Sign in with
Sign up | Sign in
System Builder Marathon Q3 2014: High-End Performance PC
By ,
1. Can We Build An Even Higher-End $1600 PC?

System Builder Marathon, Q3 2014: The Articles

Here are links to each of the four articles in this quarter’s System Builder Marathon (we’ll update them as each story is published). And remember, these systems are all being given away at the end of the marathon.

To enter the giveaway, please fill out this SurveyGizmo form, and be sure to read the complete rules before entering!

Day 1: The Budget Gaming PC
Day 2: Our Mainstream Enthusiast System
Day 3: The $1600 High-End Build
Day 4: Performance And Value, Dissected

Building upon a new mandate for our lowest-priced machine, my high-end build from last quarter faced a surprisingly tight budget, along with some restrictions that may have appeared odd. Adding a $100 operating system to the $600 machine’s total cost meant limiting Paul to a mere $500 of hardware. Doubling those numbers for the mid-priced machine meant it could have up to $1000 in components and up to $200 in extras, including the operating system. And tripling those numbers meant my top-range configuration included as much as $1500 in critical parts and $300 in extras.

Those add-ons could only include items that wouldn't affect our test scores, such as the Windows license and non-benchmarked storage (an optical drive and/or non-OS storage drive). Don argued that cases should be among those parts, since we weren’t testing the extra lights and front-panel connectors that often push enclosure prices skyward. I made that concession, even though cases do affect cooling and therefore overclocking.

Unable to push unrealistic budgetary compromises into my own build, I stuck to spending $1500 for all hardware (that's three times the base machine) and $1600 for the complete PC, including its operating system. Even though my specs looked great in the Newegg shopping cart and successfully beat my prior effort when it came to giving you better value, I still found enough problems last quarter that I decided to take another run at it for today’s presentation.

The previous machine’s $90 case is gone, which helps increase the cooling budget. The award-winning Ostrog GT replaces it at two-thirds the cost, similar thermal performance, and perhaps a less office-friendly appearance.

Q3 2014 $1600 Performance PC Components
ProcessorIntel Core i7-4790K: 4.0 - 4.4 GHz, Quad-Core, 8 MB Shared L3 Cache$340
GraphicsPowerColor PCS+ AXR9 290X 4GBD5-PPDHE Radeon R9 290X$530
MotherboardMSI Z97 Gaming 5: LGA 1150, Intel Z97 Express$138
MemoryG.Skill Ripjaws X F3-14900CL8D-8GBXM: DDR3-1600 C8, 8 GB (2 x 4 GB) $89
System DrivePlextor M6S PX-256M6S: 256 GB, SATA 6Gb/s SSD$135
PowerEVGA SuperNova 750 B2: 750 W Semi-Modular, ATX12V v2.31, 80 PLUS Bronze$90
CPU CoolerPhanteks PH-TC14PE_BK 140 mm$75
 Platform Cost
$1397
Storage DriveWD Blue WD10EZEX: 1.0 TB, SATA 6Gb/s Hard Drive
$60
OpticalLG GH24NSB0B: 24x DVD±R, 48x CD-R$18
CaseEnermax Ostrog GT ECA3280A-BR$60
 Total Hardware Cost
$1535
OSWindows 8.1 X64 OEM$100
Complete System Price
$1635

Other changes include a cheaper motherboard in the award-winning MSI Z97 Gaming 5, which makes room for better CPU. If you’re clicking those links, you’re probably starting to see a theme. These parts were given value awards in our lowest-priced enthusiast product segments. Only those components that have the biggest impact on benchmarks get the high-end treatment this quarter.

2. CPU, Graphics And Memory

CPU: Intel Core i7-4790K

I rarely get a retail CPU that doesn't overclock well, but that’s exactly what happened last quarter. We have little doubt that Intel sorts its processors to put the best cores within its highest-end SKUs, but somehow I ended up with a dud, even after buying the best LGA 1150-based processor the company sold at the time. I was almost certain to get a better CPU this time.

Read Customer Reviews of Intel's Core i7-4790K CPU


The reason I knew I’d get a better processor is because the factory-set clock rate for this build’s Core i7-4790K is higher than the top overclock I achieved last quarter.

The only reason I can think of that my Core i7-4770K maxed out at 4.2 GHz and 1.28 V is that Intel was already setting aside its best dies for the Core i7-4790K in advance of the new model’s release.

