we'll just provide a quick overview in the following table:

Test Equipment and Environment
System	Germany Intel LGA 2066Intel Core i9-7900XMSI X299 Gaming Pro Carbon AC4x 4GB G.Skill Ripjaws 4 DDR4-2600AMD Socket AM4 WorkstationAMD Ryzen 7 1800X, 1700X, 1600XAsus X370 Crosshair Hero VI2x 8GB G.Skill Ripjaws V DDR4-3200Intel LGA 2011v3Intel Core i7-6900K, Core i7-6950XMSI X99S XPower Gaming Titanium4x 4GB Crucial Ballistix DDR4-2400Intel LGA 1151Intel Core i7-7700KMSI Z270 Gaming 72x 8GB Corsair Vengeance DDR4-3200 @2400 MT/sAll SystemsGeForce GTX 1080 Founders EditionNvidia Quadro P6000 (Workstation)1x 1TB Toshiba OCZ RD400 (M.2, System)2x 960GB Toshiba OCZ TR150 (Storage, Images)Be Quiet Dark Power Pro 11, 850W Power Supply Unit (PSU)Windows 10 Pro (All Updates)Be Quiet Dark Power Pro 11, 850W Power Supply Unit (PSU)Windows 10 Pro (Creators Update)USIntel LGA 2066Intel Core i9-7900XMSI X299 Gaming Pro Carbon AC4x 8GB Corsair Vengeance DDR4-2666 @ 2666 and 3200 MT/sIntel LGA 2011v3Intel Core i7-6900K, Core i7-6950XASRock X99 Extreme44x 8GB Corsair Vengeance DDR4-2666 @ 2666 and 3200 MT/sAMD Socket AM4 Workstation AMD Ryzen 7 1800X, 5 1600XMSI X370 Xpower Gaming Titanium 2x G.Skill FlareX DDR4-3200 @ 2666 (stock), and 3200 MT/s Intel LGA 1151 Intel Core i5-7600K, i7-7500 MSI Z270 Gaming M7 2x G.Skill FlareX DDR4-3200 @ 2400 All EVGA GeForce GTX 1080 FE 1TB Samsung PM863 SilverStone ST1500, 1500W Windows 10 Creators Update Version 1703
Cooling	GermanyAlphacool Eiszeit 2000 ChillerAlphacool Eisblock XPXThermal Grizzly Kryonaut (For Cooler Switch)USCorsair H115iCorsair H100i v2Arctic MX4
Power Consumption Measurement	Contact-free DC Measurement at PCIe Slot (Using a Riser Card) Contact-free DC Measurement at External Auxiliary Power Supply Cable Direct Voltage Measurement at Power Supply 2x Rohde & Schwarz HMO 3054, 500MHz Digital Multi-Channel Oscilloscope with Storage Function4x Rohde & Schwarz HZO50 Current Probe (1mA - 30A, 100kHz, DC) 4x Rohde & Schwarz HZ355 (10:1 Probes, 500MHz) 1x Rohde & Schwarz HMC 8012 Digital Multimeter with Storage Function
Thermal Measurement	1x Optris PI640 80Hz Infrared Camera + PI Connect Real-Time Infrared Monitoring and Recording

MORE: Best CPUs

MORE: All CPU Content

VRMark, 3DMark & AotS: Escalation

Comparison Processors

Intel Core i7-6950X

AMD Ryzen 7 1800X

Intel Core i7-6900K

VRMark & 3DMark

In spite of its higher Turbo Boost Max 3.0 frequencies and new cache hierarchy, the Core i9-7900X lags behind two Broadwell-E processors at stock settings. Even a 4.5 GHz overclock does little to rectify its position. Core i7-6950X takes the lead here.

The Core i9-7900X fares better during the DX11 and DX12 benchmarks. At its stock clock rate, it even leads the overclocked Core i7-6950X under DirectX 12. This series of benchmarks sets the tone: when Core i9-7900X can leverage its strengths, it provides impressive performance; but in some games it loses out to Broadwell-E.

Raw horsepower doesn't seem to be the problem. After all, -7900X leads in the API tests as well. This could just point back to isolated shortcomings of the mesh topology that Intel conceded to previously.

Ashes of the Singularity: Escalation

Ashes of the Singularity: Escalation became the poster child for game updates after AMD's Ryzen launch. We've endured seemingly weekly patches, which necessitate constant retesting. As a result, Ryzen 7 1800X offers much better performance now than it did at launch. In fact, the 1800X registers impressive gains in most games at stock and overclocked settings compared to our last round of testing.

Optimized game code doesn't help Intel's Core i9-7900X, though. With 10 Hyper-Threaded cores at its disposal, we expect the processor to lead in workloads optimized for threading. But again, it falls behind the overclocked Core i7-6950X in spite of its frequency advantage. The rest of the field falls in according to core count. The Core i7-7700K and Ryzen 7 processors account for a majority of frame time outliers.

MORE: Best CPUs

MORE: All CPU Content

Civilization VI, Battlefield 1 & Deus Ex: Mankind Divided

Civilization VI AI Test

The stock Core i9-7900X benefits from its dual-core 4.5 GHz Turbo Boost Max 3.0 setting to score well in Civilization VI's AI test. An overclock buys us very little in this situation.

This benchmark typically favors clock rate and IPC throughput over core count, and the Core i7-6950X scales well with increased frequency. Intel's stock Core i7-7700K falls into the middle of the pack, though a bit of tuning would likely propel it into a lead.

Civilization VI Graphics Test

The Core i9-7900X at stock settings tumbles to the bottom of our chart, which runs counter to our expectations. Even 10 cores, 20 threads, lots of L2 cache, and Turbo Boost Max 3.0 can't break past the six-core Ryzen 5 1600X at 4 GHz. Overclocking to 4.5 GHz does little to help. It's clear that the mesh-imposed performance penalty will have a profound effect on some titles.

Battlefield 1 (DX11)

On the other hand, Core i9-7900X wields its Turbo Boost frequencies and adjusted cache hierarchy well in other workloads. The stock Skylake-X CPU leads the test pool, even surpassing an overclocked Core i7-6950X.

Looks like Core i9-7650X is shaping up to have a split personality, right? It'll be interesting to see which one dominates in our application testing.

The Ryzen 7 1800X and Core i7-7700K contribute most to the somewhat messy frame time chart.

Deus Ex: Mankind Divided

The Dues EX: Mankind Divided benchmark results filter into two distinct camps. The Ryzen processors land up top, while Core i9-7900X joins its IA-based counterparts at the bottom.

MORE: Best CPUs

MORE: All CPU Content

Grand Theft Auto V, Hitman & Shadow of Mordor

Grand Theft Auto V

We measure performance during Grand Theft Auto V's F-16 flight sequence with the built-in benchmark.

Intel's stock Core i9-7900X beats the Core i7-7700K by a large margin, and a bit of tuning pushes it past the overclocked Broadwell-E model.

An overclocked Ryzen 7 1800X surpasses Core i7-6900K. The Ryzen 7 1800X won't face a significant pricing challenge from the new eight-core Skylake-X chip, but it will be interesting to see them head to head when we get our hands on one.

All of these processors provide a smooth gaming experience. However, the stock Core i9-7900X experiences some frame time variability in the early stages of this benchmark. The -7700K also encounters a familiar spike near the end of the flight sequence.

Hitman (2016)

The Core i9-7900X leads in both its stock and overclocked configurations. Oddly, as we go through our results, it's evident that this processor tends to provide solid performance in lightly threaded games, but struggles in some heavily-threaded titles. Of course, we would have expected the opposite to be true.

The overclocked -7900X suffers a frame time variance spike in the opening stages of this test that's rivaled only by the overclocked Ryzen 7 1800X.

Middle-earth: Shadow of Mordor

We expect the Core i7-7700K to lead in lightly threaded games, and for once, it does.

Core i9-7900X scales well with some tuning to take second place. The overclocked Ryzen processors also fare well in this benchmark, though their stock configurations contribute the only notable frame time outliers. All the CPUs provide solid performance, but it's notable that the lowest-cost processors only trail the leaders by a few FPS.

MORE: Best CPUs

MORE: All CPU Content

Project CARS, Rise of the Tomb Raider & The Division

Project CARS

AMD's Ryzen processors trail most of the Intel competition during this test, while the overclocked HEDT Core processors deliver 100 FPS+ on average.

The overclocked Ryzen 7 1800X and Ryzen 5 1600X offer nearly the same performance as a stock Core i7-6950K. That chip sets itself apart once we crank up its clock rate, though. Only an overclocked Core i9-7900X is faster.

Rise of the Tomb Raider

Rise of the Tomb Raider recently received a patch that significantly improves Ryzen's DX12 performance, making this one more title optimized for AMD's latest platform.

Nevertheless, Core i9-7900X performs well in our DX11-based test, taking first and second place in its overclocked and stock configurations. Broadwell-E is bested just slightly.

The Division

Ryzen 7 1800X trails the Core i7-7700K in its stock form, though a bit of tuning propels it into a lead. The Core i9-7900X beats its predecessor, but trails the less expensive AMD chip and Intel's own Core i7-7700K.

The Ryzen 5 1600X and Core i7-6900K contribute a majority of frame time outliers. However, all of these CPUs provide a smooth gaming experience.

MORE: Best CPUs

MORE: All CPU Content

Workstation & HPC

2D Benchmarks: DirectX and GDI/GDI+

Adding Intel’s new CPU doesn’t yield any surprises in our AutoCAD 2D and GDI/GDI+ graphics benchmarks. The newcomer falls in place exactly where you’d expect based on its frequency.

2D Benchmarks: Adobe Creative Cloud

The same is true in Adobe's CC apps. Again, performance comes down to clock rate. The Intel Core i9-7900X does well in workloads that challenge more than four cores. However, two of the individual scores do show some need for optimization on Intel’s part, which may be weaknesses related to the mesh topology.

3D Benchmarks: DirectX and OpenGL

The Core i9-7900X’s graphics performance across individual applications and suites is consistent with our previous findings. Turbo Boost plays a big role here, since few of the popular applications are (yet) able to utilize more than two cores. This calls for a high IPC, which the Core i9-7900X has a much easier time supplying than the Core i7-6950X.

CPU Performance: Workstation

AMD’s Ryzen CPUs look good here, which is to say that the Core i9-7900X has to go all-out, and even then has a hard time justifying its price tag.

CPU Performance: Photorealistic Rendering

Final rendering doesn’t need a jack of all trades. Instead, efficiency and fast parallel computation are key. This is why photorealistic rendering gets its own test section. Intel's Core i9-7900X does really well, beating some of the contenders by huge margins. Only the much less expensive AMD Ryzen 7 1800X can keep up every now and then, which is a nice surprise. The Core i7-6950X can’t compete.

CPU Performance: Encoding & Compression/Decompression

These types of tasks also fall into the Core i9-7900X's wheelhouse. Decompression is the only field where Intel’s new processor slows down a bit. It's in good company, though. In the end, this workload is only lightly threaded, preventing the -7900X from really shining.

High Performance Computing (HPC)

Intel’s Core i9-7900X shines once again. However, we also have to call out the Ryzen 7 1800X for its exceptional performance, particularly given a much more affordable price tag.

In most tests, Intel’s new processor does dominate the field. This makes it an interesting alternative to expensive Xeon workstations for semi-professional work.

In the end, Intel’s Core i9-7900X is a good choice for enthusiasts running well-threaded professional tasks, particularly when a Xeon-based workstation might be overkill. Thanks to Turbo Boost, the new processor also performs well when only a small number of cores are in use, emphasizing IPC and clock rate more than parallelism.

This means that the Core i9-7900X kills two birds with one stone by providing high frequencies when just a few cores are active and great multi-threaded performance when dealing with well-threaded tasks. The same certainly can’t be said for its predecessor, the Core i7-6950X.

To be fair, the significantly less expensive AMD Ryzen 7 1800X keeps up with Intel's latest in a number of places. This comes as a surprise, since the numbers we generated when Ryzen launched were markedly worse. Power users who want top performance across the board will definitely get it from Intel, but this choice won’t always result in a speed-up commensurate with Core i9's price premium.

MORE: Best CPUs

MORE: All CPU Content

Power Consumption & Overclocking

First, we had to use our industrial-grade Alphacool Eiszeit Chiller 2000 cooler to achieve meaningful test results. Air cooling is out of the question for Core i9-7900X, and even a decent all-in-one liquid cooling solution won’t get far.

Power consumption is measured after the voltage converters and CPU, using points on the motherboard. The 230A we saw during our overclocking tests prompted us to add a fan to the motherboard. We only had one platform for this story, so we played it safe with extra cooling around the LGA interface.

Power Consumption

These numbers are generated using stock motherboard settings; any significant under-volting had no effect, except to cause stability problems.

At idle, the Core i9-7900X comes in below its predecessor and AMD’s Ryzen models. Getting there required a firmware update to correct problems with P-states, among other issues plaguing this platform right up until launch.

Even lightly threaded workloads like AutoCAD 2015, which uses just a few of the Core i9-7900X’s cores, catapult Intel’s -7900X to the back of the pack in power consumption metrics.

This is almost certainly due to higher clock rates than previous-gen HEDT CPUs like the Core i7-6950X. Taking into account Skylake-X's increased performance, though, the Core i9-7900X remains a more efficient solution. One percent more power consumption gets you 29 percent more 2D performance and a massive 39 percent more 3D performance. Both tasks profit from Turbo Boost Max Technology 3.0 in a big way.

If more than three cores are used, the Core i9-7900X’s advantage shrinks considerably. It still beats the Core i7-6950X by approximately 17 percent, but the power consumption increase is a lot larger (around 10 percent).

Pushing all of the Core i9-7900X’s cores with Prime95 or LuxRender propels power consumption to incredible heights. You do get 48 percent more rendering performance in LuxRender, but at the expense of 58 percent-higher power use. This approach has the elegance of a sledgehammer. Then again, if you need speed at any cost, Core i9-7900X is top-notch.

Overclocking and Stability

We now know that the Core i9-7900X’s performance to power consumption ratio turns negative as you utilize more of its on-die resources. Of course, this has to be factored into your overclocking plans, since many coolers can't cope with the heat dissipated by a >200W processor.

Stable overclocking, defined as reliable operation under Prime95 for prolonged periods of time without hitting a temperature limit, wasn’t possible beyond 4.4 GHz. Reports of >5 GHz with all cores active should be taken with a grain of salt. We did boot into Windows at 5.1 GHz, but running actual applications resulted in either a BSOD or a motherboard emergency shutdown.

We did manage to achieve a stable 4.8 GHz overclock under the single- and multi-core Cinebench R15 benchmarks. However, our cooling solution was probably the decisive factor there. Realistically, 4.5 GHz should be achievable with an all-in-one liquid cooler. Let’s take a look at the comparison curves for Cinebench R15 with all cores active:

At a Vcore of 1.4V, the system stayed stable for 10 benchmark runs. Intel's Core i9-7900X consumed an average of 261W, while individual peaks jumped as high as 293W. A test at 4.8 GHz using 20 instances of a year-long shading computation for a rooftop photovoltaic array, including profit calculation, pushed power consumption all the way to 335W. The motherboard shut down after we started Prime95 without limiting AVX. The last recorded value was 364W.

The performance and power consumption curves yield some interesting findings. Performance scales with clock rate in an almost linear fashion. Power consumption increases faster, but not as rapidly as we feared that it might.

The more pronounced curve is due to relatively high values at lower frequencies, where the “background noise” (load caused by other system components) is more noticeable.

So, what’s the final verdict on overclocking Intel's Core i9-7900X? It depends on the quality of your cooler and its ability to move heat off the CPU as quickly as possible.

MORE: Best CPUs

MORE: All CPU Content

Temperature & Thermal Problems

Cooling Nuclear Option: The Chiller

As mentioned, we had to use Alphacool's Eiszeit Chiller 2000 to achieve usable overclocking results. More conventional thermal solutions just wouldn't cut it. All-in-ones like Corsair's H100i and Enermax's LiqTech 240 hit their limits at stock frequencies under Prime95. The custom loop threw in the towel at 4.6 GHz.

Why can't those liquid coolers keep up with a CPU like the -7900X? Back in the day, a normal all-in-one was good enough to keep the Core i7-5960X running cool, even overclocked to 4.8 GHz. We measured power consumption numbers of up to 250W back then. So, why did we have to force a constant 20°C in the loop to even start experimenting?

High Temperature Differences Challenge Cooling Performance

The reason that Skylake-X is so much harder to cool traces back to the thermal paste Intel chose to use instead of solder between the processor die and heat spreader. Although paste is cheaper, it's also less than ideal for cooling performance.

Intel’s digital temperature sensors report reliable results from 35 to 40°C and up, prompting us to only include values above that range in our analysis. The difference between the water cooling block's temperature, which is held at a constant 20°C, and the CPU temperature reported by Intel's sensors shows just how bad of a choice thermal paste really was.

We measured the CPU heat spreader’s temperature the same way we did when AMD launched Ryzen 7 1800X, by using a thin copper plate. The resulting curve shows very clearly that waste heat can't be dissipated quickly enough. A solution good enough for a thermal lightweight like Intel's Core i7-7700K just doesn’t work for Core i9-7900X.

This curve represents the temperature delta, which is to say the thermal difference between Core i9-7900X’s cores and the top of its heat spreader. The outcome is shocking:

In the end, the delta between the cores and top of the heat spreader reaches 71°C, and that's using one of the best cooling setups money can buy. Naturally, lesser thermal solutions start running into trouble at stock frequencies when you run a stress test.

To illustrate our point, we plotted the temperature for all of the Core i9-7900X’s cores at stock settings running Prime95 or LuxRender. A good custom water-cooling loop does fairly well, which shouldn't come as a surprise. However, no other thermal solution will be able to keep up. Even the motherboard manufacturers we spoke to agree, telling us about their all-in-one liquid coolers running out of headroom as soon as they ran Prime95 without limiting AVX.

A Tcore of up to 65°C and a heat spreader temperature of approximately 24°C make for a difference of more than 40°C. That's at 230W. Once the 300W line is crossed, even the Alphacool Eiszeit Chiller 2000 taps out. This isn’t even difficult to do: with a Core i9-7900X running at 4.6 or 4.7 GHz, using the voltages needed to get there, even simple rendering applications trigger those levels. The highest power consumption numbers we saw were just north of 300W, which had the CPU hitting its 100°C thermal limit consistently. An emergency shutdown followed soon after.

Leakage

Next, we measured power consumption under a constant load using different coolers. For a temperature increase of approximately 40°C, power consumption increases by five percent. This isn’t just an acceptable result, but a really good one. The values above 100°C are not as reliable due to throttling. Consequently, we made an exception and used a low-pass filter that smoothed out the brief decreases.

Everything could have been great, if it wasn't for the thermal paste between Intel's die and heat spreader. Admittedly, most workstation or semi-pro users won't overclock, cutting down on the number of customers affected by this problem. But we all know that affluent enthusiasts attracted to Skylake-X's balance between high frequencies and core counts will have to face a significant cooling challenge. Your choices come down to high-end all-in-one packages or a custom water-cooling loop. Air cooling is completely out of the question if you expect the -7900X to run comfortably under full load.

MORE: Best CPUs

MORE: All CPU Content

Conclusion

Intel’s market dominance burdens the company with certain expectations when it launches new hardware. Naturally, we expect more performance. And although we're quick to deride incremental updates, forward progress is what we want to see. At no point is a step backward alright in our books, and we saw a handful of those in today's tests.

Intel's mesh fabric and AMD's Infinity Fabric demonstrate how highly parallel architectures require more sophisticated interconnects. In some cases, they introduce performance regressions compared to simpler configurations that connect subsystems more directly. Remember, both companies used their previous-generation layouts for over a decade, and early implementations weren't without weaknesses that were later improved.

Most recently, Ryzen faced some puzzling performance issues at launch. More than three months later, a steady stream of firmware, chipset, and software updates has rectified a lot of the issues we initially identified. Even in this story, revisiting Ryzen 7 1800X leaves us with a very positive impression, particularly compared to Intel's $1000+ alternatives. Ryzen didn't magically become the fastest CPU out there, but it's impossible to ignore at its price point.

Enthusiasts might hope for similar improvements from Intel. After all, AMD is overcoming its roadblocks with a fraction of the R&D budget. We asked Intel if it expects software-based optimizations to fix what disappointed us, and company representatives responded that software tuning for the new architectural enhancements and cache hierarchy could improve performance. Remember, though, Core i9-7900X is based on the same micro-architecture as older Core CPUs. It's improbable that mere code updates will rectify issues introduced by Skylake-X's layout when Skylake-S and its derivatives are already well-supported.

As it stands, aggressive Turbo Boost frequencies and a re-balanced cache hierarchy go a long way to improving on Broadwell-E's minor weaknesses. When the Core i9-7900X does well, it really shines. Often, the chip beats every competitor we throw up against it, including Core i7-6950X. In other workloads, latency imposed by its mesh topology causes Core i9 to stumble. That isn’t to say performance falls off completely. But we do see anomalies unfitting of a $1000 CPU. If you're strictly a gamer, Core i9-7900X won't make you want to buy a new CPU, motherboard, and memory kit.

Enthusiasts also want to see robust overclocking capabilities, and Skylake-X does offer a higher frequency ceiling than Core i7-6950X. You're going to cope with a lot of heat in the process, though. Given Intel’s insistence on using thermal paste between its die and heat spreader for longer-term reliability, the processor can’t dissipate heat as effectively, so thermal performance becomes a limiting factor. Plan on investing in a beefy open loop if you want to push the Core i9-7900X much further than its stock frequencies.

Core i9-7900X performs well in our productivity, workstation, and HPC tests. The mesh-imposed disparities aren't as pronounced in those benchmarks. But we also have re-run Ryzen 7 1800X benchmarks to think about. Pressure to size up has pushed AMD's flagship down to $460, less than half of what a Core i9-7900X would cost. While Intel may capture the top 1% of high-end enthusiasts with Skylake-X, everyone else has to consider whether Ryzen may be the smarter buy.

Moreover, AMD's upcoming Threadripper CPU has to have Intel worried. How do we know? The X299 motherboards we used needed firmware updates to address very serious performance issues right up until launch. Intel didn't seem nearly as ready for Skylake-X's introduction as we'd expect. A number of Core i9s with even more cores won't be ready until later this year. However, it looks like Intel couldn't get the four-, six-, eight-, and 10-core models out fast enough. They'll ship later this month.

Unfortunately, this story won't be ready to wrap up until we have Threadripper to compare against. Given Core i9-7900X’s high price and performance caveats, enthusiasts should probably hold off on a purchase until we know more about the competition, even if Skylake-X looks like a bigger step forward than Intel's past HEDT designs.

MORE: Best CPUs

MORE: All CPU Content

AMD Processor Price List

Michael Justin Allen Sexton — Sun, 14 May 2017 01:00:00 +0000

AMD Processors

AMD recently breathed new life into its microprocessor business with the release of its Ryzen CPU architecture, which is significantly faster than the company's older Bulldozer, Piledriver, and Kaveri based processors. While Ryzen occupies the high-end of the CPU market, AMD still produces APUs and CPUs based on its older microarchitecture designs to handle the low-end.