Graphics Card: PowerColor PCS+ AXR9 290X 4GBD5-PPDHE

Some of our readers don’t care for the PowerColor brand, and the company justified their distrust once with a batch of incorrect firmware. That move could have crossed this model off my short list had I not already bought it, and PowerColor might not have received a second chance had I not already had the card in-hand. But I’m now glad this company has a chance to redeem itself.

Read Customer Reviews of PowerColor PCS+ AXR9 290X 4GBD5-PPDHE Graphics Card


Quick to address the firmware issue, PowerColor proved it really would sell the same card that earned our award. That value award was based partly on better-than-reference performance, and partly on an improved temperature-to-noise ratio of its triple-fan cooler. We’re happy that the current run of cards provides the expected performance, and PowerColor even took the time to correspond publicly with those customers who had previously been affected by the bad run of parts.

Memory: 8 GB G.Skill DDR3-1866 CAS 8

Last quarter’s memory worked so well that I couldn’t find a good excuse to change it. This is the same G.Skill memory that I’ve used in most of my recent builds due to a relatively low price and overclocking consistency.

Read Customer Reviews of G.Skill's 8 GB DDR3-1866 CAS 8 RAM


It does look a little different than the memory of some of those builds, but that’s only because G.Skill offers the same ICs and circuit boards at several speeds, under multiple part numbers, and with several different heat spreader styles.

Among the least-expensive of those various iterations, this Ripjaws X F3-14900CL8D-8GBXM dual-channel kit comes with a DDR3-1866 C8 rating that means it’s probably binned a little higher than the similar C9 version.

3. Motherboard, Case And Cooling

Motherboard: MSI Z97 Gaming 5

The Asus motherboard in last quarter’s $1600 PC was a good value back then, but MSI’s Z97 Gaming 5 is a slightly better value today. Prices don’t always fluctuate synchronously.

Read Customer Reviews of MSI's Z97 Gaming 5


Originally recognized with an award for its good value at $160, the same Z97 Gaming 5 sells for less than $150 today (or at least at the time of writing).

Case: Enermax Ostrog GT

Last quarter’s build received a sleek-looking, somewhat quiet case with good cooling performance to match its looks. One of our readers even complained that the case looked a little too upscale to be a good value, even though it was only $90. That won’t be a problem this time.

Read Customer Reviews of Enermax's Ostrog GT


Though somewhat stylish, Enermax’s Ostrog GT looks like a nicely designed, inexpensive gaming case. Assets include its great cooling performance and modest noise isolation. Deficits include a top panel with room for a radiator without fans, or fans without a radiator. I was hoping to use alternative mounting methods to put a radiator and fans up there anyway before I figured out that I wouldn’t be able to afford a closed-loop liquid cooler.

CPU Cooling: Phanteks PH-TC14PE

The war between Noctua and Phanteks has been raging for a long time, with fans of the latter brand often making wild claims about superior performance and a lower price as almost everyone else shrugs. Indeed, our own tests have shown little to no performance difference between the NH-D14 and PH-TC14PE, and every attempt to purchase the more colorful brand for a lower price has been fruitless, until now.

Read Customer Reviews of Phantek's PH-TC14PE


Phantek’s temporary discounts finally aligned with a purchase day, and that discount applied only to the black version. The discount will probably apply to a different color by the time you buy one, but value is colorblind.

4. Power Supply And Storage

PSU: EVGA SuperNova 750 B2

I’m always looking for a good price on a high-quality power supply, and this time I found a deal on EVGA’s 750W SuperNova with an 80 PLUS Bronze rating.

Read Customer Reviews of EVGA's SuperNova 750 B2


For a mere $90, this SuperNova offers 67 A on the +12 V rail in a semi-modular design, where only the mandatory cables are soldered-on. It boasts up to 85% efficiency and a five-year warranty, too. Based on its connector package, the power supply also supports a second Radeon R9 290X graphics card, just in case our winner wants to upgrade.

SSD: Plextor M6S PX-256M6S

Working for a hardware site has some advantages when it comes time to shop. In this case, I was able to get an early look at our best SSD prices for September back in August. That data points to current pricing and the performance from our April review as reasons to purchase the 256 GB M6S.

Read Customer Reviews of Plextor's M6S PX-256M6S


Currently available for only $135, our storage editor though it was a good value even back at its earlier $170 price. “The M6S provides plenty of pep thanks to Marvell's 9188 controller. Even though the processor is limited to just four channels, it keeps pace with Plextor's well-regarded M5 Pro.”

Hard Drive: Western Digital Blue WD10EZEX

We don’t need more than the 256 GB of capacity to run the tests in our benchmark suite, yet the winner of today’s system will probably want a place to store photos and media. Western Digital’s 1 TB Blue drive provides that capacity without busting my budget.

Read Customer Reviews of WD's Blue WD10EZEX


The WD10EZEX has a 7200 RPM spindle to improve response time and a SATA 6Gb/s interface for speedy accesses to 64 MB of cache. But really, I only bought this disk to store information that doesn't need to live on the precious solid-state storage.

Optical Drive: LG GH24NSB0

Many readers claim that they no longer want or need an optical drive, but our operating system shipped to us on optical media. Besides, I'm willing to bet that you have something precious stashed away on a DVD or CD that you’d like to read someday.

Read Customer Reviews of LG's GH24NSB0


The GH24NSB0 also burns DVD media at 24x, just in case you have a few blank discs laying around. Plus, it's great for loading an unrecognized GbE controller driver...since you can't exactly download that until the network connection is up and running. Better still, it's 10% cheaper than the Lite-On drive I used last time.

5. Hardware Installation

The installation guides of our Phanteks CPU cooler and Enermax case reviews provide all of the information a casual builder would need to figure out our rather basic configuration. I do have a couple installation notes, however.

The PH-TC14PE cooler’s fans are larger than its sink height, so that air will escape either above or below the fins. Below is the way to go, I think, since that air helps cool on-board components. In this case, the height of my RAM forced me to raise it slightly.

Better results might have been available by substituting G.Skill’s Ares modules for the chosen Ripjaws X.

The Ostrog GT offers 11.5” of card mounting space, but also features a removable center drive cage to enable the use of longer cards like my chosen PCS+ R9 290X. I relocated the bottom-mounted 2.5” cage atop the lower 3.5” cage to simplify the power cable’s path.

The height of the remaining drive cage was the one thing I didn’t think about when designing a system to support a CrossFire upgrade. An extended fan shroud would prevent the installation of an identical graphics card, though shorter versions of the Radeon R9 290X are still an option.

6. Overclocking

I’ve long observed that the best Haswell-based dies will reach 4.6 GHz at less than 1.30 V, and long heard reports of CPUs that were pushed beyond 1.30 V degrading over time. That is, they slowly become less stable at full speed, which eventually requires an underclock. We don’t want that to happen to our readers.

My strategy is to exceed 4.6 GHz at no more than 1.30 V, though that doesn’t often work. Finding 4.7 GHz stable only at voltages that slightly exceeded 1.30 V, I dialed in 4.6 GHz and started looking for the lowest core voltage that would support it.

I found complete 4.6 GHz stability at 1.24 V. The large voltage difference between 4.6 and 4.7 GHz indicates to me that 4.6 GHz was also a great place to stop. In fact, 1.24 V is less than the peak operational voltage of this CPU at its stock settings.

Scroll up a couple of images and you’ll find that I also pushed G.Skill’s DDR3-1866 C8 to DDR3-2400 C10. That setting was far easier to find, partly because my experience with this memory shows that it reaches top stability at 1.60 V. The other half of this overclocking ease comes from the MSI motherboard’s integrated DRAM timing slope, which appeared to slow memory timings by almost-precisely the amount needed to support the overclock. I was only able to reduce the command rate (from 2T to 1T) and tRAS (from 37 to 28) without causing instability.

I overclocked the GPU with similar ease, but with less enthusiasm. After finding that the default fan slope would support all stable overclocks, I reverted to Catalyst Control Center’s built-in overclocking tool. Though these overclocks won’t impress anyone, any further increases in GPU clock caused visual errors even at low temperatures, and any further increase in memory clock caused black screens.

Though I didn’t gain much by overclocking, I certainly beat last quarter’s build. On the other hand, the previous machine was beset by a CPU that needed 1.28 V to reach only 4.20 GHz and, lacking the advantage of Intel’s newest Devil’s Canyon improvements, couldn’t accept more voltage without throttling.

7. How We Tested

Last quarter’s high-end machine focused on financially smarter choices that might provide similar overall performance to its predecessor, but at two-thirds the price. Even overclocked, though, it struggled to reach the baseline performance of the even earlier $2400 build. Naturally, we have high hopes for this quarter's effort.