MORE: AMD AM4 X370 Motherboard Price List

MORE: AMD AM4 B350 Motherboard Price List

Socket AM1

Sempron

AMD Sempron 2650View Deal

MORE: Best Builds

MORE: Best Cases

MORE: Best Cooling

Socket FM2

A4

AMD A4-6300View Deal

AMD A4-7300View Deal

A6

AMD A6-5400KView Deal

AMD A6-6400KView Deal

MORE: Best CPUs

MORE: Best Deals

MORE: Best Graphics

Socket FM2+

Athlon

AMD Athlon X4 860KView Deal

AMD Athlon X4 870KView Deal

AMD Athlon X4 880KView Deal

A6

AMD A6-7400KView Deal

A8

AMD A8-7600View Deal

AMD A8-7650KView Deal

A10

AMD A10-7860KView Deal

AMD A10-7870K Black EditionView Deal

AMD A10-7890K w/WraithView Deal

MORE: Best Memory

MORE: Best Monitors

MORE: Best Motherboards

Socket AM3+

AMD FX

AMD FX-4300View Deal

AMD FX-4350View Deal

AMD FX-6300View Deal

AMD FX-8300View Deal

AMD FX-8320EView Deal

AMD FX-6350View Deal

AMD FX-8320View Deal

AMD FX-6350 w/WraithView Deal

AMD FX-8350View Deal

AMD FX-8350 w/WraithView Deal

AMD FX-9590View Deal

AMD FX-9370View Deal

AMD FX-8370 w/WraithView Deal

AMD FX-8370EView Deal

MORE: Best Power Supplies

MORE: Best SSDs

Socket AM4

Ryzen 5

AMD Ryzen 5 1400 w/Wraith StealthView Deal

AMD Ryzen 5 1500X w/Wraith SpireView Deal

MORE: AMD Ryzen 5 1500X CPU Review

AMD Ryzen 5 1600 w/Wraith SpireView Deal

AMD Ryzen 5 1600XView Deal

MORE: AMD Ryzen 5 1600X CPU Review

Ryzen 7

AMD Ryzen 7 1700View Deal

MORE: AMD Ryzen 7 1700 CPU Review

AMD Ryzen 7 1700XView Deal

MORE: AMD Ryzen 7 1700X CPU Review

AMD Ryzen 7 1800XView Deal

MORE: AMD Ryzen 7 1800X CPU Review

How To Overclock AMD Ryzen CPUs

Jean Michel "Wizerty" Tisserand — Thu, 04 May 2017 15:30:00 +0000

Our Ryzen Overclocking Platform

Unless you've been living under a rock, then you already know that Ryzen is AMD's newest brand that now covers eight-, six-, and four-core desktop-oriented CPUs. In the months to come, we'll also see Ryzen-branded APUs that you can see in our CPU Benchmark Hierarchy.

For now, though, we're interested in how receptive the available host processors are to CPU overclocking. After all, AMD arms the entire portfolio with adjustable ratio multipliers, practically inviting enthusiasts to throw their skill and cooling budgets up against the company's first 14nm CPUs.

Is Ryzen an overclocker's dream, then? We did some preliminary tuning for our launch coverage using early firmware. And as AMD fills in its Ryzen 7 and Ryzen 5 families, we revisit the architecture's headroom using fairly high-end air and closed-loop liquid cooling. But so far we've seen a common ceiling in the 3.9 to 4 GHz range.

We've already done quite a bit to explore Ryzen's performance in stock and overclocked form. For more, check out our:

Today's story goes a step further. We're tweaking the BIOS in every direction using an eight-core model to gauge the impact of various settings on your overclocking experience. For cooling, we're going with water-cooling, though liquid nitrogen will follow as the platform matures. And we'll test the influence of memory speed and reference clock rate.

Test Configuration

Some of our labs got their CPUs from AMD, some received processors from motherboard manufacturers, and others had to buy retail chips when AMD's supply ran out. Tom's Hardware France was fortunate enough to receive all three launch models, Ryzen 7 1700, 1700X, and 1800X, from the source. To torture them, we surrounded ourselves with some of the best hardware available.

G.Skill Flare X (2x 8GB)View Deal

Cooler Master - MasterWatt Maker 1200View Deal

For cooling, we chose the Silent Loop 280 liquid cooling kit from be quiet! Although our various offices are using different motherboards, we went with Asus' Crosshair VI Hero for this story. We're arming it with two memory modules from a G.Skill Flare X kit, which was developed specifically for Ryzen. In fact, the kit employs Samsung B-die ICs, considered some of the best for overclocking. Is this a perfect combination? We'll see. To avoid graphics bottlenecks, we're using an Asus ROG Strix GTX 1080. To finish, our faithful Cooler Master MasterWatt Maker 1200 is tasked with providing clean power to our platform.

MORE: Best CPUs for Gaming
MORE: CPU Benchmark Hierarchy
MORE: How to Overclock a CPU

AMD Ryzen 7 1700View Deal

BIOS Options

Overclocking Options

Overclocking through Windows is convenient, but we still prefer locking in new settings through the BIOS. This is doubly important, we think, when it comes to a brand new platform like AMD's Ryzen.

What follows are the most notable settings from our Asus board's BIOS:

Ai Overclock Tuner: Set this to manual for access to reference clock adjustments, or set it to D.O.C.P. to pick a memory overclocking profile (related parameters are adjusted automatically).
BCLK Frequency: We'll come back to this very important setting later in the article. It can be adjusted from 85 to 145 MHz, according to the settings. By default it should be near 100 MHz, but we have recommendations of our own.
CPU Core Ratio: This is a multiplier used to determine the processor frequency. Raise it to overclock your CPU. Be careful, though. When the ratio is not set to Auto, AMD's XFR technology is deactivated. The processor switches automatically to overclocking mode and all power-saving features are deactivated. It is adjustable in increments of 0.25x.
Memory Frequency: System memory frequency, from 1333 to 3200 MT/s. You can go even further using the BCLK Frequency field. This is a very important parameter for optimizing performance, and we'll go into more depth on it shortly.
SMT Mode: Simultaneous Multi-Threading is similar to Intel's Hyper-Threading. Be careful; we had trouble with our motherboard when this option was set to Activated. Leave it on Auto; SMT will still be activated and you won't run into the bugs we did.
CPU Core Voltage Override: This allows you to regulate your processor's core voltage. Asus advises against exceeding 1.4V. AMD recommends a 1.35V maximum voltage for long-term overclocks, and although the company says Ryzen can withstand 1.45V, longevity may be affected.
DRAM Voltage: Raising the RAM voltage can help stabilize an overclock. A value of 1.35V is generally sufficient. For lofty overclocks coupled with aggressive timings, this can be raised up to 1.8V without additional cooling if you're using Samsung B-die chips.
PLL Voltage: We recommend manually locking this to 1.8V. If you leave it set to Automatic, there's a risk that your motherboard will raise it, resulting in a temperature increase. It's a useful knob to have when overclocking under liquid nitrogen (notably for the RAM).

If you click on the External Digi+ Power Control sub-menu, you get these options:

Load-line Calibration: According to our testing, Level 1 yields the best results.
CPU Current Capability: To avoid overclocking limitations, we recommend raising this to 140%.

As a reminder, LLC helps stabilize the core voltage under taxing CPU loads. When the processor is at rest (idle), it consumes little energy; therefore, the 1.35V it needs is easily supplied. Once the workload intensifies, though, voltage drops a bit (to 1.3V, for example). Obviously this is not good for stability. Motherboards equipped with load-line calibration functionality can increase core voltage under load to offset this effect.

LLC Testing

We tried a number of the modes offered by Asus' Crosshair VI Hero, and recorded their results using our voltmeter.

With the CPU set to 1.35V, all of these profiles fed our sample close to 1.357V at idle. Under load, however, they can increase significantly.

Setting	Idle	Load
LLC 1 (Auto)	1.357V	1.36V
LLC 2	1.357V	1.37V
LLC 3	1.357V	1.4V
LLC 4	1.357V	1.42V
LLC 5	1.357V	1.44V

It looks like the LLC setting is a little aggressive, given the voltage overshoot at idle. At Level 1, the increase under load is reasonable and won't cause any problems. Our testing showed the Auto mode's results to be similar, but lock in Level 1 just to be sure.

The last level is fairly crazy, pushing 1.44V instead of 1.35V.

Be careful with CPU voltage. Setting 1.35V in the BIOS does not mean the processor receives this voltage, proven here with LLC 5, which adds nearly a tenth of a volt. For the remainder of today's story, the values we're reporting are verified with help from Asus' test points.

	Ambient Recommended	Ambient Maximum	LN2 Recommended	LN2 Maximum
CPU Core Voltage	1.40V	Up to 1.45V	1.80V	Up to 1.95V
SOC Voltage	1.15V	Up to 1.30V	1.20V	Up to 1.40V
DRAM Voltage1	1.40V	Up to 1.90V	1.80V	Up to 1.90V
1.8V PLL Voltage	1.80V	Up to 2.10V	3.00V	Up to 3.20V
1.05V SB Voltage	1.05V	Up to 1.40V	1.30V	Up to 1.40V
1.8V Standby Voltage	1.80V	Up to 2.10V	2.10V	Up to 2.30V
2.5V SB Voltage	2.50V	Up to 2.80V	2.70V	Up to 2.80V
1 Depends on the DRAM sticks; the limit is considered from CPU IMC side.

DRAM Timings

The DRAM Timing Control sub-menu grants access to memory timings. No, you aren't seeing things; there are currently only five options. The right memory settings are very important on this platform, so we'll dedicate an entire page to optimizing them properly.

Not all of the other parameters available in the DRAM Timing Control After Training menu are applied. At least, that's the case in firmware version 5803, which we used. Perhaps AMD will allow motherboard manufacturers to unlock these settings in future BIOS releases.

A BIOS Cheat?

To improve performance in Windows, AMD recommends using the High performance power plan. What impact does this have? In order to answer that question, we tried it out.

Also, we turned on the BIOS-based “Performance Bias” option and ran Geekbench 4. This option purportedly allows overclockers to score more points in competitions. Our Geekbench scores are the average of three consecutive runs.

Geekbench 4

Configuration	Single-Core	Multi-Core	Mem. Copy	Mem. Latency	Mem. Bandwidth
Mode: NormalBias: Auto	4705	23774	8183	5878	7276
Mode: PerformanceBias: Auto	4729	24240	8278	5885	7312
Mode: PerformanceBias: Geekbench 4	4756	24739	8528	6004	7369

Setting the power plan to High performance does yield a speed-up (a small one in our case). But it's going to have more of an impact on a CPU that hasn't been overclocked, and can drop its frequency lower at idle.

The “Performance Bias” option found in the BIOS adds even more performance to Geekbench, even set to the default Auto mode. By changing it from Auto to Aida/Geekbench, we record gains in each of the five values tested. Certain motherboard reviewers may see this as cheating, but it'll be a blessing during overclocking competitions.

AMD Ryzen 7 1700XView Deal

Overclocking BCLK Frequency & PCIe

Software: Asus TurboV versus Ryzen Master

Even though we don't think that Asus' TurboV Core software is practical for daily usage, it does facilitate BCLK frequency adjustments from Windows (though it's called APU Frequency for some reason). To its credit, TurboV Core is very comprehensive, allowing you to modify a number of different voltages, along with multiplier coefficients, on the fly.

We did find several bugs while using TurboV Core. During testing, we fiddled with various voltage settings. We'd make our adjustments via Windows and observe their effect on temperatures. Suddenly, after setting the core voltage to 1.2V, the temperature shot from 50°C to 120°C in one-tenth of a second. Cinebench R15 didn't even have a chance to display its first pixel. We though we had applied a bad setting. Several minutes later, we were at +114°C again, just before another crash (at least the built-in protection features work). Finally, as we were setting the voltage to 1.1V, the temperature spiked at 98°C. This time the OS didn't crash. So we grabbed our voltmeter and carefully read the supply voltage.

The result? Nothing died, but we discovered a 1.1V software setting was really 1.7V on the board. We have no idea what voltage was applied when the temperature offsets were showing +120°C. But if 1.7V pushed our chip to "only" 98°C after a few seconds, then the voltage needed to immediately hit 120°C must have been close to 2V. These Ryzens must be pretty robust.

The bug was reported to Asus, and a representative told us it'd be corrected in the next version of TurboV Core. Just before publishing, we tried the latest version of TurboV Core and discovered that, instead of dialing in an incorrect voltage, the application now crashes instead.

Bugs aside, we prefer Asus' software to AMD's Ryzen Master utility. The latter can be difficult to use, it exposes fewer settings, and our performance numbers simply tumble when it's open.

Overclocking BCLK Frequency

Overclocking Ryzen isn't that different from other platforms. There is a reference clock that Asus calls the BLCK Frequency, and it's more or less equivalent to the BCLK you already know from Intel's CPUs.

It's particularly important in this case because the setting is linked to several subsystems: the processor and memory clock rates, as well as the USB ports, the PCIe bus, and the SATA interfaces. Increasing it effectively overclocks almost everything on your motherboard.

Raising the BCLK Frequency setting can create stability problems with each interface. Asus therefore suggests using the interfaces connected directly to the processor, which hold up to overclocking well. Asus even recommends using M.2 storage when you're going for an aggressive overclock, since it's attached to the CPU via PCIe.

Modifying BCLK Frequency

Many motherboards do not allow the BCLK Frequency to be modified, locking it down to 100 MHz. But certain manufacturers include an external clock generator on their high-end platforms. If you plan on exploring BCLK Frequency, make sure the board you pick does indeed feature an external clock generator. At the time of this writing, only three or four models do.

On our configuration, we achieved a stable BCLK Frequency of 148 MHz with no USB- or SATA-related issues. To prevent the reference clock frequency from rising too high, it is necessary to lower the CPU and memory multipliers. The PCIe bus does have a hard time with this frequency increase, though.

Bug: To go beyond 110 MHz on our motherboard, we were forced to leave the motherboard's SMT option on “Auto.” SMT is active in that state, but when we set the option to “Activated” instead, it prevented us from booting above 110 MHz.

Back on the subject of BCLK Frequency, here is an example of the progression trends we achieved in Windows. Of course, you aren't going to boot at 100 MHz and then switch to 150 MHz from the operating system. But there is a small amount of headroom, though slippery and inconsistent. In the table below, we indicate the frequency at boot, followed by the maximum in Windows before locking up.

Boot	OS
100 MHz	104 MHz
110 MHz	115 MHz
120 MHz	127 MHz
130 MHz	138 MHz
140 MHz	148 MHz
148 MHz	154 MHz

PCI Express Management

The Socket AM4 platform manages the third-gen PCIe. When you increase the BCLK Frequency beyond a certain point, though, the motherboard rolls back to PCIe 2.0. The higher the frequency you select, the higher the offset frequency increase. Thinking that this was tied to the PCIe bus, we tried to impose first-gen transfer rates, but saw no change.

BCLK Frequency Interval	PCI Express Bandwidth	Real Bandwidth Per Lane
85 to 104.8 MHz	Gen 3, 8 GT/s (985 MB/s)	837 to 1032 MB/s
105 to 144.8 MHz	Gen 2, 5 GT/s (500 MB/s)	525 to 724 MB/s
145 MHz+	Gen 1, 2.5 GT/s (250 MB/s)	313 MB/s+

Graphics performance isn't significantly affected by these changes, at least not in any of the benchmarks we ran with our GeForce GTX 1080. As such, we suggest using either 104.8 or 144.8 MHz. If you want, you can manually lock in the PCIe generation you want through Asus' BIOS. Be careful, though: too high of a BCLK Frequency setting with a transfer rate that doesn't adapt well risks introducing instability on the PCIe bus.

AMD Ryzen 7 1800XView Deal

Temperatures & PLL Impact

As you may know, Ryzen processors contain numerous sensors that allow the frequency and voltages to be driven as a function of CPU temperature. It is therefore important to study this relationship. Once the chip's temperature exceeds 95°C, performance is throttled.

Throughout the course of our benchmarks, we were surprised to see high temperatures, especially at idle, on our 1800X. We were even further astonished when we started trials with the 1700. Since then, AMD has published a community update that explains offsets in Ryzen temperature reporting. To figure out the “real” temperature on the R7 1800X and R7 1700X, you must apply an offset of -20°C. For the 1700, no correction is necessary.

Also, we observed that this offset of 20°C is a loose approximation, and that it can change as a function of the voltage and load applied to the processor. As a result, our measurements and temperatures for the 1700X and 1800X are not assuredly precise.

PLL Influence

As a reminder, we're using be quiet!'s Silent Loop 280 liquid cooling kit. The processor was tested at a range of voltages, and each time an appropriate frequency was applied. Logically, the higher the voltage, the higher the temperature.

Settings (V)	Frequency (MHz)	PLL Voltage @1.8V Temperature (°C)	PLL Voltage @1.9V Temperature (°C)	PLL Voltage @1.8V Corrected Temp. (°C)	PLL Voltage @1.9V Corrected Temp. (°C)
1.0	3450	46	55	26	35
1.1	3700	52	60	32	40
1.2	3850	56	65	36	45
1.3	4000	63	71	43	51
1.4	4100	69	78	49	58

Our table shows the temperature we measured at the default PLL voltage of 1.8V, but also 1.9V. Why? Simply because the motherboard changes this parameter on its own any time you raise the core frequency. Therefore, we ended up with incoherent data. After a deeper analysis, we realized that this voltage varied, and that it had a significant influence on temperature. So we redid our tests while setting the PLL Voltage to a fixed value of 1.8V, which posed no stability problems.

It does not seem useful to increase this voltage. We did not see any frequency gains, while the temperature rose 8°C on average. We recommend that you manually fix the PLL Voltage to 1.8V, or even less if your processor remains stable. The last two columns show the corrected temperatures (with the -20°C offset applied).

We plotted curves with the measured data points, and the temperature increase is practically linear. The straight lines show the corrected temperatures with -20°C offset applied. An interesting observation, by fixing the PLL Voltage at 1.8V we can add an additional 0.15V to the Vcore without even hitting the temperatures measured with the PLL set to 1.9V.

AMD Ryzen 7 1700View Deal

Overclocking Ryzen 7 1700 vs 1700X vs 1800X

Overclocking at Equal Voltages

Knowing that all of these processors are identical and come from the same manufacturing line, we have to ask whether there's any real reason to splurge on the highest-end model. To that end, we compared their base frequencies and maximum overclock with all eight cores under load. The memory was set to 3200 MT/s, and all of the other BIOS settings were left at their defaults.

Model	Frequency (MHz)	Voltage (V)	Temperature (°C)
Ryzen 7 1700	3200	1.07	35
Ryzen 7 1700X	3500	1.16	52 / 32
Ryzen 7 1800X	3700	1.23	58 / 38

The "entry-level" Ryzen 7 1700 seems to have an advantage, given a base core voltage of 1.07V (compared to the 1.16 and 1.23 volts measured on the pricier chips). Temperature is affected by the Vcore of course, but our comparison is made imprecise by AMD's offsets.

We applied the recommended -20°C offset, but again, we're reminded that the correction isn't perfect. The discrepancy between the reported and real values isn't constant. It changes as a function of multiple parameters. Applying this offset gets us closer to the right temperature, but it isn't guaranteed to be accurate.

Model	Frequency (MHz)	Voltage (V)	Temperature (°C)
Ryzen 7 1700 OC	3975	1.35	45
Ryzen 7 1700X OC	3950	1.35	63 / 43
Ryzen 7 1800X OC	4050	1.35	66 / 46

While the 1800X has a clear base advantage, the difference diminishes after overclocking. If tuning your CPU doesn't scare you, we recommend going for the Ryzen 7 1700 in light of its more attractive price.

Performance

The Ryzen 7 1700 serves as our baseline. Operating at just 3200 MHz (across eight cores), it obtains a score of 1438 points in Cinebench R15. That's a long way from the Ryzen 7 1800X's 1640 points. Given the 1800X's 500 MHz advantage, though, we don't have a difficult time explaining the 14% performance difference.

Overclocked, that delta disappears. Our Ryzen 7 1700 even beats the 1700X. Of course, you can expect overclocking results to vary from one CPU to another, as we mentioned in our Kaby Lake overclocking test.

The 1800X maintains a less-than 2% lead over AMD's Ryzen 7 1700. It is quite astonishing that the 1700X achieves a lower score; this cannot be explained simply by a lower frequency. Even at the same clock rate, our 1700X sample consistently fares worse than the 1700.

MORE: How to Overclock a CPU

AMD Ryzen 7 1700XView Deal

1800X: Maximum Overclock & Scaling

With the 1800X as our best overclocking candidate, we shift our focus to it for the rest of our experiments.

This section explores how the processor behaves at various core voltages. We begin with an under-volting scenario and finish with overclocking. In the interest of time, we're using a fairly brief test (Cinebench R15), so the reported values probably aren't their stable ceilings. However, under a heavier load, the observed progression should be similar.

Voltage (V)	Frequency (MHz)	Score (pt)	Temperature (°C)	Frequency %	Score %
1.0	3450	1540	46	0.0	0.0
1.1	3700	1642	52	7.2	6.6
1.2	3850	1710	56	11.6	11.0
1.3	4000	1770	63	15.9	14.9
1.4	4100	1822	69	18.8	18.3
1.4 SMT Off	4175	1318	62	21.0	-14.4

With a core voltage of 1.0V, we had to lower the frequency to 3450 MHz. This serves as our comparison's starting point.
Moving to 1.1V, the frequency jumps 250 MHz. That's a progression of 7.2% for a performance gain of 6.6%.
At 1.2V, we are slightly under the default core voltage. The frequency can be increased to 3850 MHz. The overclocking headroom, with respect to XFR frequency, is not high on the 1800X.
With 1.3V we reach the 4 GHz.
Our last increase is to 4100 MHz at 1.4V.
Since there are certain applications that do not use SMT, we wanted to see if deactivating the feature would allow us to push clock rate higher. Without SMT, we picked up an additional 75 MHz. The temperature also benefited: it dropped 7°C.

As is true with most components, a small increase in voltage allows a clear gain in frequency. But the magnitude of this gain lessens as Vcore goes up. Transitioning from 1.0 to 1.1V allows us to boost clock rate by 250 MHz, but going from 1.3 to 1.4V only yields 100 MHz.

Given these conditions, it is difficult to recommend a core voltage in excess of 1.3 to 1.35V for daily usage. For benchmarking, the voltage can be pushed to 1.4V without much risk.

Performance in Cinebench is heavily dependent on clock rate, so we naturally see a strong correlation between frequency and score. But it's also a threaded benchmark. While deactivating SMT (and thus eight of the 16 logical cores) helps us hit a higher clock rate, it also heavily penalizes our results.

AMD Ryzen 7 1800XView Deal

Memory Performance

Overclocking the execution cores is limited by available headroom. But if you really want to improve this platform's effectiveness, you cannot overlook the memory bus.

We begin by illustrating the gains achievable by moving from DDR4-2400 to -3200. On our Crosshair VI Hero, the 2400 MT/s is actually the default. On other motherboards, it may be lower (and not without consequence).

Cinebench R15

This test is not sensitive to memory bandwidth or latency, so the gains we measure should be trivial. Still, the influence of RAM performance is quantifiable, given a ~1.4% improvement.

Configuration	Score
2400	1639
3200	1663

Geekbench 4

This test stresses both the host processor and memory subsystem. Its single- and multi-core metrics reveal a performance increase of 5 and 6% from dialing in a faster data rate. That's quite a gain from simply modifying a multiplier setting, and memory capable of supporting 3200 MT/s isn't extremely expensive. Again, our Geekbench 4 scores are an average of three consecutive runs.