Test Hardware Configurations
 Q3 $1600 Performance PCQ2 $1600 Performance PCQ1 $2400 Performance PC
Processor
(Overclock)
Intel Core i7-4790K: 4 - 4.4 GHz, Four Physical Cores
O/C to 4.6 GHz, 1.25 V
Intel Core i7-4770K: 3.5 - 3.9 GHz, Four Physical Cores
O/C to 4.2 GHz, 1.29 V
Intel Core i7-4770K: 3.5 - 3.9 GHz, Four Physical Cores
O/C to 4.50 GHz, 1.25 V
Graphics
(Overclock)
PowerColor Radeon R9 290X: 1050 MHz GPU,  GDDR5-5400
O/C to 1082 MHz, GDDR5-5600
PowerColor Radeon R9 290X: 1050 MHz GPU,  GDDR5-5400
O/C to 1100 MHz, GDDR5-6200
2x EVGA GeForce GTX 780: 902 MHz GPU, GDDR5-6008
O/C to 1059 MHz, GDDR5-6720
Memory
(Overclock)
8 GB G.Skill DDR3-1866 CAS 8-9-9-24, O/C to DDR3-2400 CL 10-12-12-28, 1.60 V8 GB G.Skill DDR3-1866 CAS 8-9-9-24, O/C to DDR3-2133 CL 9-10-10-27, 1.60 V16 GB G.Skill DDR3-1866 CAS 9-10-9-28, O/C to DDR3-2133 CL 9-10-11-10, 1.585 V
Motherboard
(Overclock)
MSI Z97 Gaming 5:
LGA 1150, Intel Z97 Express
Stock 100 MHz BCLK
Asus Z97-A:
LGA 1150, Intel Z97 Express
Stock 100 MHz BCLK
ASRock Z87 Extreme4:
LGA 1150, Intel Z87 Express
Stock 100 MHz BCLK
CaseEnermax Ostrog GTCM Storm Scout 2 AdvancedNZXT Phantom 410
CPU CoolerPhanteks PH-TC14PE 140 mmThermaltake NiC L32Thermaltake CLW0217 Water 2.0 Extreme
StoragePlextor M6S PX-256M6S 256 GB SATA 6Gb/s SSDSamsung MZ-7TE250BW 250 GB SATA 6Gb/s SSDSanDisk SDSSDHP-256G-G2 256 GB SATA 6Gb/s SSD
PowerEVGA SuperNova 750 B2: 750 W, 80 PLUS BronzeRosewill HIVE-750: 750 W Semi-Modular, 80 PLUS BronzeCorsair HX750: 750 W Semi-Modular, 80 PLUS Gold
Software
OSMicrosoft Windows 8 Pro x64Microsoft Windows 8 Pro x64Microsoft Windows 8 Pro x64
GraphicsAMD Catalyst 14.4AMD Catalyst 14.4Nvidia GeForce 335.23
ChipsetIntel INF 9.4.0.1026Intel INF 9.4.0.1026Intel INF 9.4.0.1026

More evidence that last quarter’s CPU was a dud comes from this quarter’s CPU. The stock clock didn’t leave much room for an O/C, but the stock clock was higher than the previous machine’s overclock.

This month’s overclocking dud was the graphics card. Fortunately, its factory overclock provides fairly high performance.