Configuration	Single-Core	Multi-Core	Mem. Copy	Mem. Latency	Mem. Bandwidth
2400	4417	20786	6229	4697	5568
3200	4635	22150	7546	5651	7094

Time Spy

Graphics-bound workloads don't benefit as much from a memory bandwidth increase. We do measure a gain in the overall score, but it's very small. Conversely, the CPU-oriented benchmark jumps by 343 points.

Configuration	Graphics	CPU	Score
2400	7204	8010	7314
3200	7217	8353	7367

AMD Ryzen 7 1700View Deal

BCLK Frequency

With our G.Skill Flare X memory kit, we have five preset RAM profiles through the Crosshair VI Hero's D.O.C.P. (Direct Over Clock Profile) menu. This is an Asus/AMD implementation analogous to Intel's XMP.

Each profile uses a final data rate of 3200 MT/s and a principal timing setting of 14, but with different BCLK Frequency values to get there. The bottom section of our screen capture explains each mode:

D.O.C.P. 1: This is purportedly the system's optimal setting, which offers the best compatibility by trading off performance.
D.O.C.P. 2: A 2133 DRAM ratio and 150 BCLK Frequency are used to arrive at a 3200 MT/s data rate.
D.O.C.P. 3: 2400 DRAM ratio and 133 MHz BCLK Frequency.
D.O.C.P. 4: 2666 DRAM ratio and 120 MHz BCLK Frequency.
D.O.C.P. 5: 2933 DRAM ratio and 109 MHz BCLK Frequency.

Why go through all of this trouble? Well, AMD locked a number of RAM timings in the BIOS, so they are not accessible to the user. That is to say they aren't directly accessible.

Here is a copy of the timings we achieved by booting the system with different memory ratios, highlighted with a yellow box in each of the six screenshots (you can enlarge the image above by clicking on it to see the details). The principal timings are marked in blue. These change as a function of the memory ratio, but can be manually forced in the BIOS.

Other timing components are shown in green, and these are not accessible. If you look carefully, you can see that the higher the RAM ratio, the higher these values appear. Higher timings reduce performance, so we have every interest in keeping these as low as possible.

To ensure that you understand what we are talking about, here are two examples where the RAM is set to 3200 MT/s:

2133 ratio and 150 MHz BCLK Frequency -> RAM frequency is 150 MHz x 21.33 = 3200 MT/s, tRC is 51, tRTP is 8, tFAW is 23
3200 ratio and 100 MHz BCLK Frequency -> RAM frequency is 100 MHz x 32.00 = 3200 MT/s, tRC is 75, tRTP is 12, tFAW is 34.

It goes without saying that the first configuration should be faster. Let's see how this translates to actual performance. The reference clock is modified, then we manually adjust the multiplier coefficient of the processor so that its final frequency is as close as possible to 4050 MHz. We didn't run into any stability issues during these changes.

Cinebench R15

We begin again with Cinebench R15. Still just as insensitive to memory throughput/latency, the gain between D.O.C.P. 5 and 2 is small. While 10 points can be a mountain for a competitive overclocker, it isn't that interesting for a regular user. By manually changing the timing settings to 12-12-12-12 instead of 14-14-14-14 we gain another five points.

Configuration	Score
3200 D.O.C.P.5 - BCLK Freq. 109	1803
3200 D.O.C.P.3 - BCLK Freq. 133	1810
3200 D.O.C.P.2 - BCLK Freq. 150	1812
3200 C12 - BCLK Freq. 150	1817

Geekbench 4

Single-core performance tests are usually less sensitive to RAM, and this one reflects a 2% speed-up from D.O.C.P. 5 to 2. Tightening up the timings yields an additional 1% gain.

We do 4% better in the multi-core test. Put another way, by assuming that the score progression is perfectly linear, this 4% increase would be the equivalent of a 200 MHz clock rate boost.

Configuration	Single-Core	Multi-Core	Mem. Copy	Mem. Latency	Mem. Bandwidth
3200 D.O.C.P.5 - BCLK Freq. 109	4649	23640	7730	5677	7062
3200 D.O.C.P.3 - BCLK Freq. 133	4726	24397	8845	5899	7456
3200 D.O.C.P.2 - BCLK Freq. 150	4741	24603	8505	5926	7470
3200 c12 - BCLK Freq. 150	4786	24634	8706	6106	7456

If we take into account the 6% gain from increasing the RAM from 2400 MT/s to 3200, we have an overall increase of 10%. That's equivalent to almost 500 MHz. Not bad!

The results of our mem copy test are difficult to reproduce. The trend is visible, but hundreds of points separate the best runs from the worst. Take those results with a grain of salt, even if the values we provide are an average of three passes. It is clear in this test that there is a a visible gain, with a 12% difference between the best and the worst profile. The two remaining columns show gains between 5 and 7%.

Time Spy

The graphics-bound benchmark didn't seem to appreciate a higher memory data rate. Let's see if tighter timings can improve this: the faster the memory, the more the score drops!

At barely 0.5%, the differences are small. This represents barely 0.2 frames per second lower in a test that averages 30 FPS. These results are quite surprising, particularly since the CPU score soars at the same time.

Configuration	Graphics	CPU	Score
3200 D.O.C.P.5 - BCLK Freq. 109	7198	8893	7410
3200 D.O.C.P.3 - BCLK Freq. 133	7180	9291	7433
3200 D.O.C.P.2 - BCLK Freq. 150	7177	9336	7434
3200 c12 - BCLK Freq. 150	7161	9485	7427

Ashes of the Singularity

We made one last comparison using the Ashes of the Singularity benchmark. Our settings included the Crazy preset, and we chose the CPU-oriented metric. The test was run before this game was patched to optimize for Ryzen, but in relative terms, its utility is the same.

Compared to a data rate of 2133 MT/s, pushing the memory to 2400 MT/s increases our score by 4.1%. The frequency is higher, yes, but timings are also more relaxed, so the two effects partially cancel one another.

Changing to 3200 MT/s jumps us up 13%. This test is therefore very sensitive to memory bandwidth, and we suspect that lots of data must be transferred from one CCX to another. With the interface between CCXes affected by memory frequency, it seems natural that this bottleneck would clear by increasing RAM performance.

By applying the technique we're showing, optimizing memory timings with a higher data rate (achieved by increasing the BCLK Frequency setting), our scores jump by no less than 31%.

Ryzen 7 1800X Configuration	Within-Core Latency Range	Within-CCX Core-to-Core Latency Range	Cross-CCX Core-to-Core Latency Range	Cross-CCX Average Latency	Bandwidth	Std. Deviation
1333 MT/s	14.6 - 14.8ns	39.5 - 41.7ns	230 - 243.4ns	237.65ns	43.74 GB/s	2.84ns
2666 MT/s	14.6 - 14.8ns	39.6 - 42.3ns	117.8 - 124.6ns	120.4ns	50.16 GB/s	1.86ns
3200 D.O.C.P. Standard - BCLK Freq. 100	14.6 - 14.8ns	40.1 - 42.1ns	108 - 114.6ns	114.66ns	52.02 GB/s	1.69ns
3200 D.O.C.P.3 - BCLK Freq. 133	14.6 - 14.9ns	39.4 - 42.0ns	108.4 - 112.4ns	111.51ns	55.24 GB/s	0.90ns

Maximum RAM Frequency

Having heard many times that Ryzen doesn't have a lot of headroom for memory overclocking, we wanted to find out for ourselves. After all, achieving DDR4-3200 at CAS12 wasn't difficult. What comes after, though?

In all honesty, it's pretty easy to hit 3400 MT/s. You select a setting in the BIOS and the motherboard boots up. At least, that's the case with our Crosshair VI Hero. Early on in this platform's life, other models seemed to struggle more. Fortunately, the situation does seem to be improving.

Once you get to 3400 MT/s, overclocking gets more complicated. The results seem to vary from one processor to another and are strongly dependent on your motherboard. Between 3400 and 3650 MT/s, we saw crash after crash. It was impossible to get the system to boot. There was, however, a very small “magic zone” at 3800 MT/s where the system was usable. The timing defaulted to CAS14, but manually specifying CAS12 didn't pose a problem.

Of course, we tried to run our tests at this data rate, but the system was too unstable. After three hours of trying to get a result from Cinebench, we threw in the towel.

The take-away is that Ryzen can hit these frequencies. It's difficult today due to platform immaturity. Asus' board is one of the better ones out there, but it's not perfect. In the future, after additional BIOS updates, it's possible that everyone will gain access to higher data rates and the additional performance they enable.

AMD Ryzen 7 1700XView Deal

Conclusion

The many hours we put into overclocking Ryzen and Asus' Crosshair VI Hero were made more difficult than we're used to due to the many bugs in AMD's nascent Socket AM4 platform.

Of course, it is never easy to start with a new platform in-hand, but when every single component is unfamiliar, and complicated by the fact that AMD seems to have rushed its launch without disseminating information to everyone who needed it, don't even bother counting the hours you spend trying to explore various settings for the first time.

The target is constantly moving, too. Regular BIOS updates intended to smash show-stopping bugs often require trashing old results and starting over from scratch. We didn't think we'd ever finish. And indeed, many updates have happened since this piece was published on Tom's Hardware's French site.

At a certain point, though, with all of the data in front of us, overclocking Ryzen becomes child's play. Increase a multiplier, fiddle with the memory data rate; it's naturally pretty intuitive.

The BCLK Frequency & RAM Surprise Attack!

G.Skill Flare X (2x 8GB)View Deal

If you want to add 4 to 5% more performance, you must have a motherboard that lets you modify the reference clock, such as our Crosshair VI Hero. You'll want to pore over our data and try replicating some of our experiments using the options and settings we dialed in.

For tinkerers, Ryzen 7 1700 is the eight-core model we suggest. Once these CPUs are pushed to their limit, the final difference between them is negligible.

Be sure to configure your memory with relation to the BCLK Frequency setting, which can significantly improve gaming performance. Don't hesitate to spend extra on high-performance memory. G.Skill's Flare X kit proved to be very practical thanks to Asus' D.O.C.P profiles.

The Ryzen platform, in all of its newness, still suffers from several bugs. Watch carefully for BIOS updates from your motherboard vendor. Little by little, these will correct problem spots, improving stability and increasing performance.

MORE: Best CPUs for Gaming
MORE: CPU Benchmark Hierarchy
MORE: How to Overclock a CPU

AMD Ryzen 7 1800XView Deal

AMD Ryzen 7 1700 CPU Review

palcorn@outlook.com (Paul Alcorn) — Thu, 06 Apr 2017 13:00:00 +0000

Introduction

AMD is trying to claw back lost market share with its eight-core Ryzen processors, and in the process, the company is generating a tremendous amount of excitement. Most of the enthusiasm stems from competitive pricing and solid performance in content creation and productivity workloads, even if Ryzen isn't shaping up to be universally superior, as many hoped prior to launch. The chips still suffer an IPC deficit compared to Intel’s Kaby Lake processors, and the unique Zen microarchitecture falls behind in some games at lower resolutions. While the Ryzen 7 CPUs we've tested provide a smooth experience in most titles, they don't oust Intel from its comfortable position atop the benchmark charts. This makes it difficult to universally recommend those high-end parts.

But it appears the bottom of AMD's Ryzen 7 stack offers the best value. The company claims that its Ryzen 7 1700 is the most efficient eight-core CPU available. And priced at $330, it's undoubtedly the cheapest one with modern amenities. The 1700 wields the same design as its more expensive counterparts, including the same Zen-based architecture, two CCXes enabling 16 logical cores, and 16MB of L3 cache. It also sports an unlocked ratio multiplier, AMD's SenseMI suite, and Socket AM4 compatibility.

As a result, the Ryzen 7 1700 might hit a value sweet spot that the $500 Ryzen 7 1800X and $400 1700X couldn’t.

AMD Ryzen 7 1800X

Ryzen 7 1700X

Ryzen 7 1700

The 1700’s 65W TDP stands out as its most notable differentiator compared to the higher-end 95W Ryzen 7s. A more conservative power rating means lower voltages (and heat), so its 3 GHz base and 3.7 GHz boost frequencies understandably trail the 1700X and 1800X as well.

Both X SKUs do benefit from AMD's XFR (eXtended Frequency Range) technology, which provides an additional 100 MHz over the boost ceiling if your thermal solution is beefy enough. In contrast, the 1700 comes equipped with a reduced XFR feature set that doesn’t boost beyond 3.7 GHz, though it does facilitate an all-core 3.1 GHz boost in threaded workloads. You also save a few bucks with the bundled 95W Wraith Spire cooler, and although we wouldn’t recommend using the stock heat sink for overclocking, it’s a nice addition.

The 1700 wades into a brutally competitive segment; its $330 price tag is only slightly lower than the $350 Core i7-7700K rocking a 4.2 GHz base and 4.5 GHz Turbo Boost clock rate. The Ryzen 7 1700 beckons with twice as many cores and double the L3 cache, though that doesn't translate to more performance in every application. After all, even Intel's Core i5-7600K competes with the top Ryzen 7s in certain workloads.

The Ryzens we've tested thus far suffer curious performance losses in some games. But AMD contends that many developers will patch their software to improve frame rates. Encouragingly, Stardock/Oxide recently patched Ashes of the Singularity: Escalation to optimize for Ryzen. Valve also released a patch for Dota 2. Both updates improve performance, and more important, they hint at what we might see from other devs in the future.

In the meantime, Ryzen 7 1700 offers a great starting price for eight cores and an unlocked multiplier. We think it can match its bigger brothers with a bit of tuning. Let's test that theory out.

MORE: Best CPUs

MORE: Everything Zen: AMD Presents New Microarchitecture At HotChips

MORE: Intel Kaby Lake Core i7-7700K, i7-7700, i5-7600K, i5-7600 Review

MORE: Broadwell-E: Intel Core i7-6950X, 6900K, 6850K & 6800K Review

Overclocking & Test Setup

Overclocking

We dialed in a 3.9 GHz overclock for all three Ryzen processors, and the tuning experience was similar across the trio. A 1.3875V core voltage and 1.2V CPU SoC voltage paved the way for extended stability in Prime95. During our tests in the U.S. lab, we saw the 1700 peak at 65°C (according to AIDA), while the 1700X and 1800X reached 70°C.

Waste heat appears to accumulate quickly at higher clock rates (we hit 82°C with the 1800X at 4 GHz using Corsair's H100i v2). The Wraith Spire cooler can dissipate up to 95W, but it’s best to purchase a beefy cooler if you plan on pushing the limits of Ryzen's frequency headroom. Regardless of the heat sink we used, or the unsafe voltages we dialed in, though, we couldn't run at the same 4 GHz with our 1700. Our sample just wouldn't make it past our stress tests. Your mileage may vary, of course.

AMD recommends a "safe" limit of 1.35V for long-term overclocks, but doesn't provide warranty coverage for voltage settings beyond default. The company also claims Ryzen can take up to 1.45V, though it doesn't recommend this for long-term use.

Ryzen Memory Support	MT/s
Dual-Channel/Dual-Rank/Four-DIMM	1866
Dual-Channel/Single-Rank/Four-DIMM	2133
Dual-Channel/Dual-Rank/Two-DIMM	2400
Dual-Channel/Single-Rank/Two-DIMM	2677

We bumped Corsair's LPX memory modules up to DDR4-2933 for our overclocked tests. There are reports that higher data rates yield notable gains from Ryzen processors. Unfortunately though, even after the most recent wave of BIOS updates, support for 3000 MT/s+ is dicey. The motherboard ecosystem continues to mature, and we plan to follow up when AMD opens up additional timings and improves memory overclocking with an update later in April.

Test Setup

For this piece, we split testing between our German and American labs. The team in the U.S. ran the gaming benchmarks, while the Germans measured performance in HPC and workstation apps, and then collected thermal/power data. They used MSI's X370 XPower Gaming Titanium motherboard and we went with Asus' Crosshair VI Hero and an EVGA GeForce GTX 1080 FE. We stuck with AMD’s recommended presets for our stock configurations to minimize issues attributable to the dissimilar boards.

If you want to know more about how the Tom's Hardware DE system looks and is controlled, check out How We Test Graphics Cards.

Test Systems and Measurement Setups
Systems	Germany AMD 1Ryzen 7 1800X, 1700X, 1700MSI X370 XPower Gaming Titanium2x Corsair Vengeance LPX DDR4-3000 @2666 MT/sIntel LGA 2011-v3 Intel Core i7-6900K MSI X99S XPower Gaming Titanium 4 x 4GB Crucial Ballistix DDR4-2400Intel LGA 1151 Intel Core i7-7700K MSI Z270 Gaming 7 2 x 8GB Corsair Vengeance DDR4-3200 @2400 MT/sAMD Socket AM3+ FX-9590 Asus Crosshair V Formula 2 x 8GB Corsair Dominator DDR3-2133 @1866 MT/sGermany All1x 1TB Toshiba OCZ RD400 (M.2, System SSD)2x 960GB Toshiba OCZ TR150 (Storage, Images)be quiet Dark Power Pro 11, 850WWindows 10 Pro (All Updates)US AMD 1Ryzen 7 1800X, 1700X, 1700Asus ROG Crosshair VI Hero2x Corsair Vengeance LPX DDR4-3000 @2666 MT/sUSA Intel 1Intel Core i7-7700KMSI Z270 Gaming M72x Corsair Vengeance LPX DDR4-3000 @2400 MT/sUSA Intel 2Core i7-6900KASRock X99 Extreme44x Crucial DDR4-2400US All1TB Samsung PM863SilverStone ST1500, 1500WWindows 10 Pro (All Updates) Version 1607
Cooling	Germany- Alphacool Eispumpe VPP755 Pump - Alphacool NexXxoS UT60 Full Copper 240mm- Alphacool Eisblock XPX CPU-Alphacool Cape Corp Coolplex Pro 10 LT- 5x be quiet! Silent Wings 3 PWM- Thermal Grizzly Kryonaut US-Corsair H100iv2-Noctua NH-U12S SE-AM4-Arctic MX-4
Case	Lian Li PC-T70 with Expansion Kit and Mods
Power Consumption Measurements	- Contact-free DC Measurement at PCIe Slot (Using a Riser Card) - Contact-free DC Measurement at External Auxiliary Power Supply Cable - Direct Voltage Measurement at Power Supply- 2 x Rohde & Schwarz HMO 3054, 500MHz Digital Multi-Channel Oscilloscope with Storage Function - 4 x Rohde & Schwarz HZO50 Current Probe (1mA - 30A, 100kHz, DC) - 4 x Rohde & Schwarz HZ355 (10:1 Probes, 500MHz) - 1 x Rohde & Schwarz HMC 8012 Digital Multimeter with Storage Function
Thermal Measurements	- 1 x Optris PI640 80Hz Infrared Camera- PI Connect Analysis Software with Profiles
Noise Measurements	- NTI Audio M2211 (with Calibration File)- Steinberg UR12 (with Phantom Power for Microphones)- Creative X7, Smaart v.7- Custom-Made Proprietary Measurement Chamber, 3.5 x 1.8 x 2.2m (L x D x H)- Perpendicular to Center of Noise Source(s), Measurement Distance of 50cm- Noise Level in dB(A) (Slow), Real-time Frequency Analyzer (RTA) - Graphical Frequency Spectrum of Noise

3DMark, Ashes of the Singularity, Battlefield 1 & 4

3DMark

Synthetic benchmarks usually don't provide an accurate measure of real-world gaming performance, but 3DMark's DX11 physics and DX12 CPU tests provide useful insight into the raw horsepower available to the game engine.

The DX11 and DX12 CPU tests reveal a similar trend. Ryzen 7 1700 trails the other eight-core processors, but passes the Core i7-6900K once we overclock it. Of course, the same treatment applied to Intel's chip would have a similar effect, so we'll stop short of calling this an AMD victory.

The quad-core Core i7-7700K and i5-7600K are far less potent in the threaded Time Spy and Fire Strike tests. However, they leverage their superior IPC throughput and clock rate to turn in a solid showing during the single-thread DX11 API overhead metrics. Intel's four-core CPUs are also competitive against the Ryzen processors in multi-threaded DX11 draw call performance, which helps explain their commanding lead in most gaming benchmarks.

DX12 leverages additional cores more efficiently, which turns the tables in favor of the eight-core chips. AMD's 1700 makes a solid showing, though it naturally trails the quicker 1700X and 1800X at stock settings. The overclocked 1700 pulls closer to its companions, but all overclocked Ryzen 7 processors suffer curiously reduced DX11 multi-threaded performance compared to stock settings. Through it all, the stock Core i7-6900K stubbornly refuses to budge from its commanding lead in DX11 threaded and DX12 tests.

Ashes of the Singularity: Escalation

Lackluster results in Ashes of the Singularity was a notable thorn in Ryzen's side, which surprised us given the purportedly close collaboration between AMD and Stardock/Oxide. Before the Ryzen launch, AMD stated that Ashes of the Singularity's Nitrous game engine wasn't optimized to fully leverage Ryzen's unique cache topology and SMT implementation, but that a patch was inbound. We gained access to the pre-release update and recorded significantly better performance from Ryzen 7. We saw the Core i7-6900K speed up too, though.

The stock Ryzen 7 1700 and 1700X can't beat a stock Core i7-7700K, but a little overclocking puts them in the lead. Unsurprisingly, though, the Core i7-6900K asserts its dominance with a large lead over the other processors.

Battlefield 1

We dialed Battlefield 1 up to the Ultra preset and repetitively took an armor-laden stroll across the O La Vittoria landscape.

We run into a graphics bottleneck during the test, so we don't observe much variation between contenders. They all enable smooth game play through our test sequence. This is why AMD says Ryzen is great for high-res gaming!

Battlefield 4

Battlefield 4 also leans on the GPU more than the CPU, so again, we notice little variation at the upper end of the 1920x1080 benchmark results.

The Ryzen 7 1700 and 1700X trail the other contenders, but pull up to the faster processors after overclocking. In many ways, these results mirror the common trend we see at higher resolutions. All of the processors offer acceptable performance when you encounter a graphics bottleneck.

Civilization VI AI & Graphics Test, Deus Ex: Mankind Divided, GTA V

Civilization VI AI Test

Civilization VI's AI benchmark measures the amount of computational horsepower available to the system during a turn-based strategy gaming session. The AI benchmark is very consistent, so it was strange when we noticed the Ryzen 7 1700 outpacing the rest of the field. Alas, the remarkable performance came from a Civilization VI Spring 2017 update, which dramatically improved AI performance, forcing us to toss our old numbers.

We retested all of the processors and noted higher performance across the board, so it appears this wasn't an AMD-targeted patch. The finishing order remains the same, reflecting a ~3-second improvement for all CPUs. The stock Ryzen 7s can't keep up with Core i7-6900K, though they do jump in front after overclocking.

The Core i7-7700K and i5-7600K continue to dominate this test. We did log CPU utilization during our run and saw the benchmark scale across all 16 logical cores (it spawns 45 threads). But scaling and scaling well are two different matters, so the four-core processors continue enjoying a lead.

Civilization VI Graphics Test

The update had little impact on our graphics test, though we did notice a slight decline in average frame rates for most processors, along with more consistent minimum frame rate measurements.

The results line up familiarly, though the Core i5-7600K isn't as competitive as it is in other titles. Meanwhile, Intel's Core i7-6900K leads, while the company's Core i7-7700K falls into its wake. The stock Ryzen 1700 isn't competitive and trails its siblings by a large margin. An overclock helps put the chip within range of the two faster SKUs.

Deus Ex: Mankind Divided

Deus Ex: Mankind Divided remains the only game that Ryzen processors dominate in convincing fashion. That trend continues as we look at Ryzen 7 1700's performance. The CPU secures its spot in the leading group at stock settings. And due to an apparent graphics bottleneck, it experiences minimal uplift from our overclocking efforts.

Grand Theft Auto V

We measure performance during Grand Theft Auto V's F-16 flight sequence with the built-in benchmark. The scripted path yields a solid and consistent benchmark, while constantly changing terrain applies a significant graphics workload.

The stock Ryzen 7 1700 shows up at the bottom of our performance chart. But once again, it climbs into a more competitive position after we overclock. Unfortunately for AMD, all of the Ryzen processors trail their Intel competition by a significant margin. Even the affordable Core i5-7600K at stock clock rates outperforms. And overclocking would certainly propel it into a much higher position.

Hitman (2016), Metro: Last Light Redux, Middle-earth: Shadow of Mordor

Hitman (2016)

Hitman prefers high clock rates and additional logical cores, so the Core i7s lead while the Core i5-7600K tumbles to the bottom.

We recorded moderate improvements from the overclocked Ryzen 7s, but they still aren't competitive with Intel's Core i7s. A clear delineation between the processor families, apparent in our average frame rate over time chart, suggests that the game could benefit from Ryzen-specific optimizations.

Metro: Last Light Redux

We encounter a graphics bottleneck during our Metro: Last Light Redux tests. The stock 1700 only loses by 1.5 FPS compared to the Ryzen 7 1800X, but trails the Core i7-7700K's average frame rate by 4.1.

Middle-earth: Shadow of Mordor

Middle-earth: Shadow of Mordor also reflects a GPU-imposed ceiling at the top end of the scale. A stock Ryzen 1700 doesn't have the clock rate to keep up, so it trails the leader by an average of 8.9 FPS. Overclocking helps put the 1700 into a more competitive position, though.

Project CARS, Rise of the Tomb Raider, The Division

Project CARS

Project CARS is a CPU-intensive title that promotes parallelism by breaking tasks into smaller chunks and spreading them among the cores. Intel's quad-core models take a commanding lead during our benchmark run, though, indicating a particular sensitivity to clock rate.

The Ryzen processors aren't as fast as Intel's Core i7-6900K, even after overclocking. We measure a 27.4 average FPS gulf between the Ryzen 7 1700 and Core i7-7700K. Although this gap narrows to 17.5 FPS after overclocking the 1700, we expect the -7700K would also benefit greatly from a bit of tuning.

Rise of the Tomb Raider

Rise of the Tomb Raider is another sore spot for AMD's architecture, as evidenced by a 55.9 FPS gap between the Core i7-6900K and stock Ryzen 7 1700. The -6900K's solid performance points to respectable scaling based on host processing resources. But even after our overclocking efforts, the eight-core Ryzen 7s still trail Intel's chips by more than 30 FPS. The 1700, for its part, averages 24.2 FPS higher after overclocking to 3.9 GHz.

The Division

We spot little variation between the overclocked Ryzen processors as they slip by Intel's Core i7-6900K.

Workstation & HPC Benchmarks

2D Benchmarks: DirectX And GDI/GDI+

Even though it sports considerably lower base and boost frequencies (and doesn't benefit from XFR), the Ryzen 7 1700 stays within striking distance of its faster 1700X and 1800X counterparts.

2D Benchmarks: Adobe Creative Cloud

As with most productivity-focused workloads, AMD's Ryzen 7 processors battle fiercely for a position near the top of these charts. They're especially attractive in light of their low price points. Intel's less expensive Core i7-7700K leads though, so it takes a Ryzen 7 1700 to make a value case.

3D Benchmarks: DirectX And OpenGL

We see more of a delta between the 65W 1700 and 95W X SKUs during the DirectX and OpenGL tests. Many of these metrics respond best to single-threaded performance at high clock rates, so Kaby Lake enjoys a natural advantage.

CPU Performance: Workstation

Parallelized workloads are in Ryzen's wheelhouse, so all three models turn in strong results in our SolidWorks 2015 and 3ds Max 2015 composite/computing benchmarks. In many cases, they're faster than Intel's pricier Core i7-6900K in its stock configuration.

CPU Performance: Photorealistic Rendering

The 95W Ryzen processors compete aggressively against the Core i7-6900K in our CPU rendering test suite, while the 65W 1700 trails its faster counterparts and battles the stock Core i7-7700K during many of the Blender tests. AMD's 1700 does beat the Core i7-6900K in 3ds Max, but falls behind during the LuxRender benchmark.

CPU Performance: Encoding & Compression/Decompression

The 1700 is particularly competitive in HandBrake. It beats the -6900K during high-quality encoding tasks, and puts up a fight in our test using default settings. At $330, it's impossible to dismiss the 1700's value proposition compared to Intel's $1100 price point. Likely due to its lower frequency, the 1700 loses steam when we decompress an archive using 7-Zip (it even trails the old FX), but it fares better in the more parallelized compression workloads.

HPC Benchmarks (High Performance Computing)

The results of these HPC-oriented benchmarks vary depending on how they utilize each architecture. As a general trend, though, when a test is able to exploit Ryzen 7's eight cores, AMD beats the Core i7-7700K. It's usually able to give Core i7-6900K a run for its money, too. In the cases where Intel's -6900K is significantly faster, we might suspect specific optimizations for efficiency on Intel CPUs. We know, for example, that there are accelerated packages for LAMMPS, one of which includes Intel CPUs and Xeon Phi. SPEC doesn't say if any of these are part of its wpc 2.0 suite, though. AMD's market penetration with the Naples server chips will likely determine how much reaction we see in the HPC software ecosystem.

Power Consumption & Temperatures

Direct Comparisons of Power Consumption

The 65W 1700 consumes slightly more power at idle than the 95W Ryzens.

Meanwhile, our mildly overclocked Core i7-6900K consumes more power at idle than its stock configuration because we reduced its single-core Turbo Boost frequency to achieve a 3.9 GHz clock rate.

The 1700's power consumption is impressive during the AutoCAD 2015 workload; it only consumes 29.3W. A stock Core i7-7700K uses considerably more power. But looking at these figures on their own can be misleading. Remember that Intel's top Kaby Lake-based CPU demonstrated a commanding lead in the previous page's AutoCAD workloads, so it ends up offering superior performance per watt.

AMD's Ryzen 7 1700 proves its frugal nature by drawing only 44.3W during our gaming benchmark. The -6900K consumes less power than Intel's Core i7-7700K, likely because the workload doesn't fully utilize all eight cores.

The 32nm FX-9590 is in a class of its own, which isn't a good thing. Still, it highlights one of the 14nm process' main advantages.

Prime95's Small FFT stress test pushes power consumption to the max, revealing one of the 65W 1700's best attributes: it consumes 23.3W less than the 1700X. The 1700's modest power use, coupled with a small performance delta between it and the 1700X in our application benchmarks, paints a convincing picture of efficiency.

Temperatures

We optimized our CPU cooler for Socket AM4 by using two nuts between the spring and bracket to increase the force on the package to 0.4Nm. That is why these results differ from those in our launch article, where we only used washers.

Both AMD and Intel employ different temperature measurement methodology. While these readings aren't entirely comparable, they do serve as a close approximation.

In its stock form, Ryzen 7 1700 runs cooler than the rest of the field due to its lower TDP. Of course, all bets are off once you start overclocking and dialing in higher voltages. In any case, AMD uses solder between its die and heat spreader, which generally provides better thermal transfer than thermal paste. Intel famously uses thermal paste and contends that it boosts processor longevity.

Even though our results aren't apples-to-apples, it's clear that the 1700's 65W TDP does convey an expected power and thermal advantage over the other Ryzen 7s.

Conclusion

Ryzen 7 1700 pulls AMD's entry-level eight-core price point below the quad-core Core i7-7700K. That's powerful. The 1700's compelling performance in desktop productivity and content creation workloads, even at stock settings, is impressive. Moreover, the chip's power consumption and thermal characteristics align with our expectations of a CPU with a 30W-lower TDP. Combine those two strengths and you have a recipe for hard-to-beat efficiency.

The 1700 also challenges or beats Intel's Core i7-6900K in several of our application tests. Moreover, it sells for roughly one-third of the -6900K's $1100 price tag, and the bundled cooler is a nice bonus. Overall, Ryzen 7 1700's price to performance ratio is very attractive for most productivity use-cases.

AMD's higher-end Ryzen 7 1700X and 1800X are fast enough for smooth gaming. But in light of their $400 and $500 prices, there's no real reason to recommend them over Intel's Core i5-7600K or Core i7-7700K. Ryzen 7 1700 demonstrates the same behaviors as both X-series SKUs, albeit with lower frame rates if you refrain from overclocking. Tuning the 1700 somewhat aggressively breaks the 65W chip free of its shackles and allows it to trade blows with the faster Ryzen 7s. It's a bummer that our sample didn't overclock quite as well; it just wouldn't crack 4 GHz like the X-series parts. Still, a respectable clock rate ceiling enables similar performance as the 95W models, so if you expect to overclock, spending more money on a 1700X or 1800X may not make sense.

The recent Ashes of the Singularity: Escalation and Dota 2 patches bring hope that game developers can, and will, address Ryzen's most quantifiable weakness. Ideally, all developers would follow suit, but in reality, most older games won't be changed. It's the future of gaming we must look to. And in that, the more accessible Ryzen 5 and 3 families may prove to be even better options for gaming, perhaps encouraging devs to spend more time improving the FHD experience on Ryzen-based platforms.

For now, Ryzen 7 1700 provides good-enough gaming performance at an acceptably low price point, which makes it a viable option for anyone shopping for an eight-core workstation-class chip ready for some entertainment, too.

MORE: Best CPUs

MORE: Everything Zen: AMD Presents New Microarchitecture At HotChips

MORE: Intel Kaby Lake Core i7-7700K, i7-7700, i5-7600K, i5-7600 Review

MORE: Broadwell-E: Intel Core i7-6950X, 6900K, 6850K & 6800K Review

AMD Ryzen 7 1700X Review

palcorn@outlook.com (Paul Alcorn) — Mon, 27 Mar 2017 13:00:00 +0000

Introduction

AMD is trying to shake up the market with shockingly low prices for its 8C/16T Ryzen 7 line-up. And while these CPUs don't dominate every workload, there is hope the company's newest architecture is compelling across enough segments to put much-needed pressure on Intel.

One component of AMD's strategy involves attractive pricing. The flagship Ryzen 7 1800X grabbed attention for its ability to battle Intel's Broadwell-E-based Core i7-6900K for $550 less (and with the same number of execution cores). We agree that the 1800X is compelling in threaded productivity and content creation apps. But we think you'll derive more value out from the cheaper Ryzen 7 1700X ($400) and 1700 ($330). The former goes up against Intel's $450 Core i7-6800K, while the latter undercuts Core i7-7700K. In both cases, the AMD chips wield more processing resources than the Intel competition.

Of course, that doesn't mean AMD runs the table in benchmarks. In our AMD Ryzen 7 1800X CPU Review, we saw the new chip offer a solid price-to-performance ratio for productivity and workstation-class workloads. Unfortunately, we couldn’t say the same about gaming. We followed up with testing in a wider range of popular games, and while Ryzen delivered playable performance in most titles, cheaper Core i7-7700K and Core i5-7600K chips were typically better.

Zen, In-Depth

From a high level, various enhancements over Zen’s predecessors, such as Simultaneous Multi-Threading (SMT), a new micro-op cache, better branch prediction, a wider instruction scheduler window, and faster caches contribute to the architecture's increased IPC throughput. Check out Everything Zen: AMD Presents New Microarchitecture At HotChips for more.

AMD claims that the gaming performance issues stem from how applications interact with the intricacies of its new architecture. The company expects a wave of updates from various developers that will eventually remedy this (though so far only two devs have publicly committed to optimizing their engines for the new processors). Until something concrete happens, though, we don't see much value in gaming-specific Ryzen 7 1800X builds.

Might the Ryzen 7 1700X cast a more favorable light on gaming? After all, it costs $100 less, carries over the eight-core configuration with 16MB of L3 cache, and continues to offer an unlocked ratio multiplier.

AMD Ryzen 7 1800X

Ryzen 7 1700X

Ryzen 7 1700

The unlocked multiplier is especially interesting, given the similarities up and down the Ryzen 7 family. Given a similar 95W TDP between the $500 1800X and $400 1700X, then, the only technical differences between them are their base, two-core Precision Boost, and XFR clock rates. Out of the box, 1800X enjoys a 200 MHz advantage down low and up top. But we've heard claims that 1700X hits a similar ceiling as 1800X when it comes to overclocking.

Right out of the gate, Ryzen 7 1700X looks like a smarter buy than 1800X. But is it smart enough to maintain AMD's strong position in well-threaded desktop apps and make up some value ground in gaming, where the architecture isn't as strong?

MORE: Best CPUs

MORE: Intel & AMD Processor Hierarchy With CPU Comparisons

MORE: Everything Zen: AMD Presents New Microarchitecture At HotChips

MORE: Intel Kaby Lake Core i7-7700K, i7-7700, i5-7600K, i5-7600 Review

MORE: Broadwell-E: Intel Core i7-6950X, 6900K, 6850K & 6800K Review

Overclocking And Test Setup

Overclocking

You can download AMD's Ryzen Master utility from the company's website. The tool makes overclocking from within Windows nice and convenient. However, we still prefer making firmware-based adjustments.

Our 1800X hit 4 GHz at 1.425V across all of its cores in our launch article, but we stepped that back to a more conservative 3.9 GHz overclock in this piece to represent a safer long-term frequency. We achieved Prime95 stability on both the 1800X and 1700X by dialing core voltage up to 1.3875V and bumping CPU SoC voltage to 1.2V. We left Load Line Calibration set to Auto, but boosted the Crucial LPX memory modules to DDR4-2933 for our overclocked setups. That's only a slight increase over the DDR4-2699 we used for the stock gaming configurations.

While we've seen many enthusiasts hit 4.0-4.1 GHz with much higher voltages, AMD says 1.35V is the "safe" ceiling for long-term overclocking. We didn't run into any thermal issues at 3.9 GHz, and recorded 70°C (per AIDA) after a two-hour Prime95 run with the capable Noctua NH-U12S SE-AM4 heat sink in place. Waste heat appears to accumulate quickly at higher clock rates (we hit 82°C at 4 GHz using Corsair's H100i v2). Invest in a beefy cooler if you plan on pushing the limits of Ryzen's frequency headroom.

Test Setup

We've experienced general platform instability with numerous AM4-based motherboards and early firmware revisions. This is typically pinned on unoptimized AMD Generic Encapsulated Software Architecture microcode. AGESA is a bootstrap protocol that initializes processor cores, memory, and the HyperTransport (now Infinity Fabric) controller. We're using Asus' Crosshair VI Hero for this round of testing, along with the latest 1001 firmware. This version demonstrated improved stability at stock and overclocked settings. We'll continue to monitor the firmware updates from other vendors and note when they've achieved a similar degree of stability. AMD is expected to release an AGESA revision in late March that should enhance memory support, including opening up more sub-timings.

Early firmware updates yielded large performance increases over AMD's original press kit. As time goes on, though, the gains get smaller and smaller. Asus mentioned that we shouldn't expect big changes from future builds.

In general, we've recorded more performance variation from Ryzen processors than we're accustomed to in a few games. Ashes of the Singularity is one example. Incidentally, that's also a title expected to improve once the developer follows up on promises to optimize for AMD's architecture. We suspect that some of this inconsistency stems from the impact of cross-CCX thread migration and other peculiarities of Zen. In response, we've disabled several transitory background services that fire up with little to no provocation and may promote thread migration, such as Windows Search and Defender. Frequent reboots during testing also helped weed out obvious outliers, yielding more consistent results. We're again testing under Windows' High performance power plan with HPET disabled. As we demonstrated in our Ryzen 7 1800X review, some games benefit from disabling SMT, though that also leads to lower performance in other titles. Plus, we don't think you should be expected to toggle this feature on and off. As such, we're leaving SMT enabled for today's story.

As expected, we also encounter graphics bottlenecks in several titles at 2560x1440. AMD argues that its Ryzen processors fare best at higher resolutions, for obvious reasons: as you start shifting the burden to your GPU, host processing weaknesses are easier to mask. Further, outside of one exception in Middle-earth: Shadow of Mordor, we only recorded minor hierarchy changes between FHD and QHD in our Ryzen Versus Core i7 in 11 Popular Games article. In other words, the same performance trends at 1920x1080 carry over to 2560x1440.

If you want to know more about how the Tom's Hardware DE system looks and is controlled, check out How We Test Graphics Cards.

Test Systems and Measurement Setups
Systems	Germany AMD 1Ryzen 7 1800X, 1700XMSI X370 XPower Gaming Titanium2x Corsair Vengeance LPX DDR4-3000 @2666 MT/sIntel LGA 2011-v3 Intel Core i7-6900K MSI X99S XPower Gaming Titanium 4 x 4GB Crucial Ballistix DDR4 2400Intel LGA 1151 Intel Core i7-7700K MSI Z270 Gaming 7 2 x 8GB Corsair Vengeance DDR4-3200 @2400 MT/sAMD Socket AM3+ FX-9590 Asus Crosshair V Formula 2 x 8GB Corsair Dominator DDR3 2133 @1866 MT/sGermany All1x 1TB Toshiba OCZ RD400 (M.2, System SSD)2x 960GB Toshiba OCZ TR150 (Storage, Images)be quiet Dark Power Pro 11, 850WWindows 10 Pro (All Updates)US AMD 1Ryzen 7 1800X, 1700XAsus ROG Crosshair VI Hero2x Corsair Vengeance LPX DDR4-3000 @2666 MT/sUS AMD 2AMD FX-8350MSI 970 Gaming2x Kingston HyperX DDR3-2133USA Intel 1Intel Core i7-7700KMSI Z270 Gaming M72x Corsair Vengeance LPX DDR4-3000 @2400 MT/sUSA Intel 2Core i7-6900KASRock X99 Extreme44x Crucial DDR4-2400US All1TB Samsung PM863SilverStone ST1500, 1500WWindows 10 Pro (All Updates) Version 1607
Cooling	Germany- Alphacool Eispumpe VPP755 Pump - Alphacool NexXxoS UT60 Full Copper 240mm- Alphacool Eisblock XPX CPU-Alphacool Cape Corp Coolplex Pro 10 LT- 5x be quiet! Silent Wings 3 PWM- Thermal Grizzly Kryonaut US-Corsair H100iv2-Noctua NH-U12S SE-AM4-Arctic MX-4
Case	Lian Li PC-T70 with Expansion Kit and Mods
Power Consumption Measurements	- Contact-free DC Measurement at PCIe Slot (Using a Riser Card) - Contact-free DC Measurement at External Auxiliary Power Supply Cable - Direct Voltage Measurement at Power Supply- 2 x Rohde & Schwarz HMO 3054, 500MHz Digital Multi-Channel Oscilloscope with Storage Function - 4 x Rohde & Schwarz HZO50 Current Probe (1mA - 30A, 100kHz, DC) - 4 x Rohde & Schwarz HZ355 (10:1 Probes, 500MHz) - 1 x Rohde & Schwarz HMC 8012 Digital Multimeter with Storage Function
Thermal Measurements	- 1 x Optris PI640 80Hz Infrared Camera- PI Connect Analysis Software with Profiles
Noise Measurements	- NTI Audio M2211 (with Calibration File)- Steinberg UR12 (with Phantom Power for Microphones)- Creative X7, Smaart v.7- Custom-Made Proprietary Measurement Chamber, 3.5 x 1.8 x 2.2m (L x D x H)- Perpendicular to Center of Noise Source(s), Measurement Distance of 50cm- Noise Level in dB(A) (Slow), Real-time Frequency Analyzer (RTA) - Graphical Frequency Spectrum of Noise

3DMark, Ashes of the Singularity, Battlefield 1 & 4

3DMark

We don't consider synthetic benchmarks to be a good measure of real-world gaming performance, but 3DMark's DX11 physics and DX12 CPU tests provide useful metrics that quantify the amount of processing power available to the game engine.

The stock Ryzen 7 1800X comes close to unseating Intel's Core i7-6900K during the DX12 CPU test. AMD's 1700X trails its faster counterpart by 383 points. However, overclocking both chips to 3.9 GHz shrinks the gap to 72 points...in favor of the 1700X.

Futuremark's DX11 Fire Strike benchmark runs 32 parallel soft and rigid body physics simulations that tax the processor specifically. We notice a similar trend during the DX11 physics test; both Ryzen processors lag the i7-6900K at stock settings, but overtake Intel when overclocked. Rest assured that we're planning a follow-up that has each of these unlocked CPUs tuned up to show how they ultimately fall in relation to each other.

The Ryzen processors offer better single-threaded DX11 performance when we overclock them, yet curiously turn in worse results in the threaded DX11 test compared to the stock settings. We repeated this test multiple times just to be sure. In spite of the odd outcome, both AMD CPUs beat the Core i7-6900K in 3DMark's DX11 single-threaded test, regardless of clock rate.

Ryzen 7 demonstrates a tremendous DX11 single-thread deficit compared to the Core i7-7700K and i5-7600K, which helps explain Intel's commanding lead in titles that are primarily sensitive to single-threaded performance. The 3DMark DX12 API overhead test reveals a large gap between the stock Ryzens and Core i7-6900K. Even overclocking can't make up the difference.

Ashes of the Singularity

Ashes of the Singularity is notoriously CPU-bound, but it responds well to higher core counts and clock rate, which makes it particularly well suited for examining the impact of increased processing performance on CPU-intensive titles. Unfortunately, the game engine doesn't play well with AMD's new architecture, which is a punishing reality considering its hefty 16-thread SMT implementation. Oxide Games has voiced its intention to optimize for Ryzen, but the timing specifics remain unclear.

Ryzen 7 CPUs lag behind the Intel competition (even its Core i5) by a considerable margin.

At stock clock rates, the 1800X outpaces AMD's 1700X by 2.1 FPS on average. However, the overclocked Ryzen 7 1700X roughly matches the higher-end model. Given the frequency boost and correspondingly small speed-up, something other than clock rate is holding Ryzen back.

Fortunately, all of the processors except AMD's FX-8350 provide smooth-enough frame rates at 1920x1080.

Battlefield 1

We dialed Battlefield 1 up to Ultra preset and trudged across the landscape in O La Vittoria. The game flirts with a graphics bottleneck, even at 1920x1080, so there isn't as much performance variation between host processors.

While both overclocked Ryzen 7 CPUs trump the stock configurations, average frame rates in Battlefield 1 are so close that there's little reason to compare performance. Only the FX-8350 stands out as a notably slower CPU unworthy of your high-end graphics hardware.

Battlefield 4

Battlefield 4 behaves similarly, also emphasizing graphics performance (even at 1920x1080). As a result, the Ryzen 7 CPUs are right there in the mix with Intel's fastest desktop processors.