Benchmark Configuration
3D Games
Battlefield 4Version 1.0.0.1, DirectX 11, 100-Sec. Fraps "Tashgar"
Test Set 1: High Quality Preset, No AA, 4x AF, SSAO
Test Set 2: Ultra Quality Preset, 4x MSAA, 16x AF, HBAO
Grid 2Steam Version, In-Game Test
Test Set 1: High Quality Preset, No AA
Test Set 2: Ultra Quality Preset, 8x AA
Metro: Last LightSteam version, Built-In Benchmark, "Frontline" Scene
Test Set 1: DX11, Med Quality, 4x AF, Low Blur, No SSAA, No Tesselation, No PhysX
Test Set 2: DX11, High Quality, 16x AF, Normal Blur, SSAA, Tesselation Normal, No PhysX
Far Cry 3V. 1.04, DirectX 11, 50-sec. Fraps "Amanaki Outpost"
Test Set 1: High Quality, No AA, Standard ATC., SSAO
Test Set 2: Ultra Quality, 4x MSAA, Enhanced ATC, HDAO
Adobe Creative Suite
Adobe After Effects CCVersion 12.0.0.404: Create Video that includes three streams, 210 frames, Render Multiple Frames Simultaneously
Adobe Photoshop CCVersion 14.0 x64: Filter 15.7 MB TIF Image: Radial Blur, Shape Blur, Median, Polar Coordinates
Adobe Premiere Pro CCVersion 7.0.0 (342), 6.61 GB MXF Project to H.264 to H.264 Blu-ray, Output 1920x1080, Maximum Quality
Audio/Video Encoding
iTunesVersion 11.0.4.4 x64: Audio CD (Terminator II SE), 53 minutes, default AAC format 
LAME MP3Version 3.98.3: Audio CD "Terminator II SE", 53 min, convert WAV to MP3 audio format, Command: -b 160 --nores (160 Kb/s)
HandBrake CLIVersion: 0.99: Video from Canon EOS 7D (1920x1080, 25 FPS) 1 Minutes 22 Seconds
Audio: PCM-S16, 48,000 Hz, Two-Channel, to Video: AVC1 Audio: AAC (High Profile)
TotalCode Studio 2.5Version: 2.5.0.10677: MPEG-2 to H.264, MainConcept H.264/AVC Codec, 28 sec HDTV 1920x1080 (MPEG-2), Audio: MPEG-2 (44.1 kHz, 2 Channel, 16-Bit, 224 Kb/s), Codec: H.264 Pro, Mode: PAL 50i (25 FPS), Profile: H.264 BD HDMV
Productivity
ABBYY FineReaderVersion 10.0.102.95: Read PDF save to Doc, Source: Political Economy (J. Broadhurst 1842) 111 Pages
Adobe Acrobat 11Version 11.0.0.379: Print PDF from 115 Page PowerPoint, 128-bit RC4 Encryption
Autodesk 3ds Max 2013Version 15.0 x64: Space Flyby Mentalray, 248 Frames, 1440x1080
BlenderVersion: 2.68A, Cycles Engine, Syntax blender -b thg.blend -f 1, 1920x1080, 8x Anti-Aliasing, Render THG.blend frame 1
Visual Studio 2010Version 10.0, Compile Google Chrome, Scripted
File Compression
WinZipVersion 18.0 Pro: THG-Workload (1.3 GB) to ZIP, command line switches "-a -ez -p -r"
WinRARVersion 5.0: THG-Workload (1.3 GB) to RAR, command line switches "winrar a -r -m3"
7-ZipVersion 9.30 alpha (64-bit): THG-Workload (1.3 GB) to .7z, command line switches "a -t7z -r -m0=LZMA2 -mx=5"
Synthetic Benchmarks and Settings
3DMark ProfessionalVersion: 1.2.250.0 (64-bit), Fire Strike Benchmark
PCMark 8Version: 1.0.0 x64, Full Test
SiSoftware SandraVersion 2014.02.20.10, CPU Test = CPU Arithmetic / Multimedia / Cryptography, Memory Bandwidth Benchmarks
8. Results: 3DMark And PCMark

A single Radeon R9 290X certainly won’t beat a pair of GeForce GTX 780s in 3DMark, so the performance deficit of both $1600 machines doesn’t bother me.

Weaker overclocking applied to this quarter's Radeon R9 290X allows the previous machine to pass it in this specific test, in spite of the previous machine’s terrible host processor clock rate.

The new build’s drive is significantly cheaper than Q2’s, pleasing me with only a slight drop in PCMark’s storage score. Other test scores are up, thanks to a faster CPU.

9. Results: SiSoftware Sandra

Sandra’s Arithmetic scores shows off the Core i7-4790K's clock rate advantage in stock form compared to Intel's Core i7-4770K. The new CPU also overclocks better, edging out the good overclock I achieved on the first quarter's $2400 PC and smashing the horrible overclock from Q2, which employed the same CPU model apparently cut from a different part of the wafer.

Sandra Cryptography saw marked improvements too, especially in the bandwidth-sensitive encoding/decoding test. A quick look at bandwidth numbers reveals the reason.

Even though I used the same memory in both the current and previous build, this quarter's PC significantly outpaces Q2’s at stock settings. Use of the same RAM points towards the motherboard as the problem, where MSI and ASRock both beat Asus.