The stock Ryzen 7 1700X trails the field slightly, but bumping it to 3.9 GHz adds a few FPS, edging out the stock 1800X. Meanwhile, the FX-8350 is our only real loser.

Civilization VI AI & Graphics Test, Deus Ex: Mankind Divided, GTA V

Civilization VI AI Test

Civilization VI's AI benchmark measures the amount of computational horsepower available to the system during a turn-based strategy gaming session.

The overclocked Ryzen processors nudge past Intel's Core i7-6900K, but can't catch Kaby Lake. While it's likely threaded, this metric clearly isn't able to utilize more than four cores.

Civilization VI Graphics Test

The Core i7-6900K rockets to the top of this chart, but Intel's Core i7-7700K achieves a better minimum frame rate. Remember that all of the Intel CPUs in our story also sport unlocked multipliers, and would benefit handsomely from overclocking as well.

The two overclocked Ryzen 7s enjoy quantifiable gains from the jump to 3.9 GHz. Our 1800X registers a 3.6% gain over its stock settings, and the 1700X scores a 6.5% performance increase. A stock Core i5-7600K can't carry over its top position from the AI Test, and instead lands under the stock Ryzen 7 1700X.

Deus Ex: Mankind Divided

Deus Ex: Mankind Divided provides rare respite for the Ryzen processors; they beat their Intel competition in convincing fashion during the game's benchmark. We asked Eidos for technical detail about the game engine's behavior and await more information.

The difference between Intel and AMD processors is obvious from their FPS result.

Although Ryzen appears graphics-bound, the two CPUs at 3.9 GHz do average higher frame rates than the stock configurations.

Grand Theft Auto V

We pushed the graphics settings as high as they'd go to characterize real-world gaming under Grand Theft Auto V with a high-end GPU. We measure performance during the F-16 flight sequence in the built-in benchmark. The constantly changing terrain of the expansive scene yields a solid and consistent benchmark.

The Core i7-7700K leads with a 91.1 FPS result, and the rest of the field scales down from there. Intel's i5-7600K also turns in a stellar performance, particularly in light of its budget-friendly price point. Although Ryzen 7 provides smooth-enough performance, it lags the competition by a quantifiable margin. Overclocking does clearly help, and the two CPUs at 3.9 GHz fare similarly.

Hitman (2016), Metro: Last Light Redux, Middle-earth: Shadow of Mordor

Hitman (2016)

The Intel processors maintain a lead during this test. Although we observe a moderate gain from overclocking Ryzen, AMD's deficit suggests something else is bottlenecking these CPUs other than clock rate.

Also interesting is that Core i5 trails the i7s and Ryzen. Could its lack of Hyper-Threading really incur such a performance hit compared to Core i7-7700K?

Further, we notice several frame time outliers with the stock 1800X that shows up in both our frame time and unevenness charts. Overclocking smooths out some of the inconsistency.

Metro: Last Light Redux

Even at 1920x1080, Metro Last Light is largely graphics-bound. Nevertheless, all four Ryzen data points trail behind the three Intel CPUs, and not even overclocking changes this.

Middle-earth: Shadow of Mordor

Middle-earth also appears to be limited by graphics horsepower. Still, the Core i7-770K leads slightly, followed by the Core i5-7600K. The Ryzen 7 processors trail at their stock settings, though overclocking them is enough to beat a stock Core i7-6900K.

Project CARS, Rise of the Tomb Raider, The Division

Project CARS

The Core i7-7700K takes a beastly lead during our Project CARS test run. AMD's Ryzen 7 CPUs obvious struggle at their stock settings, though overclocking helps quite a bit.

The 1800X does fare somewhat better than the 1700X right out of the box, and as a result of its higher base and XFR frequencies, it doesn't benefit as much from an overclock to 3.9 GHz.

AMD's old FX-8350 lags woefully behind the rest of the field, suffering from latency spikes throughout our run. This was perceivable as intermittent bouts of choppy performance.

Rise of the Tomb Raider

Quite the opposite of Deus Ex, Rise of the Tomb Raider is a marked weakness for AMD's Ryzen CPUs. They lag behind the Intel competition by 35 to 45 FPS in their stock configurations.

The Ryzens jump up to the ~120 FPS range in response to our moderate overclock, improving their standing somewhat.

The Division

Tom Clancy's The Division appears more graphics-bound, allowing the Ryzen chips to compete alongside Intel's Broadwell-E and Kaby Lake architectures. Overclocking even helps propel the Ryzen 7s in front of the stock Core i7-6900K. If you're primarily buying new hardware to game, though, that Core i5 still looks like a pretty solid performer.

Workstation And HPC Benchmarks

Our German team re-tested all of this hardware for today's story using the latest firmware. This especially affects the Ryzen 7 1800X and its 3.8 GHz overclocked results. Therefore, some of the updated test data no longer agrees with the launch results. Consider this a good thing. Maturity is helping Ryzen perform better.

2D Benchmarks: DirectX And GDI/GDI+

We extended our AutoCAD 2D and graphics throughput benchmark for the GDI/GDI+ functions to the new CPUs and summarized the results in a common section. Absolutely nothing changed from the key conclusions in our launch article.

2D Benchmarks: Adobe Creative Cloud

The Ryzen 7 1700X is well-suited for a number of threaded workloads, even if it trails the 1800X due to lower stock clock rates. Still, the differences we measure are often not noticeable during normal use.

3D Benchmarks: DirectX And OpenGL

The performance of these CPUs in our professional graphics workloads doesn't really change from what we saw in our 1800X launch coverage. However, we see that Ryzen 7 1700X is at a disadvantage without overclocking, since the architecture's IPC throughput is lower than Intel's. This particularly hurts in single-threaded tasks.

CPU Performance: Workstation

As we shift to more parallelized CPU-bound workloads, the Ryzen CPUs hit their stride and start beating some of the Intel processors that previously appeared much faster.

CPU Performance: Photorealistic Rendering

Workloads tuned for multi-core/multi-processor configurations love Ryzen. The stock 1800X and 1700X consistently appear in the top half of our test field, while the overclocked 1800X takes first or second place across the board. If you do a lot of rendering in Blender, 3ds Max, or LuxRender, Ryzen represents a good way to save money without compromising performance.

Considering the similar overclocking headroom available on 1700X and 1800X CPUs, a tuned 1700X should serve up even better value.

CPU Performance: Encoding And Compression/Decompression

AMD straight-up rocks our HandBrake encoding tests.

It's not as dominant in 7-Zip, but does trade blows with the Intel competition. Take a look at the 7-Zip decompression chart, specifically. Since Core i7-7700K lands at the top, we can assume this workload can't fully utilize the 8C/16T CPUs. This explains why a stock Core i7-6900K lands among the Ryzen configurations.

So long as the behavior of your workload is somewhat predictable, it's possible to guess how Ryzen will fare against Intel's nimble Kaby Lake models and beefy Broadwell-Es.

HPC Benchmarks (High Performance Computing)

Power Consumption And Temperatures

Direct Comparisons of Power Consumption

Similar to our performance numbers, we re-ran the 1800X review's power data to reflect the continual improvements being made to motherboard firmware.

Let's start by looking at idle power consumption. Intel's overclocked Core i7-6900K turns in a better result than the stock configuration because we also reduced the one-core Turbo Boost frequency.

The 95W AMD CPUs clearly use less power in our combined CAD benchmark. When we weigh average performance against power consumption, however, the eight-core CPUs from AMD and Intel are fairly similar.

The same story applies to our measurements in games, where Ryzen 7 blows away the Core i7s. It's been a long time since AMD was at least equal in terms of efficiency.

The 3.8 GHz Ryzen 7 sucks down more than 140W , but the Core i7-6900K is even worse at 166W. The Core i7-7700K down-clocked to the same frequency uses a conservative 86W or so.

Based on AMD's technology briefings, we know it has more granular control over clock rate. And it's notable that the 95W Ryzen 7 1700X we're reviewing today uses less power under our stress test than Intel's 91W Core i7-7700K. It takes a significant underclock to put the Kaby Lake flagship in first place.

Temperatures

We optimized our CPU cooler for Socket AM4 by using two nuts between the spring and bracket to increase force on the package to 0.4Nm. That is why these results differ from those in our launch article, where we only used washers.

The temperatures we recorded for the FX-9590 are a bit uncertain, since AMD’s older Bulldozer CPUs don't measure with 100% confidence. Moreover, the Ryzen 7 and Core i7 CPU readings aren't exactly comparable; both companies employ different sensor approaches.

Intel's Core i7-7700K is the only processor in our test field handicapped by cheap thermal paste between its die and heat spreader. Thankfully, AMD solders Ryzen's heat spreader, which results in good thermal transfer. This naturally shows up in the relationship between power converted to heat.

Enthusiasts should be happy with Ryzen's power consumption and its resulting waste heat.

Conclusion

When AMD announced that all of its Ryzen 7 CPUs would get unlocked multipliers and eight cores, we immediately imagined opportunities for enthusiasts to snag the lower-end models for less and overclock to great effect. Our testing shows that, if you're willing to spend some time tuning, the Ryzen 7 1700X offers better overall value than the 1800X at a similar 95W TDP.

This shouldn't be surprising. After all, Ryzen 7 1700X bears many of the flagship's attributes, differing mainly in base and peak clock rates. As a result, you get satisfactory performance in heavily threaded workloads at a lower price. Though most professionals eschew overclocking in favor of stability, even at its stock settings the 1700X serves up a solid price-to-performance ratio.

The story gets better once you push Ryzen 7 1700X up to and beyond the 1800X's default configuration. And overclocking is simple, too. We bumped the core voltage up to 1.3875V and increased the SoC voltage to obtain a rock-solid 3.9 GHz across all of the chip's cores. With the right cooler, a moderate overclock doesn’t come close to pushing Ryzen's thermal boundaries, and there is likely some additional headroom if you're willing to tolerate higher voltages. Memory tuning is still somewhat limited, but we're told that future firmware updates should make more settings available.

Of course, the 1800X's challenges in games carry over to the 1700X, too. Even overclocked, Ryzen lags behind cheaper Intel CPUs in much of our gaming suite. Your best chance of seeing parity comes from graphics-bound resolutions and detail settings. AMD claims that patches may address some of our concerns, but we aren't holding our breath. It's far more probable that future titles include optimizations for AMD's new architecture.

It would be easier for us to recommend Ryzen 7 to gamers if it was less expensive. But with Core i7-7700K and Core i5-7600K performing so well, and both CPUs less expensive than the 1700X we're reviewing today, Kaby Lake maintains its leadership. But there remains near-term hope for the Ryzen family: AMD's Ryzen 5 series will surface early in April at price points better suited to take on mainstream Core CPUs.

We've established that Ryzen 7 1700X tells a better value story than the 1800X, and it gets even better if you're willing to live with the B350 chipset's reduced feature set. Enthusiasts willing to overclock should be able to match or exceed the 1800X's stock performance with little effort. Now, what we really want to know is whether Ryzen 7 1700 is the best model of all, or if dipping down to the 65W model means giving up some overclocking headroom. Stay tuned!

MORE: Best CPUs

MORE: Everything Zen: AMD Presents New Microarchitecture At HotChips

MORE: Intel Kaby Lake Core i7-7700K, i7-7700, i5-7600K, i5-7600 Review

MORE: Broadwell-E: Intel Core i7-6950X, 6900K, 6850K & 6800K Review

AMD Announces Ryzen 5 Processors; $169 Four-Cores And $219 Six-Cores; AMD Confirms 2+2 Alignment

palcorn@outlook.com (Paul Alcorn) — Thu, 16 Mar 2017 02:05:00 +0000

We're still exploring the Ryzen 7 processors, but AMD is forging ahead quickly with its Ryzen 5 lineup. The new Ryzen 5 processors take aim at the heart of the market with 6-core and 4-core variants at competitive price points lining up against Intel's four-core i5 series. AMD has the core count advantage in its shockingly low price range, but it hasn't released many technical details. We do know that the processors will come to market on April 11, and AMD will release more details as review units begin shipping in a few weeks.

How To Disrupt The Market

Like their Ryzen 7 siblings, all of the Ryzen 5 processors come with an unlocked multiplier and the SenseMI suite. None of the new models come with integrated graphics.

The Ryzen 5 1600X comes to market as the flagship offering, but it carries a budget price. AMD offers six cores and twelve threads for a mere $249. The 1600X wields a 3.6/4.0 GHz base/boost frequency and 16MB of L3 cache, but the company hasn't revealed the XFR clocks. We do know that it features a 95W TDP, which is the same as the Ryzen 7 1800X and 1700X. The CCX alignment is important, but we'll dive back in on that shortly with some context.

From a pricing perspective, a $249 six-core/twelve-thread chip is rather impressive compared to a $350 four-core/eight thread i7-7700K, but of course, performance will tell the tale. It's really aimed at the $240 four-core i5-7600K, which comes without Hyper-Threading. The 1600 features 16MB of L3 cache.

	Ryzen 5 1600X	Ryzen 5 1600	Ryzen 5 1500X	Ryzen 5 1400	Ryzen 7 1800X	Ryzen 7 1700X	Ryzen 7 1700
MSRP	$249	$219	$189	$169	$499	$399	$329
Interface	1331	1331	1331	1331	1331	1331	1331
Process	14nm GloFo	14nm GloFo	14nm GloFo	14nm GloFo	14nm GloFo	14nm GloFo	14nm GloFo
Cores/Threads	6/12	6/12	4/8	4/8	8/16	8/16	8/16
TDP	95W	65W	65W	65W	95W	95W	65W
Base Freq. GHz	3.6	3.2	3.5	3.2	3.6	3.4	3.0
All-Core Precision Boost	?	?	?	?	3.7	3.5	3.1
Precision Boost Freq (dual core)	4.0	3.6	3.7	3.4	4.0	3.8	3.7
XFR Frequency GHz	?	?	3.9 GHz	?	4.1	3.9	3.8
Cache (L3)	16MB	16MB	16MB	8MB	16MB	16MB	16MB
Unlocked Multiplier	Yes	Yes	Yes	Yes	Yes	Yes	Yes

The Ryzen 5 1600 is the entry-level six-core that comes without XFR functionality. It retails for $219 and features a 3.2/3.6 GHz base/boost frequency, 16MB of L3 cache, and a 65W TDP.

AMD also offers two four-core/eight-thread Ryzen 5 processors. The 1500X retails for $189 and has a 3.5/3.7 GHz base/boost clock, XFR functionality, and a 65W TDP. XFR (eXtreme Frequency Range) clocks the 1500X up to 3.9 GHz automatically if the cooling solution provides enough thermal headroom. The Ryzen 7 processors provide a meager 100 MHz boost, so it’s encouraging to see a step up to 200 MHz, but we aren't sure if that applies to the entire range of Ryzen 5 processors. It comes with a 16MB L3 cache.

The four-core Ryzen 5 1500 comes without the benefit of XFR technology and retails for $169. The clock rate takes a step back to 3.2/3.4 GHz base/boost. The 1500X pulls back to 8MB of L3 cache.

AMD Ryzen 7 1800X

Ryzen 7 1700X

Ryzen 7 1700

Queue The Cinebench

Once again, AMD treated us to a multi-threaded Cinebench workload against an Intel CPU. AMD claimed that the Ryzen 5 1600X beats the Intel Core i5-7600K in the multi-threaded workload by 69%. The results aren't entirely surprising considering the i5-7600K has only four cores/threads (3.8/4.0 GHz base/boost) at its disposal, whereas the Ryzen 5 1600X weighs in with six cores and twelve threads (3.6/4.0 base/boost Ghz).

As everyone hopefully learned with AMD's pre-launch Ryzen 7 Cinebench benchmarks, the test offers only a narrow view of overall performance and doesn't tell us anything about gaming performance. AMD is tight-lipped about important details (as usual) while it builds up the hype. We will have to wait for our product reviews to determine how well the 1600X stacks up in more diverse workloads (such as gaming).

In-Box Coolers

The Ryzen 5 1400 comes with the Wraith Stealth stock cooler, whereas the 1600 and 1500X come with the beefier Wraith Spire. AMD doesn't include a stock cooler with the 1600X, as it predicts that most enthusiasts will use a third-party cooler with the high-end processor. AMD isn't offering its bling-bling LED stock heatsink (the Wraith Spire LED cooler) with the Ryzen 5 processors, and it isn't available separately.

One Socket To Rule Them All

AMD pointed out that most users will use the Ryzen 5 processors with the B350 chipset, which allows for overclocking via the unlocked multiplier (an advantage over Intel's value chipsets). Like the rest of the Ryzen processors, the Ryzen 5 chips will drop into any Socket AM4 motherboard, which provides a wide range of options and also provides an easy upgrade path in the future.

Back To The CCX (Core Complex)

Before we get started on the CCX, here's a small primer for the uninitiated (pulled from our recent Ryzen gaming piece):

Theories abound as to why Ryzen processors are struggling in gaming metrics, but some of the disparity no doubt comes from an IPC and clock rate deficit compared to Intel's Kaby Lake design. The issue also appears to stem from AMD’s Zen architecture and how applications navigate its cache hierarchy. The Zen architecture employs a four-core CCX (CPU Complex) building block. AMD outfits each CCX with a 16-way associative 8MB L3 cache split into four slices; each core in the CCX accesses this L3 with the same average latency. Two CCXes come together to create an eight-core Ryzen 7 processor (image below), and they communicate via AMD’s Infinity Fabric interconnect. Data that traverses the void between CCXes incurs increased latency, so it's ideal to avoid the trip altogether if possible. Unfortunately, threads migrate between the CPU Complexes, thus suffering cache misses on the local CCX's L3. Threads might also communicate with other threads (and their data) running on the CCX next door, again adding latency and chipping away at overall performance.

UPDATE, 3/16/17, 7:52am PT : AMD has confirmed that the Ryzen 5 processors employ a dual-CCX design. The 6-cores have a 3+3 alignment (three cores on each CCX), and the four-cores use a 2+2 alignment (two cores per CCX). This means the processors will likely suffer the same lower performance in gaming scenarios, and in some cases it might be worse, as the processor will be forced to allocate threads to the second CCX more frequently.

The CCX alignment appears pretty straightforward for the six-core Ryzen 5 models. The Ryzen 5 1600X's base/boost frequencies and TDP matches the Ryzen 7 1800X, so the 1600X employs the same dual-CCX (Core Complex) architecture, albeit with a few cores disabled due to defects (or just for the sake of product segmentation).

Performance could vary depending upon how AMD aligns the disabled cores. We know that latency increases as data navigates the chasm between the two CCX (via the Infinity Fabric). Lopsided core allocations (four cores on one CCX and two on the other, for instance) that vary per processor (perhaps 3+3 on some) could result in varying levels of application performance even among the same models (Update 3/16/17, 7:52am PT: AMD has confirmed that the cores are disabled in a symmetrical alignment). AMD is disabling cores on all of the models, but most still employ the entire 16MB cache, which makes some of the six-core models particularly attractive (based on price) compared to the eight-core Ryzen 7's. The Ryzen 5 1400 is the only model that features a trimmed L3 cache (8MB). The four-core 1500X with 16MB of cache is also impressive- the competitive i5's only come with 6MB of L3 cache, while the i7-7700K has a comparatively paltry 8MB. It will be interesting to measure the performance impact of the healthy L3 cache allotment.

Most importantly, the four-core models will employ a dual CCX design. This means the processors will likely suffer the same lower performance in gaming scenarios, and in some cases it might be worse, as the processor will be forced to allocate threads to the second CCX more frequently.

A single four-core CCX, or only one active CCX on a dual-CCX chip, would help avoid many of the problems that appear to restrict Ryzen 7's gaming performance in many popular titles, and frankly, many enthusiasts were hoping this would be the the case. The four-core 1500X and 1400 have the same 65W TDP as the eight-core Ryzen 7 and the six-core Ryzen 5 1600X and 1600. We would expect a lower TDP or significantly higher clocks within the same TDP envelope with a single-CCX architecture. Retaining the dual-CCX design with active cores on both CCX is far from ideal, and the gaming performance disparity will likely continue.

Competition Returns

The battle between AMD and Intel will be the most fierce in the mid-range and low-end because these segments comprise the vast majority of the market. AMD's Ryzen 5 processors are quite competitively priced, but performance will also dictate whether Intel has to adjust its pricing strategy. The lack of integrated graphics can be both a blessing and a curse. It allows AMD to significantly undercut Intel on pricing, but it also restricts the addressable market.

Once again, AMD doesn't have to beat Intel's Core i7 and i5 series in all tests; it just has to be close, and then let its brutally competitive pricing do the work.

AMD has its much-anticipated Ryzen 3 series coming to market in the second half of the year, and Intel has already added the overclockable dual-core i3-7350K to its lineup. Intel's addition of Hyper-Threading to the Pentium lineup also appears to be an attempt to solidify its product stack in preparation for the low-end Ryzen competition. We will probably have to wait until the Ryzen 3 products land before we see any measurable Intel pricing reaction.

AMD will have the Ryzen 5 processors on store shelves April 11, but much like the Ryzen 7 launch, we wouldn't be surprised to see a pre-order window open up prior to general availability. As always, we would advise those with itchy trigger fingers to wait for our review--particularly, when we explore the performance impact of the CCX implementation on the four-core models.

AMD Posts Ryzen Community Update; Power Patch Coming

palcorn@outlook.com (Paul Alcorn) — Tue, 14 Mar 2017 20:40:00 +0000

AMD's relatively low performance in some games has led to a flurry of theories, with the most popular being that the Windows scheduler bears the blame. There have been widespread reports, based on a now-deleted tweet as far as we can gather, that Microsoft will issue an update to rectify a scheduling conflict. AMD released a community update that contradicts the theories.

AMD also clarified its temperature reporting methodology and announced that it's pushing an update in April to refine its power management policy. The company also released a separate blog post outlining several tweaks to get the best gaming experience.

The Windows Scheduler Is Not To Blame

Ryzen's unique cache topology and simultaneous multi-threading (SMT, which is akin to Intel's Hyper-Threading) has led to reduced performance in many games, which AMD claimed can be fixed through updates that allow software to interact correctly with the new architecture. Many opine that the Windows scheduler is improperly assigning threads to physical and logical cores, but AMD contradicts that theory:

We have investigated reports alleging incorrect thread scheduling on the AMD Ryzen processor. Based on our findings, AMD believes that the Windows 10 thread scheduler is operating properly for “Zen,” and we do not presently believe there is an issue with the scheduler adversely utilizing the logical and physical configurations of the architecture. As an extension of this investigation, we have also reviewed topology logs generated by the Sysinternals Coreinfo utility. We have determined that an outdated version of the application was responsible for originating the incorrect topology data that has been widely reported in the media. Coreinfo v3.31 (or later) will produce the correct results.

We touched on the subject on the first page of our Ryzen Versus Core i7 In 11 Popular Games article. Unfortunately, there is no global fix for the scheduler, so we will have to wait for updates from each individual developer.

Windows 10 Is Great, Thanks

Many users have also recorded elevated game performance in Windows 7 compared to Windows 10, which has fanned the scheduler-theory flames. AMD also addressed the issue in its statement:

Finally, we have reviewed the limited available evidence concerning performance deltas between Windows 7 and Windows 10 on the AMD Ryzen CPU. We do not believe there is an issue with scheduling differences between the two versions of Windows. Any differences in performance can be more likely attributed to software architecture differences between these OSes.

According to AMD, the scheduler isn't the issue, but we believe the performance delta may be due to Windows 10's more aggressive power management policy. As we outlined in our original Ryzen 7 1800X review and backed up with testing, the Ryzen processors perform better with the high-performance power profile.

The high-performance profile shifts control of the processor power states from Windows to the CPU, thus negating the operating system's incessant desire to park cores, which results in decreased performance. Windows 7 doesn't appear to park cores as aggressively, which is likely contributing to increased performance in the older operating system.

Speaking Of Power Plans

AMD also touched on the performance disparity between the balanced and high-performance power plans, and the company plans to issue an update in the April time-frame to address the issue.

Users may have heard that AMD recommends the High Performance power plan within Windows 10 for the best performance on Ryzen, and indeed we do. We recommend this plan for two key reasons: Core Parking OFF: Idle CPU cores are instantaneously available for thread scheduling. In contrast, the Balanced plan aggressively places idle CPU cores into low power states. This can cause additional latency when un-parking cores to accommodate varying loads.Fast frequency change: The AMD Ryzen processor can alter its voltage and frequency states in the 1ms intervals natively supported by the “Zen” architecture. In contrast, the Balanced plan may take longer for voltage and frequency (V/f) changes due to software participation in power state changes.In the near term, we recommend that games and other high-performance applications are complemented by the High Performance plan. By the first week of April, AMD intends to provide an update for AMD Ryzen processors that optimizes the power policy parameters of the Balanced plan to favor performance more consistent with the typical usage models of a desktop PC.

Of course, the real question is how the revised power plan will affect Ryzen's power consumption and thermal profile--an issue that we're following up with the company to ascertain. Increased power consumption and thermals inevitably lead to higher cooling requirements, and thus acoustical output from the cooling solution, so the modified power profile might have an effect on many fronts.

Speaking of Thermals

AMD also addressed Ryzen's somewhat confusing temperature reporting.

The primary temperature reporting sensor of the AMD Ryzen processor is a sensor called “T Control,” or tCTL for short. The tCTL sensor is derived from the junction (Tj) temperature—the interface point between the die and heatspreader—but it may be offset on certain CPU models so that all models on the AM4 Platform have the same maximum tCTL value. This approach ensures that all AMD Ryzen processors have a consistent fan policy. Specifically, the AMD Ryzen 7 1700X and 1800X carry a +20°C offset between the tCTL° (reported) temperature and the actual Tj° temperature. In the short term, users of the AMD Ryzen 1700X and 1800X can simply subtract 20°C to determine the true junction temperature of their processor. No arithmetic is required for the Ryzen 7 1700. Long term, we expect temperature monitoring software to better understand our tCTL offsets to report the junction temperature automatically. The table below serves as an example of how the tCTL sensor can be interpreted in a hypothetical scenario where a Ryzen processor is operating at 38°C.

Product Name	True Junction Temp (Example)	tCTL Offset for Fan Policy	Temp Reported by tCTL
AMD Ryzen 7 1800X	38°C	20°C	58°C
AMD Ryzen 7 1700X	38°C	20°C	58°C
AMD Ryzen 7 1700	38°C	0°C	38°C

In the near term, a bit of math will rectify the issue nicely, and we expect that software utilities will evolve over time to provide accurate measurements. Much like the issue with cache measurements, though, it might take some time for the optimizations to trickle down to us.

About That Whole Gaming Thing...

Finally, AMD posted a separate blog post outlining several steps we can take to enjoy optimized gaming performance, and although most of the optimizations are no-brainers for the enthusiast set, there are a few interesting tidbits for Ryzen users.

AMD also confirmed the fact that some games perform better with SMT disabled, which we already characterized with targeted testing in our launch piece.

All in all, it's been a rather busy 13 days since the Ryzen launch, and we continue to test the 1700X and 1700 for our full reviews. The latest tidbits of information are coming rapidly, and we expect to learn much more in the days to come.

Ryzen Versus Core i7 In 11 Popular Games

palcorn@outlook.com (Paul Alcorn) — Tue, 14 Mar 2017 13:00:00 +0000

Introduction

Pre-release Cinebench and Blender benchmarks showing Ryzen ahead of Core i7-6900K gave enthusiasts hope they'd have a cheaper alternative to Intel's brawny Broadwell-E-based CPUs. And while it's fair to say the Ryzen launch went well for AMD in comparisons of pricing and professional application performance, gaming didn't paint the processor in a very good light at all.

We are always willing to make some concessions in the name of value, so Ryzen doesn't have to beat Intel's offerings across the board. It just needs to be competitive. Where that line exists for you is completely subjective. But for many, Ryzen’s frame rates are too low, even in light of its attractive pricing. And if gaming is the primary purpose for your PC, it's hard to ignore faster and cheaper Kaby Lake-based Core i7s and i5s that serve up better results in many popular games.

The Zen architecture employs a four-core CCX (CPU Complex) building block. AMD outfits each CCX with a 16-way associative 8MB L3 cache split into four slices; each core in the CCX accesses this L3 with the same average latency. Two CCXes come together to create an eight-core Ryzen 7 processor (image below), and they communicate via AMD’s Infinity Fabric interconnect. Data that traverses the void between CCXes incurs increased latency, so it's ideal to avoid the trip altogether if possible.

Unfortunately, threads migrate between the CPU Complexes, thus suffering cache misses on the local CCX's L3. Threads might also communicate with other threads (and their data) running on the CCX next door, again adding latency and chipping away at overall performance.

AMD noted in a recent blog post that most games aren’t optimized for its implementation of simultaneous multi-threading, which is particularly painful due to Ryzen’s core advantage. In fact, we’ve found that disabling SMT actually improves the chip's performance in games like Ashes of the Singularity, Arma 3, Battlefield 1, and The Division.

Ryzen represents AMD's first attempt at an SMT technology, so teething pains on the application side are understandable. Two game developers have come forward and voiced their intention to support AMD’s implementation in future updates, and AMD says it seeded the industry with 300 developer kits to jump-start the optimization effort. There are thousands of games, though. While many existing titles probably won't receive patches written with AMD in mind, we do hope that newer titles incorporate the code needed to run more smoothly.

According to AMD, this problem doesn't relate to the Windows scheduler. Normally we'd say that's a good thing, since it doesn't depend on Microsoft to fix. But if the issue was tied to the operating system, a single update could optimize for AMD's processors, similar to what we saw with Bulldozer in the Windows 8 days. Instead, we have to look out for improvements one application at a time.

AMD Ryzen 7 1800X

Ryzen 7 1700X

Ryzen 7 1700

AMD also points out that Ryzen is more competitive at 3840x2160 than at lower resolutions, such as 1920x1080. Obviously, gaming at higher resolutions shifts the bottleneck over to your GPU. So while AMD's observation is true, it isn't indicative of better processor performance, but rather the architecture's weakness hidden behind a slammed GPU. Many of us use our CPUs for several years, and as we swap out for faster graphics cards, the bottleneck will start swinging back to host processing. In many ways, today’s 4K is tomorrow's QHD.

The Socket AM4 ecosystem holds great promise, but our experience with the top motherboard manufacturers (and indeed their experience with AMD's chipsets) has been less than ideal. We've received a flurry of updates in the days leading up to and following Ryzen 7's launch. In some cases, new firmware improves performance. In others, the fix shines light on the underlying issues. General platform instability aside, we get the sense there wasn't enough preparation pre-launch, and AMD's partners are scrambling now as a result. But there's hope that things will get better. AMD recently announced it's working on an updated power profile to better accommodate normal desktop usage patterns (more on that in a bit).

In the meantime, we want to better understand the state of gaming with a Ryzen CPU. Today's feature includes a number of popular titles and all three Ryzen CPUs. Though we're still working on our reviews of the Ryzen 7 1700X and 1700, digging deeper on gaming, specifically, was our top priority for follow-up after publishing the AMD Ryzen 7 1800X CPU Review.

MORE: Best CPUs

Performance Power Profiles and Test Platforms

Game Testing Particulars

Due to time constraints, our original review only included five games and the Ryzen 7 1800X. This time around, we're expanding to include 11 games and all three Ryzen 7 processors. At launch time, AMD gave us a list of titles that respond favorably to its new architecture, including Sniper Elite 4 (DX12), Battlefield 1 (DX12), The Division, Star Wars: Battlefront, Assassin's Creed: Syndicate, Battlefield Hardline, Overwatch, Witcher 3, and Dead Rising. We added Battlefield 1 and The Division to our line-up for this round of testing.

AMD recommends disabling the HPET (High Precision Event Timer) and using Windows' "High performance" power profile to improve gaming results. We can confirm that both adjustments do confer gains. The High performance profile, in particular, offers the biggest boost. AMD adds:

Core Parking Off: Idle CPU cores are instantaneously available for thread scheduling. In contrast, the Balanced plan aggressively places idle CPU cores into low power states. This can cause additional latency when un-parking cores to accommodate varying loads.Fast frequency change: The AMD Ryzen processor can alter its voltage and frequency states in the 1ms intervals natively supported by the “Zen” architecture. In contrast, the Balanced plan may take longer for voltage and frequency (V/f) changes due to software participation in power state changes

AMD also announced that it will provide an update in the April time frame that adjusts parameters for the Balanced profile to increase performance.

In an effort to give Ryzen the most favorable conditions possible, we test AMD and Intel CPUs alike using the High performance power plan, and with the HPET disabled. We tested the CPUs with the stock clock settings. In our original coverage, we included numbers with SMT disabled. This time around, however, it's turned on. Enthusiasts should not be expected to toggle back and forth depending on the application they're running.

Test Platforms

We're using the same hardware you saw in our Ryzen 7 1800X review, except that we switched over to MSI's X370 XPower Gaming Titanium motherboard. The rapidly-evolving nature of the Ryzen ecosystem (and by that we mean incessant firmware updates) means that our gaming results are only representative of today's test environment; it may change in the future.

Systems	AMD 1Ryzen 7 1800X, 1700X, 1700MSI X370 XPower Gaming Titanium2x Corsair Vengeance LPX DDR4-2666AMD 2AMD FX-8350MSI 970 Gaming2x Kingston HyperX DDR3 2133Intel 1Intel Core i7-7700KMSI Z270 Gaming M72x Corsair Vengeance LPX DDR4-2666Intel 2Core i7-6900KASRock X99 Extreme44x Crucial DDR4 2400All1TB Samsung PM863SilverStone ST1500, 1500WWindows 10 Pro (All Updates) Version 1607
GPU	EVGA GeForce GTX 1080 FE
Cooling	Corsair H100iv2Noctua NH-U12S SE-AM4Arctic MX-4

MORE: Best Graphics Cards

MORE: How Well Do Workstation Graphics Cards Play Games?

3DMark Physics, Ashes of the Singularity, Battlefield 1 & 4

3DMark Physics

3DMark's DX11 physics and DX12 CPU tests give us a good (albeit synthetic) measure of processing resources available to a game engine.

The Time Spy metric illustrates DX12-based CPU scaling based on its demanding physics simulation, occlusion culling, and procedural generation operations. This test shows Ryzen 7 1800X with a healthy lead over Intel's Core i7-6900K, while Ryzen 7 1700X and 1700 outpace the Kaby Lake CPUs with lower core counts.

Futuremark's DX11 Fire Strike benchmark runs 32 parallel soft and rigid body physics simulations that tax the processor specifically. We see the same trends emerge in this benchmark, though Ryzen 7 1700 curiously beats the 1700X, even after repeated tests.

The system initiates thousands of draw calls per frame rendered, so 3DMark's API Overhead metric is interesting as well. Intel's Core i7-6900K establishes a healthy lead with DX12 draw calls at 1920x1080, but provides fewer DX11 single-threaded draw calls during the test.

Ashes of the Singularity

Ashes of the Singularity is notoriously CPU-bound, responding well to higher core counts and clock rate. But the Ryzen CPUs lag behind Intel's Broadwell-E- and Kaby Lake-based chips by a significant margin. Even the four-core, non-Hyper-Threaded Core i5-7600K beats the 16-thread Ryzen 7s.

We noticed a pronounced speed-up in our Ryzen 7 1800X review when we disabled SMT on the AMD platform, so it's clear that Ashes isn't utilizing this architecture to its fullest. Oxide Games released a statement saying that it plans to optimize for Zen, so a performance improvement is expected in the future. We didn't see any sort of time commitment, though.

Benchmarks run at 2560x1440 give us a look at high-resolution gaming without hitting the GPU bottleneck imposed by 4K. And while the frame rates do drop a bit, Core i7-6900K continues dominating with its eight-core, 16-thread configuration. AMD's FX-8350 suffers the highest frame time measurements during the test run, and the Core i5-7600K experiences more frame time variance than the Ryzen 7 processors.

Battlefield 1

We cranked Battlefield 1's quality settings to Ultra to represent a normal gaming experience with a beefy GPU, then we repetitively took the armor-laden walk at the opening scene of O La Vittoria. Intel CPUs take the lead again, but there is less separation between them and the Ryzen 7s.

There is also little differentiation between the three Ryzen SKUs; the purportedly value-oriented 1700 lags the $500 1800X by 1.6 FPS on average. All of the Ryzen 7s seem to offer similar overclocking headroom, so the 1700 might stand out as a good selection due to its $330 price point. Then again, a Core i5-7600K sells for $240 and has much more room to overclock.

Although the game engine rewards Intel's Core i7-6900K with a lead over the rest of the test pool, the CPUs fall into the same finishing order. AMD's Ryzen processors trail Intel's chips, but we notice less separation between the three Ryzen 7 SKUs. Incidentally, all of these CPUs provide a smooth gaming experience, though the FX-8350 predictably demonstrates the most frame time variance.

Battlefield 4

Our Battlefield 4 tests consist of a bumpy jeep ride through hostile territory at the beginning of the Tashgar level. This benchmark appears wholly graphics-bound (there's an average of 1.3 FPS separating the CPUs we're evaluating).

Due to the GPU-imposed ceiling, we observe similar frame time variance for all of the processors, except for AMD's aging FX-8350.

The Ryzen processors effectively tie Intel's Core i5-7600K in this test. Looks like there's truth to AMD's claim that Ryzen's gaming performance is best when you're wholly GPU-bound.

Civilization VI AI & Graphics Test, Deus Ex: Mankind Divided, GTA V

Civilization VI AI Test

Civilization VI's AI benchmark measures the amount of computational horsepower available to the game engine during the workload.

The Core i5-7600K takes a minor lead over the i7-7700K, though that outcome is really too close to call definitively. Perhaps Hyper-Threading is hurting the Core i7 in this case, which wouldn't bode well for AMD's implementation of SMT. Sure enough, the Ryzen processors show behind Intel's hardware, neatly falling in order of 1800X, 1700X, and 1700.

We also recorded frame rate through the AI test's first 60 seconds, yielding another look at how the faster CPUs drive this turn-based metric at a quicker pace.

Civilization VI Graphics Test

Shifting over to Civilization VI's graphics test shows the Core i7-6900K taking a lead with its eight cores, though it's trailed closely by the higher-clocked Core i7-7700K. The Core i5-7600K's four physical cores don't hold up as well to the 8C/16T Ryzen chips, so it lands just under AMD's two fastest models. We also notice more variation between the 1700 and 1700X than we recorded in other tests.

Bumping the resolution up a notch doesn't change the finishing order. However, we see the Core i7-6900K average a slightly higher frame rate. The game is obviously CPU bound, so 2560x1440 doesn't give us anything new to report.

Deus Ex: Mankind Divided

Deus Ex: Mankind Divided responds well to the Ryzen family; all three models manage to beat Intel's fastest contenders. You'd better believe we're looking for a plausible technical explanation.

There's a distinct drop-off between AMD's Ryzen CPUs and the Intel chips. The Core i7-6900K trails Ryzen 7 1800X by 7.2 FPS on average, which is a comparatively large win. Our "frame rate over benchmark run" chart shows how both processor vendors segment out into tight groupings.

The same pattern emerges at 2560x1440, though with lower frame rates across the board. Ryzen 7 1800X leads the Core i7-6900K by 3.7 FPS, shrinking the gap (this time to Intel's benefit) as resolution goes up.

Grand Theft Auto V

Grand Theft Auto V flips the performance story around again. We dialed the graphics settings up as high as they'd go in an effort to mimic a real-world experience on a GeForce GTX 1080.

The Core i7-7700K lands in first place by averaging 91.1 FPS, while Ryzen 7 1800X achieves 75.0 FPS (a 17% difference). Separately, the Ryzen processors scale better in GTA than they do in our other game tests. A 7% delta exists in moving from Ryzen 7 1700 to 1800X.

Intel's Core i5-7600K beats the Ryzen line-up at 2560x1440 as well, boasting a surprisingly good 58 FPS minimum frame rate. But it's the Core i7-7700K that outperforms everything else.

Hitman (2016), Metro: Last Light Redux, Middle-earth: Shadow of Mordor

Hitman (2016)

The Core i7-7700K and -6900K offer the best performance though Hitman's benchmark sequence at 1920x1080, jumping out to an ~18 FPS lead over the Ryzen processors.

Interestingly, the Core i5-7600K lands behind AMD's Ryzen trio in our average frame rate measurement, but demonstrates a better minimum frame rate.

A GPU bottleneck nudges all of the CPUs closer together, helping Ryzen 7 1800X come closer to its Intel competition at 2560x1440.

This time around, Core i5-7600K sneaks in above the Ryzen 7 1700 when we compare average frame rates. But the Kaby Lake-based CPU continues to post a higher minimum frame rate than the Ryzen family.

Metro: Last Light Redux

The Ryzen and Core processors settle into two distinct categories during the Metro: Last Light Redux benchmark, though the groups are distinguished by just 2 FPS.

We run into the graphics wall at 2560x1440. The top two Intel CPUs eke out a narrow win, but there's just not much room for host processors to stretch their legs under these conditions.

Middle-earth: Shadow of Mordor

It looks like a graphics bottleneck of some sort limits performance, even down to 1920x1080, in the Shadow of Mordor test. Only AMD's FX-8350 fares notably worse than the rest of the pack.

The Ryzen processors make a resurgence when we bump this game's resolution up to 2560x1440. All three models sweep past Intel's Core CPUs. Shadow of Mordor is one of the few titles that demonstrated significant changes after shifting from FHD to QHD, so we retested multiple times to verify accuracy.

Project CARS, Rise of the Tomb Raider, And The Division

Project CARS

Slightly Mad Studios designed the Project CARS game engine specifically to promote parallelism by breaking tasks down into smaller chunks across available resources. The end result is a sophisticated engine that scales well with additional CPU cores and higher clock rates.

The prevalence of Kaby Lake at the top of this chart tells us that the game responds well to high IPC throughput and clock rates. After all, even the Core i5-7600K's four physical cores outpace Core i7-6900K's 8C/16T configuration (despite dropping to a lower minimum frame rate).

AMD's Ryzen CPUs line up predictably, given their frequencies.

As we've come to expect from Project CARS, we don't notice much of a performance drop as we shift to 2560x1440. This serves to underline the game's CPU-bound nature.

Rise of the Tomb Raider

Just as we saw Deus Ex split into two distinct performance groups, so to does Rise of the Tomb Raider draw attention to the different architectures being tested. This time, however, it's Kaby Lake/Broadwell-E in the lead. Clearly, there's work to be done optimizing for Ryzen.

Ryzen 7 closes the gap at 2560x1440, suggesting a more graphics-bound scenario.

The Division

Tom Clancy's The Division is graphically demanding all the way down to 1920x1080, allowing Ryzen 7 1800X to climb in ahead of Core i7-6900K. Both Kaby Lake-based CPUs land in first and second place, but their performance advantage is imperceptible. The variable we cannot ignore here is price: the Core i5-7600K, especially, is a much more affordable solution if you're gaming-focused.

The Core i7-6900K lands back in the lead as we jump to 2560x1440. But at what cost? Curiously, Intel's Core i7-7700K tumbles three positions. We tested this condition several times to verify the result, but it does stand out as a possible outlier. The Ryzen 7 1800X matches the Core i5-7600K, though it does suffer a lower minimum frame rate during the test.

Conclusion

As we noted in the AMD Ryzen 7 1800X CPU Review, gaming is a mixed bag for AMD. Some titles respond well to the new architecture, while most others still require optimization of some sort. Where those optimizations will focus remains a question mark.

Our 3DMark physics, CPU, and draw call tests indicate that Ryzen is plenty powerful when the game engine and API can utilize its resources effectively. Unfortunately, this isn't what we see consistently in the real world. We don’t expect the Ryzen 7 processors to beat Intel’s Core i7-7700K and Core i5-5600K in most games; Kaby Lake does enjoy an IPC throughput advantage and higher frequencies, after all. However, we recorded abnormally low performance from Ryzen in several titles, such as Rise of the Tomb Raider, Project CARS, Hitman, Civilization VI, and Ashes of the Singularity. Conversely, Ryzen competes more readily in Tom Clancy’s The Division, Metro: Last Light Redux, and Battlefield 4, particularly when the workload is graphics-bound. And we did record real victories for AMD, too. Ryzen scored big in Deus Ex: Mankind Divided and Middle Earth: Shadow of Mordor at 2560x1440. What we're working on now is determining how those titles are unique in the way they utilize Ryzen 7.

Between the good and the bad, we have a slightly better idea of what to expect from games that respond well to Ryzen and those requiring some work.

The Ryzen processors sell for a much lower price than Intel's Broadwell-E-based CPUs, earning them solid marks in value comparisons using workstation-class software. This same value story isn't applicable in games, though, where much cheaper Intel Kaby Lake-based CPUs are typically as fast or faster. The $240 Core i5-7600K beat all three Ryzen CPUs in several games, and the $350 Core i7-7700K nearly swept the table.

To complicate matters, AMD and its motherboard partners still have a lot of work left to get Socket AM4 platforms running stably. But it's happening as fast as firmware updates can be finalized. The company's recent announcement that a new power plan is forthcoming shows promise as well, though this probably should have been finalized before launch. And it's not clear if a modified profile simply optimizes for Ryzen's idiosyncrasies at the expense of, say, power consumption, heat, and noise.

While we're happy to have Ryzen doing serious battle with Broadwell-E for the hearts and minds of content creators, coders, and other professionals, our assessment of the gaming space suggests Ryzen 7 isn't currently the processor family to beat. Intel's Kaby Lake-based CPUs are definitely less expensive, and typically faster. Perhaps a rash of patches will change the way existing games treat Ryzen. Maybe developers are already rolling optimizations into their upcoming titles. And we definitely have high expectations for the Ryzen 5 and 3 line-ups, which should level AMD's strengths against Intel's mainstream processors using more evenly matched prices.

MORE: Best CPUs

AMD Ryzen Lineup Sold Out On Amazon (Again)

Nathaniel Mott — Thu, 02 Mar 2017 17:05:00 +0000

AMD Ryzen is going, going, gone from Amazon's website. The company previously sold all its pre-order stock of the 1800X--which we reviewed--before restocking the processor. Now the 1800X is once again sold out on Amazon, and so are its companions, the 1700 and 1700X.

Other retailers have also failed to keep Ryzen in stock for its official launch today. Best Buy was supposed to stock the lineup but merely offers pre-built rigs made with the processors; Newegg still lists the 1800X as out of stock; and Fry's says the entire Ryzen lineup is sold out. With the exception of Newegg, which said it expects to have more 1800X units around March 10, none have said when they expect to stock back up on the new CPUs.

Ryzen's popularity comes as no surprise. AMD opened pre-orders on February 22, the same day it finally revealed more information about the lineup and its pricing, and the new chips quickly rose to the top of Amazon's list of best-selling CPUs. (The top spots are mostly taken up by Intel now, though Amazon's fluctuating stock of Ryzen probably limited the series' ability to climb back up the rankings.) The hype train has gone full steam ahead.

Things hit a bump when we discovered that Ryzen motherboards could "pose a potential problem with some of the announced CPU cooling solutions" because of "the backplates designed to secure the heatsink on the processors." We confirmed the issue in our review of the 1800X:

One of the hardware components that’s seen some changes is the AM4 motherboard’s backplate. It’s different in two very significant ways: The distance between the wholes has changed, and the threaded pins are longer. Even though AMD seems to have informed manufacturers of the first change, they apparently forgot about the second one. We asked the manufacturers, and they confirmed that this is what happened. This means that if the original motherboard backplate’s used in conjunction with longer screws, then the cooler might not be held against the processor tightly enough.

The fix is simple--adding the necessary millimeter via thick ring washers or the proper nuts--but it's something to be aware of as you think about how Ryzen might fit into your rig. Assuming, that is, that you're able to buy the processor you want even though they all keep selling out left and right.

AMD Ryzen 7 1800X CPU Review

palcorn@outlook.com (Paul Alcorn) — Thu, 02 Mar 2017 14:00:01 +0000

The Ryzen Debut

The conundrum AMD currently faces started when it launched Bulldozer to lackluster reviews back in 2011. The following years found it trying to right the ship with Piledriver, Steamroller, and Excavator. But it’s safe to say the company's host processing portfolio never regained its lost luster. Meanwhile, Intel dominated the mobile, desktop, and server markets with a seemingly insurmountable performance built on the excellent Sandy Bridge design and an unrelenting cadence of incremental improvements.

Heading into today's review, nobody can argue the fact that AMD is far behind by comparison. Wouldn't that make a comeback all the more impressive, though?

We started seeing Zen micro-architecture teasers last year. Company representatives told us its next generation would usher in incredible performance gains, matching or surpassing Intel's best efforts on multiple fronts. Zen also promised to revitalize an aging platform through Socket AM4 and new core logic. And AMD says it has a clear path forward planned for future versions of Zen.

At some point, though, the rubber has to meet the road. A good first step was taking aim at a competitor. Intel's $1000+ Core i7-6900K seemed like an ambitious choice, but early hand-picked benchmark results made AMD's eight-core engineering samples look formidable. Then, announcing that the flagship model would sell for less than half of of the -6900K's price sent the masses into a frenzy. Most online vendors even sold out of their Ryzen 7 1800X allocation during pre-sales based on little more than AMD's own endorsement.

Now it's time for Ryzen to stand up on its own accord and show us what it can do in the real-world. We have several Ryzen SKUs in-house, spread across multiple Tom's Hardware labs. We've identified a number of unexpected results that bear continued investigation. We'll continue updating our coverage as answers materialize. But we want to start putting our findings in front of you so enthusiasts can make more informed buying decisions in the face of general availability, which begins today.

Finding Zen

Four years ago, AMD began its work on the Zen core, which is its first clean-sheet architecture since Bulldozer. AMD’s initial objective was to transition from the 28nm process used for its modern APUs to GloablFoundries’ 14nm FinFET node, which offers increased performance and density within a similar power envelope. The company also set an ambitious goal to increase instruction-per-clock throughput by 40% over Excavator through a series of design choices that significantly boost performance. Notably, AMD deployed a new architecture and a lithography shrink simultaneously, which is a daunting challenge.

Last year, we published Everything Zen: AMD Presents New Microarchitecture At HotChips. In that story, we stepped through the composition of Zen, from front to back, right up to describing the CPU complex (CCX) responsible for housing four execution cores, each core's 512KB L2 cache, and 8MB of shared L3 cache. If you aren't already familiar with Zen and how it differentiates from prior-gen designs, check that piece out.

Moving forward, you need to know that the Zen core is Ryzen's fundamental building block. All three SKUs we're introducing today employ two quad-core CCXes, adding up to 4.8 billion transistors across the entire die. The company says its Infinity Fabric connects the CPU complexes, but remains shy about how that's a quantifiable benefit.

As we established in our architectural deep-dive, AMD also arms Zen with simultaneous multi-threading support, allowing each physical core to operate on two threads in parallel. In theory, this improves the utilization of available hardware resources. A lot of our real-world benchmarks bear that out with phenomenal performance gains. But other workloads expose teething pains we're still trying to diagnose.

How about the 40% IPC improvement goal AMD set for itself? Well, after factoring in the new micro-op cache (bypassing the L1 and L2 for frequently-accessed micro-ops), the better branch prediction engine, the 1.75x-larger instruction scheduler window, and faster caches, the company cites a +52% final tally compared to Excavator. Naturally, we have our own single-threaded workloads to run and will gladly make comparisons using CPUs from our lab.

The Ryzen 7 Line-Up

AMD is splitting its newest CPUs into the eight-core Ryzen 7 family, a six-core Ryzen 5 series, and the quad-core Ryzen 3 line-up. Only the Ryzen 7 SKUs are shipping today, but it's easy to see that AMD is targeting Intel's Core i7, i5, and i3 portfolios with a similar naming scheme.

Aside from Intel's eight- and 10-core i7s, the Ryzen 7s deliver higher core counts across the board. The AMD CPUs also blow Intel's Broadwell-E prices out of the water, though four-core/eight-thread Kaby Lake is generally cheaper (albeit with half as many cores).

It's not entirely clear what features AMD plans to roll out across the Ryzen 5 and 3 CPUs, but we do know 7s sport the SenseMI suite. We'll go into more depth on SenseMI shortly. What's important here, though, is that SKUs with an X suffix include the eXtended Frequency Range capability. XFR automatically increases clock rate beyond the factory-set Precision Boost frequency if you provide additional thermal headroom with an aggressive cooler. This extra bit of speed applies to two of the chip's cores.

AMD Ryzen 7 1800X

Ryzen 7 1700X

Ryzen 7 1700

Ryzen 7 is solely a host processor, devoid of integrated graphics. All three models debuting today drop into the Socket AM4 interface, include eight physical cores, and boast 16MB of shared L3 cache. They also sport unlocked ratio multipliers, though you'll need a motherboard based on the X370, B350, or X300 chipsets to overclock.

The Ryzen 7 1800X features a 3.6 GHz base frequency able to hit 4 GHz under lightly-threaded workloads via Precision Boost technology. Both of those specifications are higher than Intel's eight-core Core i7-6900K. With Precision Boost enabled, all of the 1800X’s cores can operate at 3.7 GHz. And with enough thermal headroom, two cores jump as high as 4.1 GHz.

Perhaps surprisingly, given the comparisons to Intel's 140W Broadwell-E behemoths, 1800X bears a 95W TDP. If that's not enough to make you believe AMD has a new lease on life, the $500 price tag should excite professional content creators especially. Of course, if you don't regularly find yourself running heavily-threaded tasks, Ryzen 7's value isn't as pronounced. After all, Intel's Kaby Lake-based Core i5s and i7s offer solid performance and generally sell for less than the top-end AMD chips. Ryzen 7's performance in our benchmark suite will have to justify the premium.

The 95W Ryzen 7 1700X's clock rates drop to 3.4 GHz base and 3.8 GHz under Precision Boost. Those frequencies compare favorably against the 140W Core i7-6800K, which tops out at 3.6 GHz in lightly-threaded tasks and only comes equipped with six cores. Worse, Intel charges $425 for the -6800K, while AMD is introducing Ryzen 7 1700X at $400. The Core i7-7700K also becomes relevant at this point, with its $350 price tag.

AMD's Ryzen 7 1700 has a 65W TDP, making it the lowest-power eight-core desktop CPU available. A 3 GHz base clock rate and 3.7 Precision Boost ceiling are significantly lower than Intel's 91W Core i7-7700K. However, the company compensates with twice as many physical cores and a comparable price tag.

Ryzen Memory Support	MHz
Dual-Channel/Dual-Rank/Four-DIMM	1866
Dual-Channel/Single-Rank/Four-DIMM	2133
Dual-Channel/Dual-Rank/Two-DIMM	2400
Dual-Channel/Single-Rank/Two-DIMM	2677

The six-core/12-thread Ryzen 5 family should surface in Q2, and include at least two models. The Ryzen 5 1600X will feature a 3.6 GHz base and 4 GHz Precision Boost ceiling, while the 1500X is expected to start at 3.5 GHz and ramp up to 3.7 GHz in lightly-threaded workloads. AMD hasn't shared cache configurations yet for those models. Ryzen 3s are also in the queue, though those aren't expected until the second half of 2017.

AMD geared its pricing structure to target the 99% of enthusiasts it says buy CPUs priced under $500. If Ryzen 7 is successful, the stage is set for even more disruption in the mid-range and low-end segments as well. So, does Ryzen begin its life on stronger footing than Bulldozer? Let's find out.

MORE: Best CPUs

MORE: Intel & AMD CPU Comparisons

MORE: Everything Zen: AMD Presents New Microarchitecture At HotChips

MORE: Intel Kaby Lake Core i7-7700K, i7-7700, i5-7600K, i5-7600 Review

MORE: Broadwell-E: Intel Core i7-6950X, 6900K, 6850K & 6800K Review

AMD SenseMI Suite & XFR

AMD's SenseMI feature suite consists of five key technologies that allow Ryzen 7 processors to adjust performance and power consumption parameters in real time.

Pure Power

According to AMD, its Zen architecture employs an array of 1000 sensors accurate to 1mA, 1mV, and 1°C. The Pure Power feature monitors these temperatures, voltages, and currents, enabling real-time adjustments based on decisions made by what company representatives describe as learning algorithms.

Those sensors feed telemetry data across the Infinity Fabric loop to the Infinity System Management Unit at 1ms intervals. The management unit analyzes that data and issues commands across the fabric to adjust voltage and frequency settings for optimal performance. AMD also notes this functionality helps manage its speculative cache features and AI-based branch prediction.

Each piece of silicon is unique, and AMD points out that its algorithms allow the processor to optimize itself based upon its own characteristics. Notably, other semiconductor vendors employ a similar technique to control the power consumption of their processors dynamically.

Precision Boost

Precision Boost adjusts the power/performance curve to optimal settings for the operating environment, much like Intel's Turbo Boost, based on information derived from Pure Power. The algorithms that control Precision Boost facilitate changes in 25 MHz steps, which is of course more granular than Turbo Boost's 100 MHz increments.

On the Ryzen 7 1800X, for example, Precision Boost increases the 3.6 GHz base frequency to 3.7 GHz across all cores, and can push two cores up to 4 GHz. This is important: whereas Intel's Turbo Boost technology varies clock rate based on the number of active cores, Precision Boost draws that distinction between two active cores and anything in excess of two cores, at which point maximum frequency drops to the all-cores number. What you end up with is a nice speed-up in lightly-threaded tasks, but less benefit than Turbo Boost as soon as a third core spins up.

XFR (eXtended Frequency Range)

AMD's eXtended Frequency Range feature allows the processor to dynamically adjust its clock rate above the stock and Precision Boost clocks based on available thermal headroom.

XFR automatically increases the Ryzen 7 1800X's Precision Boost ceiling by 100 MHz if your cooling solution can keep the CPU running below a certain threshold. We tested with Noctua's NH-U112S SE-AM4 air cooler and Corsair's H100i v2 closed-loop liquid cooler, and both perform well enough to engage XFR, allowing a 4.1 GHz peak clock rate. Unfortunately, we don't have AMD's stock heat sink/fan, so we don't know if it's beefy enough to make XFR work.

AMD claims this feature scales with air, water, and LN2 cooling, but doesn't specify if there's a maximum frequency with LN2.

Interestingly, the Precision Boost and XFR features are intertwined on our Asus Crosshair VI Hero motherboard. If you enable the "Core Performance Boost" setting, the 1800X triggers both Precision Boost and XFR frequencies based on your workload. But you cannot disable either feature independent of the other.

Neural Net Prediction & Smart Prefetch

AMD claims that its Neural Net Prediction capability describes a built-in neural network able to learn application behavior and pre-load instructions before they're needed, while Smart Prefetch learns access patterns to pre-load data into local cache.

Both features appear to be marketing-speak for AMD's perceptron branch predictor, which debuted in the Jaguar core. Of course, the company likely improved and optimized this technology for Zen, specifically, but it's not magic.

The AM4 Platform

AMD announced its AM4 interface and corresponding chipsets during the Bristol Ridge launch. Previously, core logic was a weak point, as the 9-series chipsets lacked a lot of modern features available from Intel's platforms. Moving forward, all Ryzen CPUs share the same socket. This gives you a number of options for building a Ryzen-, Bristol Ridge-, or future Zen+-based system.

Similar to Intel's platform controller hub architecture, AMD pushes integration with the move to Socket AM4 and tasks its chipset with functions typically associated with southbridges. Meanwhile, it adds more capabilities to the processor die you wouldn't expect to find there. For instance, Ryzen 7 1800X provides four USB 3.1 Gen 1 ports. It also offers 16 lanes of PCIe 3.0 we're guessing most enthusiasts will use for graphics, along with four lanes of second-gen PCIe for SATA 6Gb/s and NVMe storage. In time, we may even find board vendors building platforms based on just the CPU's I/O.

Intel's Core i7-6900K doesn't expose any USB connectivity directly. Rather, it relies on X99 and attached controllers to enable those ports. On the other hand, Broadwell-E packs up to 40 lanes of third-gen PCIe for a lot more flexibility in what you attach. Ryzen's ability to split its 16-lane PCIe 3.0 link into two x8 slots looks a lot more like Intel's mainstream chipset line-up in comparison.

AMD breaks Socket AM4 down into five different chipsets. Enthusiasts will want to pick up X370. B350 targets the mainstream crowd, whereas A320 goes the locked-down route for "essential" builds. For those of you who anticipate deploying APUs in HTPCs and LAN boxes, X300 and A/B300 should address those markets (we're still waiting on their specifications).

Given Tom's Hardware's focus on the enthusiast and mainstream markets, our motherboard reviews will emphasize X370 and B350. The two platforms are differentiated in much the same way AMD positioned the old 990FX and 970: a higher-end chipset gets you extra I/O, mostly. X370 includes four extra USB 3.1 Gen 1 ports, two additional SATA 6Gb/s ports, and two more second-gen PCIe lanes. Moreover, B350 doesn't allow you to split the 16-lane PCIe 3.0 link into two slots like X370 does.

Both Socket AM4 platforms support overclocking out of the box though, in addition to RAID 0/1/10 and two SATAe ports. AMD's documentation notes that SATAe ports can be broken up and re-purposed for additional SATA ports or second-gen PCIe lanes, potentially opening the door to larger storage arrays or additional M.2 ports.

According to AMD, the Socket AM4 interface will carry it through 2020. By then, technologies like DDR5 and fourth-gen PCIe should be prevalent, paving the way for newer platforms.

Overclocking & Test Setup

AMD Ryzen Master

AMD's Ryzen Master software allows you to tune a number of variables, such as the CPU's ratio multiplier, voltage, and memory speeds, from inside the operating system. When you open the utility, it warns you of the perils associated with overclocking and disavows AMD of responsibility should you inadvertently nuke your processor. When you adjust parameters, the CPU automatically switches into OC mode and disables all of the normal thermal and voltage restrictions.

You're able to disable cores in groups of two, which could help you hit more aggressive frequencies. Individual cores cannot be sped up or slowed down. Rather, adjustments apply to all cores.

It's worth noting that the older AMD Overdrive tool is not compatible with Ryzen CPUs.

We prefer overclocking from the motherboard firmware, though. Using simple multiplier and voltage adjustments, we achieved a Prime95-stable 4 GHz clock rate at 1.425V using Asus' Crosshair V Hero (that was with load-line calibration set to Auto). The highest temperature we saw was 82°C during our stress test.

Of course, if you're using a cooling solution less capable than our Corsair H100i v2 at its maximum fan/pump settings, overheating could become a problem. AMD predicts that most customers should see somewhere between 3.9 and 4.1 GHz across all cores, and suggests you stick with a 1.35V ceiling if you want your chip to last. Although core voltages in excess of 1.45V are considered sustainable, they'll have a more pronounced effect on longevity.

Asus notes that pushing memory transfer rates in excess of 2933 MT/s on its Crosshair VI Hero requires memory binned at 3200 MT/s or higher with Samsung A-die ICs. AMD hasn't opened up all of Ryzen's memory sub-timings yet, but we expect the company to open this up soon. Currently, Asus' testing with Hynix A-die ICs tops out at 3000 MT/s, but that may improve with microcode updates.

Ryzen 7 1800X

Intel Core i7-6900K

Ryzen 7 1700X

Measurement System and Methodologies

Splitting our tests between two locations proved difficult this time around, since AMD sent our German lab samples two days before the launch. We wanted to benchmark with retail hardware though, not downclocked engineering samples, so we raced the clock to crank out results.

Our challenge was complicated by AMD's decision to send our labs two different motherboards. Germany received MSI's X370 XPower Gaming Titanium, so all of our workstation, HPC, and power consumption tests were run on that platform. Meanwhile, the U.S. team used Asus' Crosshair VI Hero and the EVGA GeForce GTX 1080 FE for graphics testing.

MSI’s X370 XPower Gaming Titanium, just like Asus’ model, is a flagship motherboard based on AMD's X370 chipset. We stuck with AMD’s recommended presets on both machines, minimizing issues attributable to our dissimilar boards. Our German office used two 8GB Corsair Vengeance LPX DDR4-2666 memory sticks and a custom water-cooling loop.

If you want to know more about how the Tom's Hardware DE system looks and is controlled, check out How We Test Graphics Cards.

To keep things simple, we present the highlights in the following table. The Intel systems are identical to those for the gaming tests.

Test Systems and Measurement Setups
Systems	Germany AMD 1Ryzen 7 1080XMSI RX370 XPower Gaming Titanium2x Corsair Vengeance LPX DDR4-2666Germany AMD 2AMD FX-9590Asus Crosshair Hero V2x Corsair Dominator Platinum DDR3 2133Germany All1x 1 TB Toshiba OCZ RD400 (M.2, System SSD)2x 960GB Toshiba OCZ TR150 (Storage, Images)be quiet Dark Power Pro 11, 850WWindows 10 Pro (All Updates)US AMD 1Ryzen 7 1800XASUS Crosshair VI Hero2x Corsair Vengeance LPX DDR4-2666US AMD 2AMD FX-8350MSI 970 Gaming2x Kingston HyperX DDR3 2133USA Intel 1Intel Core i7-7700KMSI Gaming M72x Corsair Vengeance LPX DDR4-2666USA Intel 2Core i7-6900KASRock Extreme44x Crucial DDR4 2400US All1TB Samsung PM863Silverstone ST1500, 1500WWindows 10 Pro (All Updates) Version 1607
Cooling	Germany- Alphacool Eispumpe VPP755 Pump - Alphacool NexXxoS UT60 Full Copper 240mm- Alphacool Eisblock XPX CPU- 2x be quiet! Silent Wings 3 PWM- Thermal Grizzly Kryonaut US-Corsair H100iv2-Noctua NH-U12S SE-AM4-Arctic MX-4
Case	Lian Li PC-T70 with Expansion Kit and Mods
Power Consumption Measurements	- Contact-free DC Measurement at PCIe Slot (Using a Riser Card) - Contact-free DC Measurement at External Auxiliary Power Supply Cable - Direct Voltage Measurement at Power Supply- 2 x Rohde & Schwarz HMO 3054, 500MHz Digital Multi-Channel Oscilloscope with Storage Function - 4 x Rohde & Schwarz HZO50 Current Probe (1mA - 30A, 100kHz, DC) - 4 x Rohde & Schwarz HZ355 (10:1 Probes, 500MHz) - 1 x Rohde & Schwarz HMC 8012 Digital Multimeter with Storage Function
Thermal Measurements	- 1 x Optris PI640 80Hz Infrared Camera- PI Connect Analysis Software with Profiles
Noise Measurements	- NTI Audio M2211 (with Calibration File)- Steinberg UR12 (with Phantom Power for Microphones)- Creative X7, Smaart v.7- Custom-Made Proprietary Measurement Chamber, 3.5 x 1.8 x 2.2m (L x D x H)- Perpendicular to Center of Noise Source(s), Measurement Distance of 50cm- Noise Level in dB(A) (Slow), Real-time Frequency Analyzer (RTA) - Graphical Frequency Spectrum of Noise

Cache Testing

The Cache Testing Dilemma

The Ryzen reviewer's guide contained an interesting note regarding L1, L2, and L3 cache measurement tools. AMD indicates that AIDA64 and SiSoftware Sandra, both commonly-used tools for cache testing, are "not yet equipped to accurately measure cache performance of the Zen architecture." AMD provided its own internally-measured reference values and noted that it is working with the FinalWire (AIDA) and SiSoftware teams to facilitate accurate Zen cache measurement methodology in the future.

We measured performance with the utilities and achieved similar results for Intel's Core i7-6900K, but we also noticed a large gap between the AMD-provided Ryzen measurements and our test results. Ryzen’s L3 cache latency measured 20 ~ 23ns, which is double the provided value. Due to some of the performance characteristics we noted during our game testing, we also tested with SMT enabled and disabled, but the results fell within expected variation. We also measured a ~10ns memory latency gap in favor of the Intel processor.

Many common utilities write zeros to the cache to measure performance. AMD responded to our inquiries and stated that Intel coalesces incoming "zero" write traffic before it passes it to the cache, which could yield artificially high cache throughput measurements, particularly because those patterns don't exist in real-world usage. In our opinion, changing the access pattern would result in reduced performance measurements for the Intel processors, but it certainly shouldn't boost AMD's cache measurements. AMD responded that the current utilities are also not optimized for Zen's unique architecture, and optimizing the utilities' code paths will expose more performance.

We contacted both SiSoftware and FinalWire to request any updated or beta versions of their utilities that can facilitate accurate testing. There is still collaboration between the company's and AMD to resolve the issue, but unfortunately, we aren't at liberty to discuss the details of those conversations. Until that time, and until the vendors and AMD agree on what's happening, we've determined that drawing any conclusions from these results would be irresponsible, and ultimately misleading. When we have final, verified, and real numbers, we'll provide them.

Ashes of the Singularity & Battlefield 4

A Note On Testing

As mentioned, we encountered lower-than-expected results in many of our game benchmarks. Concerned that the explanation could be related to Windows' handling of P-states, we ran tests on Ryzen under the High Performance and Balanced power profiles as a diagnostic measure. The former is denoted with an HP in our charts. Intel's chips are tested using Windows' Balanced profile, representing the way most of us configure our desktops.

The Zen architecture is AMD's first with simultaneous multi-threading, so we also tested the Ryzen 7 1800X with SMT disabled to flesh out any performance deltas attributable to this feature. Indeed, we observed higher performance with SMT turned off in some titles.

Finally, we normalized the clock rates of our test samples to compare per-clock performance. Game engines and DX11 do not always scale linearly, so our 3.8 GHz numbers merely serve as a reference.

In light of the unexpected results we do present, we have to warn you that some of our data could be attributable to factors like motherboard firmware or early microcode. Then again, these are products AMD and its partners are selling, so that's a risk early adopters assume. In the meantime, we continue running tests and asking questions. Expect a follow-up soon.

Ashes of the Singularity

Ashes of the Singularity scales well with additional cores and frequency. This rewards the Core i7-6900K and its broadside of eight cores and 16 threads.

Surprisingly, Ryzen 7 1800X does not reap the same benefits. We expected much higher performance in this CPU-intensive title from the 1800X due to its similar core configuration and higher clock rates. In the end, Core i7-6900K outperforms the stock Ryzen setup, despite AMD's seemingly favorable specs.

We disabled the 1800X's SMT feature and its average frame rate increased to 70.9 FPS. By comparison, we disabled Hyper-Threading (Intel's equivalent technology) and re-tested the -6900K, yielding a 7 FPS performance loss.

Surprisingly, even the 4C/8T Core i7-7700K outpaced AMD's Ryzen 7 1800X by a significant margin (and that was after trying to provide the 1800X with the most favorable settings possible). We also noted minimal uplift from overclocking the 1800X to an all-core 3.8 GHz. Due to the automatic XFR function, and the fact that we couldn't disable XFR without losing all Precision Boost capabilities, we couldn't test at AMD's base 3.6 GHz.

AMD provided its own benchmark numbers at 4K, creating a graphics-bound situation. We tested at 2560x1440 to reduce this bottleneck, exposing more difference between CPUs. The same strangeness happens at QHD too though, and 1800X trails the field by a quantifiable margin. The 1800X's behavior suggests the game isn't optimized for Zen at this point; Oxide Games released this statement:

“Oxide games is incredibly excited with what we are seeing from the Ryzen CPU. Using our Nitrous game engine, we are working to scale our existing and future game title performance to take full advantage of Ryzen and its eight-core, 16-thread architecture, and the results thus far are impressive. These optimizations are not yet available for Ryzen benchmarking. However, expect updates soon to enhance the performance of games like Ashes of the Singularity on Ryzen CPUs, as well as our future game releases.” - Brad Wardell, CEO Stardock and Oxide

Battlefield 4

We traipse into graphics-bound territory during our 1920x1080-based Battlefield 4 testing. Ryzen 7 1800X in its stock configuration (with the Balanced power profile) is the only modern processor that doesn't average 160 FPS or more. It fares better with the High Performance profile activated, pulling in with the rest of the field if you're willing to leave Windows in this mode.

The FX-8350 trails at its 3.8 GHz setting, though that's a down-clock compared to its 4 GHz base frequency. The similarly-clocked Ryzen 7 1800X provides a notable speed-up compared to the old -8350.

The Ryzen 7 1800X provides the same performance as Intel's Core i7-6900K, pushing the EVGA GeForce GTX 1080 FE into graphics-bound territory. If your favorite titles tend to be GPU-limited, you can save $500 on an eight-core processor that offers a lot of heft in more heavily-threaded applications (or save an additional $175 on a quad-core chip that games really well).

Hitman, Project CARS & Metro: Last Light

Hitman (2016)

Hitman scales well during our benchmark, obviously responding to clock rate and core count. The Ryzen 7 1800X receives a boost when we disable SMT, but it still lags the field under the influence of a slight overclock to 3.8 GHz.

The FX-8350, which experiences severe frame time variance, highlights how far AMD has come in a few short generations on the performance front.

The Core i7-7700K takes the lead with a fixed 3.8 GHz frequency, but Ryzen closes the gap as resolution increases. The 1800X at its stock clock rate shows up well behind Intel's Core i7-6900K, but to keep things in perspective, you get 87% of Broadwell-E's performance for less than half of its price with Ryzen 7 1800X. Conversely, you can opt for the $350 Core i7-7700K and enjoy more performance than AMD's 1800X in many popular titles.

The gap narrows even more between Ryzen and Intel's processors as we shift to 2560x1440. The Core i7-6900K and -7700K are still faster, but the utility question surfaces again: if you specifically need an eight-core processor for content creation, you can go with the 1800X and give up some gaming alacrity, or, if high frame rates are top priority, buy the Core i7-7700K knowing it won't be as fast elsewhere.

Project CARS

Intel's Core i7-7700 takes a big lead during our Project CARS testing, followed by the Core i7-6900K. AMD's Ryzen 7 1800X with a fixed 3.8 GHz frequency slides past the SMT-disabled results, suggesting the game works well with AMD's SMT implementation.

Project CARS delivers similar maximum frame rates, even at a higher resolution, which illustrates just how CPU-constrained the game is. The 1800X without SMT and the 1800X under Microsoft's High Performance power profile swap positions on the chart, but are only separated by 0.1 FPS.

The FX-8350 suffers significant performance variation, while Ryzen 7 1800X provides a smoother experience using the High Performance profile. Intel's Core i7-7700K continues to impress with a beastly lead over the other processors, including the -6900K overclocked slightly.

Metro: Last Light Redux

The Ryzen 7 1800X averages 91+ FPS during the benchmark, and only lags the Core i7-7700K by 2.8 FPS.

The FX-8350 falls to the bottom of our chart, and though it trails the rest of the field by a marginal average frame rate, it experiences much more frame time variance.

In all four configurations, the Intel processors are separated by a scant 0.2 FPS during the graphics-intensive workload. The Ryzen 7 1800X offers nearly the same performance as the leading CPUs.

Results: Desktop & Office

It takes more than 18 hours to run the office, workstation and HPC (High Performance Computing) benchmarks once. Consequently, we limited our tests to a total of four CPUs and two runs each (i.e., at stock clock frequency and at 3.8GHz). We used Microsoft Windows’ performance profile for all of the tests. AMD’s older generation is represented by the 5GH FX-9590, which necessitates water cooling.

We get going with a good old friend: the GDI/GDI+ benchmark. We decided to start with it, because its results will provide context for what’s to come later. This benchmark is used to test two different ways to output 2D objects. These are primarily found in older applications, print output, and, in a modified form, in today’s GUI display. We’re using an Nvidia Quadro P6000, since it represents the best available graphics performance that money can buy. This means that there’s no GPU bottleneck to worry about whatsoever.

We start with directly drawing to an output device. The graphics driver uses the CPU quite heavily for this task, but, for the most part, doesn’t use that many threads. This is due to the demise of true 2D hardware acceleration with the introduction of the unified shader architecture. The Microsoft Windows driver model’s also not exactly conducive to it. It’s interesting to see that, in spite of all of this, AMD’s new Ryzen 7 1800X manages to beat Intel’s Core i7-6900K. Since each individual action has to make its way through the entire system, these results aren’t just representative of processor performance, but also include chipset performance.

Next, we add a memory task by using the single 2D function that’s still hardware-based: copying the graphics output from the place in memory where it was built to the output device. We’re executing the exact same steps, but draw a virtual bitmap and not on the monitor. Only the completed picture is sent to the output device. The result is that the CPU load increases significantly, which, in turn, has a surprisingly large impact on the benchmark results. Without the rest of the system providing a bottleneck, the older FX-9590 manages to beat AMD’s new offerings. Only if its clock frequency is lowered to 3.8GHz does it have to admit defeat. The same goes for Intel’s Core i7-6900K.

Update (this is in reference to some reader questions about these results): Drawing into DIB is a pure 2D workload, and the stock i7-6900K can't reach its highest possible clock rate in this low-priority multi-threaded workload. To be specific, the i7-6900K spreads the workload over its 16 threads, so the processor runs at close-to-idle clock speed, because it only utilizes 20% of many of its cores. The distributed workload also strips the processor of its Turbo Boost capabilities.

By comparison, the -7700K spreads the workload among half the threads, which yields a higher load per core, thus boosting its clock speed. The -6900K at 3.8GHz also experiences a boost because it has 50% more clock speed than its stock configuration.

This test further implies that AMD is able to transition between power states faster than the Intel processors, which means it is tuned well for light workloads.

These findings are of great significance for AutoCAD, which we get to on the next page. Ultimately, AutoCAD draws all of its output in memory first, and then sends it to the monitor.

DTP & Presentation

Adobe’s Creative Cloud makes for an exemplary benchmark, and it certainly makes more sense to use a real-world application than a synthetic benchmark. It tests both single- and multicore performance.

Encoding & Multimedia

AMD’s Ryzen 7 1800X can live up to its full potential with our Handbrake benchmark. No matter what the quality is set to, AMD’s new processor just crunches away.

Compression & Decompression

Crunching lots of numbers in short order is also very useful for file compression. AMD’s Ryzen 7 1800X manages to draw even with Intel’s Core i7-6900K at the same clock frequency. Intel’s Core i7-7700K takes the lead when it comes to decompression due to its higher clock frequency. In the end, packing’s harder than unpacking.

Programming in Python

This application sits right on the boundary to the workstation space, and thus leads us seamlessly to the next page. Apart from the visual task library (e.g., graphically complex charts), which pose problems due to the usual reasons (i.e., difficulties with parallelization), the scientific and engineering fields are where AMD’s Ryzen 7 1800X shines.

Ryzen’s a well-suited CPU for both developers and users due to the fact that many libraries don’t provide an explicit advantage for Intel processors.

Bottom Line

With Ryzen, AMD has produced a processor that’s competitive in many application areas. It certainly doesn’t have to hide from Intel’s current octa-core offerings. The price/performance ratio makes it an interesting choice for commercial applications, unless the application in question specifically demands an Intel CPU.

Results: Workstation

Let's take a look at several common workstation productivity applications.

The AutoCAD results remind us of what we reported on the previous page when we looked at output devices. AMD’s Ryzen 7 1800X ends up exactly where you’d think, based on those findings. There’s no real parallelization, but there’s plenty of system memory and cache usage, which turns out to be a combination that puts its performance squarely in the middling category.

Ryzen’s 3D performance is okay, especially in light of the number of cores not being the priority here. Consequently, the results aren’t surprising.

The Cinebench R15 OpenGL benchmark gives us a preview of what we’re in for if an application has not been optimized for Ryzen. It brings up the rear in spite of noticeable multithreading. Then again, it’s also often more than 30% faster than its predecessor, which is running at 5GHz.

Important NoteMany of the professional applications in the development field have been optimized and compiled for Intel CPUs. This is reflected in the performance numbers. Still, we want to provide a complete performance picture, so we won’t skip them. The results might motivate the application developers to focus more of their efforts on AMD’s Ryzen and Naples so that their users have two options.

One of these professional applications is Solidworks by Dassault Systemès. It’s clear to see across the two composites’ performance numbers that both single- and multithreaded scenarios are affected to varying degrees, in spite of the rendering results having an influence on the overall results as well.

The same general principle applies to Creo 3.0 by PTC. Just like Solidworks, this application’s an important and de-facto standard tool for professional development.

Catia’s older, and it’s not optimized for multithreaded performance. This means that the CPUs are left with using just one single core, which results in a lack of performance for AMD’s newest offering.

Autodesk’s Maya 2013 has the same problem, since graphics output in real time isn’t among the tasks that lend themselves well to parallelization.

Parallelization vs. Single-Core Performance

We start with AMD’s very own favorite example that the company kept showing during the presentations: Cinebench R15. First, we need to note that the only way to get the following numbers is by disabling the AVX instruction set completely. This is to say that Cinebench might be a good example, but it can’t necessarily stand in for all that many other applications when it comes to the AMD Ryzen 7 1800X’s real-world performance. Nevertheless, these results provide us with a good basis of comparison for what’s to come later. Note: For all 3.8 GHz benchmarks we turned XFR off (AMD) and Turbo Core off (Intel). For the single core benchmark we also disabled SMT.

Using just one core at stock clocks, AMD’s Ryzen 7 1800X leads Intel’s Core i7-6900K. Looking at them at the same clock frequency, which we set at 3.8GHz for all tests, provides the best comparison, and here Ryzen 7 1800X @3.8 GHz is a tick slower than the Core i7-6900K @3.8 GHz.

Rendering

Next up, we have two different Blender benchmarks. First, all we do is generate a picture output with a sample size of 200 pixels. AMD’s Ryzen 7 1800X beats Intel’s Core i7-6900K at the same clock frequency, but falls behind at stock clock frequencies.

Then again, this type of rendering is just the final step in a daily workload, so we run the preceding steps in addition to the final rendering for our second benchmark. This includes preview pictures and content creation-related processes.

AMD’s Ryzen is unbeatable at pure rendering tasks, and it does a solid job otherwise.

It’s a neck-and-neck race between AMD’s Ryzen 7 1800X and Intel’s Core i7-6900K once again. At the same clock frequency, Ryzen pulls ahead when using the console version, which doesn’t include any workflow-related tasks, but just the graphics output.

As was the case with Blender, the 3ds Max benchmark distinguishes between several different areas, which don’t just include CPU rendering and, consequently, provide a good idea of what the work processes would feel like to the user. AMD’s Ryzen 7 is at a disadvantage when there’s real-time graphics output via the GPU or when only a few cores are used.

Bottom Line

There might have been a few hiccups here and there, which were usually due to the specific software, but, overall, AMD’s Ryzen 7 1800X seems surprisingly mature and offers solid performance. It’s questionable if it’s a good investment if the application in question has been heavily optimized for Intel processors, but there’s no real drawback to its use as a general-purpose CPU.

Results: Scientific & Engineering Computations And HPC

We’re using the SPECwpc benchmark suite, since it provides a good variety of workstation tasks that use very different types of mathematical computations. These are both highly parallelized and also make heavy use of memory bandwidth, cache, and latencies.

The pre-Euler3D CFD (Computational Fluid Dynamics Benchmark) test did work, but the rest of Rodinia’s benchmarks (i.e., Accelerating Compute-Intensive Applications) didn’t. We’re still showing all of the results below, since the remaining CPU’s results might still be of interest.

LAMMPS stands for Large-scale Atomic/Molecular Massively Parallel Simulator, and it didn’t just run flawlessly, but also really showed off Ryzen’s capabilities. In the resulting head-to-head race, AMD got there first more often than not.

NAMD is a benchmark for high-performance simulation of large biomolecular systems. All individual tests went Ryzen 7 1800X’s way.

The Fastest Fourier Transform in the West, or FFTW, is a popular open-source solution to compute one-, two- and three-dimensional DFTs (Discrete Fourier Transforms). The C library makes heavy use of single-precision AVX these days, which proves to be a terrible thing for the Ryzen 7 1800X.

Convolution is a benchmark for an area of functional analysis. Convolution stands for a mathematical operation on two functions (i.e., f and g) that produces a third function.

The CalculiX benchmark is based on a three-dimensional structural finite element program.

Poisson's Equation is a partial differential equation that allows the modeling of the number of incidences that occur at a constant medium rate independent of each other within a fixed time interval or area. It doesn’t prove to be a problem for AMD’s Ryzen 7 1800X, but the older AMD FX-9590 doesn’t fare well.

Sequential Reweighted Message Passing, or SRMP for short, is an algorithm used to solve discrete energy minimization problems. AMD’s Ryzen 7 1800X runs into trouble, which might be due to the specific benchmark application. It uses a total of eight threads.

The earth’s underground structure can be ascertained based on seismic processing. One of the four basic steps to do so is the Kirchhoff Migration, which is used to build a graphical representation out of the available data. AMD’s Ryzen 7 1800X does well with it.

GNU Octave is a scientific programming language. Octave Forge is a toolbox that’s an important part of it, since it allows additional functions to be added and the application’s functionality to be extended. AMD’s Ryzen 7 1800X does a solid job, but we do have to note that not all of the functions worked without problems and that we had to skip one chart graph altogether.

Bottom Line

Unfortunately, we just didn’t have time to try to find possible solutions to the problems that occurred with GNU Octave and some of the other benchmarks. Otherwise, we would have searched for the root of the problems or tested the software using a different operating system. Alas, two days just aren’t enough time to perform a comprehensive CPU test.

This is even more regrettable in light of the fact that AMD’s Ryzen generally does well with these types of tasks. It would have been nice to know the reasons behind the few times when it didn’t. Overall, we’re left with the impression that AMD’s boiled down a server CPU to the desktop level. This doesn’t have to be a bad thing, though, and it certainly makes us excited for Naples.

Results: Power Consumption And Temperatures

Power Consumption

We measure voltages and currents directly on the motherboard using the existing sensors and calculate the power consumption based on them. To achieve valid results, we take the average of the two-minute measurements for each of our scenarios and use our low-pass filter and analysis software to get rid of any extreme peaks or valleys.

These aggregated results are a lot more telling than extremely short peaks due to the latter’s very brief nature. We took a look at a total of eight different scenarios.

Some of the applications consume more power over time. This is due to the CPU heating up. We found that the difference between a cold and fully warmed up CPU can be up to 3W. If these 3W are interpreted as leakage currents, then this is actually a very good result.

We took the average of the power consumption curves after the CPU had reached its full operating temperature and put the results in a bar graph to provide a summary.

Temperatures

Keeping the processor cool is more important than ever due to AMD’s XFR and related technologies. Lower temperatures have the potential to translate to higher clock frequencies. This is why we used a custom water cooling loop, which was very much needed during the rendering benchmarks with all cores running at 3.8GHz.

The CPU diode that AMD uses should be closest to the Tpackage value, which is the part of the CPU that has to endure the most heat in the long run. Tcore doesn’t matter as much anymore due to the massive areas that the cache takes up these days.

Infrared Video and CPU Temperatures

We decided to make an infrared video and spare no expense. Shooting it also required a degree of bravery, since we had to take off the CPU cooler to record it. We installed the camera so that it was pointed downward directly facing the CPU’s heat spreader. The camera’s distance from the processor was a function of the camera’s focal length and efforts to avoid any reflections caused by the camera’s lens and the metallic surface of the heat spreader.

We also applied a total of six thin layers of special lacquer used in board manufacturing to the heat spreader. This is necessary because the pure metal’s temperatures can’t be measured accurately due to its emissivity changing along with its temperature. After the usual calibration, we recorded the following video.

It’s easy to locate the processor’s die this way. It appears to be long and narrow in shape. The heat spreader’s uniform heat distribution indicates that there’s sufficient contact between the die and the heat spreader, especially in light of the fact that we weren’t able to assert any pressure on the heat spreader while we were recording the video. We recently discovered that the die is soldered to the Integrated Heat Spreader (IHS), which is typically much more efficient than Intel's technique of using thermal paste.

AM4 Kit Incompatibility

One of the hardware components that’s seen some changes is the AM4 motherboard’s backplate. It’s different in two very significant ways: The distance between the holes has changed and the threaded pins are longer. Even though AMD seems to have informed manufacturers of the first change, they apparently forgot about the second one.

We asked the manufacturers, and they confirmed that this is what happened. This means that if the original motherboard backplate’s used in conjunction with longer screws, then the cooler might not be held against the processor tightly enough.

The solution involves adding the needed millimeter by using either thick ring washers or suitable nuts. We’ll keep in touch with the manufacturers, since the problem doesn’t seem to affect all of the kits.

Conclusion

AMD's Ryzen 7 launch represents more than just a new CPU family. For most of our readers, it signals the return of competition to the enthusiast-oriented processor market. And considering the flagship 1800X’s potent cost advantage compared to Intel's Core i7-6900K, the competitor AMD singled out months ago, Ryzen 7 does deliver. It's just not as universally superior as the company wanted everyone to believe.

We come away from today's coverage with a number of questions that couldn't be answered in time for the launch. For instance, we discovered Ryzen's tendency to perform better in games with SMT disabled. Could this be a scheduling issue that might be fixed later? AMD did respond to our concerns, reminding us that Ryzen's implementation is unique, meaning most game engines don't use if efficiently yet. Importantly, the company told us that it doesn’t believe the SMT hiccup occurs at the operating system level, so a software fix could fix performance issues in many titles. At least one game developer (Oxide) stepped forward to back those claims. However, you run the risk that other devs don't spend time updating existing titles.

The evening before launch, AMD sent us a list of games that it says should perform well with Ryzen, including Sniper Elite 4, Battlefield 1, Star Wars: Battlefront, and Overwatch, among others. Many of the titles tend to be heavily threaded, which would lend itself well to Ryzen's high core count. We plan on revisiting some of those. Further, AMD suggests adjusting several different parameters for games that suffer from low performance. It recommends using Windows' High Performance power profile (which also helps Intel CPUs). It also says to disable the HPET (High Precision Event Timer), either in your BIOS or operating system, to gain a 5-8% advantage. Our results already reflect HPET disabled, though. Interestingly, AMD's Ryzen Master software requires HPET to “provide accurate measurements,” so you may find yourself toggling back and forth for the best experience.

It’s hard to recommend the Ryzen 7 1800X over Intel's lower-cost quad-core chips for gaming, especially given the Core i7-7700K's impressive performance. That's not a knock against AMD, specifically. After all, we say the same thing about Intel's own Broadwell-E CPUs. High-end Kaby Lake processors constantly challenge pricier competitors, and the flagship -7700K sells for $350. Even after down-clocking the -7700K to 3.8 GHz, it still beats Ryzen 7 1800X in nearly every game in our suite. Those issues would only be exacerbated on a Ryzen 7 1700X, which operates at lower clock rates.

Conversely, the Ryzen 7 1800X is in its element when you throw professional and scientific workloads at it. It isn't the fastest in every high-end benchmark, but any calculation that factors in value almost assuredly goes AMD's way. For years, Intel has operated with impunity, charging inflated prices for incremental speed-ups. The 1800X’s $500 price tag and competitive performance will no doubt excite power users on a budget. To that end, when we weigh the 1800X’s strong showing in workstation and HPC workloads against its issues with games, we can't help but believe that AMD designed this specific configuration with a datacenter-driven mindset and didn’t optimize it thoroughly for desktops. Much like Intel and Broadwell-E, in fact.

AMD’s Precision Boost technology yields a nice dual-core boost during lightly threaded workloads, but it isn’t as advanced as Intel’s sophisticated multi-core Turbo Boost functionality. XFR is a nice feature that automatically offers improved performance with robust cooling solutions, but most of us only get 100 MHz out of it, so it's hard to call it a compelling advantage. Achieving a 4 GHz overclock was straightforward enough through multiplier and voltage adjustments, and there are plenty of AMD-specific firmware settings we need to explore. More headroom could certainly be available (though the Core i7-7700K is honestly more exciting to overclock if all you care about is higher numbers). On the memory overclocking side, AMD hasn’t opened all of the sub-timings yet, and the Core i7-6900K has a throughput advantage with its quad-channel controller.

Ryzen 7 1800X's aggressive price might help put enough pressure on Intel to compel price cuts on Broadwell-E, but the bigger battle is going to happen when Ryzen 5 and 3 emerge to challenge the competition's more affordable (and difficult to usurp) models. AMD is also bringing its Naples server CPUs forward soon, and with what we’ve seen from the Zen core, that should be an exciting launch.

It's a bummer the Ryzen launch was so clearly rushed. We expected AMD to have a better explanation for its gaming performance, but all of the feedback we received from the company came very last-minute. It's hard to imagine these shortcomings weren't discovered previously and diagnosed more thoroughly. We're happy to put in the time and effort, though. Expect more information as it becomes available.

In the meantime, we would recommend Ryzen 7 1800X for heavily-threaded workloads like rendering and content creation. And while we won't judge a processor on its gaming performance alone, current indications suggest AMD's $500 flagship doesn't beat Core i7-7700K for value in that specific segment.

MORE: Best CPUs