10. Results: Battlefield 4

Battlefield 4’s 200 FPS cap takes a big toll on the $2400 Q1 PC's value, though that machine finally started to show SLI's advantage at a resolution of 4900x900. Using just one Radeon R9 290X, the Q2 and Q3 machine are closely matched.

Better GPU overclocking helps last quarter's machine stay ahead of the Q3 build under Battlefield 4's Ultra quality preset. Both PCs used the same graphics card, so this advantage is attributable to the same luck of the draw that hurt Q2’s CPU overclock.

11. Results: Grid 2

Grid 2 is memory bandwidth-sensitive at mainstream detail levels. Last quarter's poor memory performance at stock settings gives way to great results when overclocked, while Q3’s loose timings at overclocked settings have far less impact compared to already-good baseline performance.

The $2400 machine is up to 50% faster at Grid 2’s Ultra preset, though the slowest system's frame rate should still be fast enough for the vast majority of users. Perhaps buyers with 4K displays and 3D glasses will need the expensive graphics?

12. Results: Arma 3

According to our chart representing standard-quality graphics settings, Arma 3 benefits from fast memory, fast host processors, and fast graphics. The transition between those subsystems becomes fairly obvious as resolution increases, with the stock Q2 machine suffering bad memory bandwidth and the overclocked Q1 machine benefiting from the once-fastest single-GPU graphics cards in SLI.

The benefits of SLI really stand out when we enable the Ultra quality preset at 5760x1080, where both of the $1600 machines are barely playable. A look at my test notes shows that the difference between average and minimum frame rates is minimal in this title.

13. Results: Far Cry 3

A high CPU frequency confers huge frame rate benefits onto the Q3 $1600 PC when we benchmark Far Cry 3 at some of its most basic detail settings. Of course, as we already know, this is where the Q1 $2400 machine needs big resolutions for its high-end graphics array to shine.

Both $1600 PCs barely cope with 5760x1080 and Far Cry 3’s Ultra quality preset, as a look at my test notes shows minimum frame rates just over 20. The Q1 machine’s SLI configuration is much more capable.

14. Results: Audio And Video Encoding

The key to deciphering our audio and video transcoding benchmarks is to remember that all three machines employ the same host processing architecture, Haswell.

This quarter's configuration boasts higher stock and overclocked frequencies, edging out the well-overclocked Q1 machine.

15. Results: Adobe Creative Cloud

Adobe After Effects is the only benchmark in our suite to benefit from more than 8 GB of RAM, and that's largely a function of how many threads each CPU can handle concurrently. The $2400 machine had 16 GB to split between four Hyper-Threaded cores. Comparing the 8 GB machines, we see how much further the Q3 build’s extra CPU frequency takes us.

The new build stumbles in Photoshop when we test using GPU-accelerated filters at baseline settings, but picks up ground when overclocked. The same tests show a preference for the $2400 machine’s Nvidia GPUs.

Last quarter's system took a bigger stumble at stock clocks in Premiere, where only its poor memory performance should have separated it from the $2400 machine’s identical frequencies. Overclocking naturally helps all three PCs.

16. Results: Productivity

All three high-end builds utilize the same CPU architecture with an identical core count, so 3ds Max only reflects changes in clock rate. The newest machine wields a better overclock and a higher stock clock rate, so it dominates comparisons in both configurations.

Visual Studio also shows a small, but noticeable performance deficit in the machine with the lowest memory performance. The rest of this category's tests appear to scale well according to CPU frequency.

17. Results: File Compression

A quick look at 7-Zip, WinRAR, and WinZip EZ results shows that last quarter’s machine really was lagging at stock speed, where only its memory bandwidth appeared to be problematic. It’s the same memory as the new build though, so I’ll keep guessing that I might have been encountering a firmware problem of some sort.

Fast memory, a faster stock CPU clock, and better overclocking make a clean sweep of file compression for this quarter’s $1600 build.

18. Power, Heat And Efficiency

Last quarter's system lagged in memory performance at stock settings, but it was also able to claim the lowest combined-load power consumption.

Meanwhile, the new system needs far less voltage to achieve its best overclock, actually reducing power consumption in our CPU-only test.

Of course, the performance of two graphics cards costs big in terms of power for the $2400 machine. These days, you could use a couple of GeForce GTX 980s and enjoy better performance for similar consumption.

The new PC’s CPU temperature doesn’t rise when it's overclocked, mostly because I was able to reduce the fixed voltage setting compared to the stock slope’s maximum. GPU temperatures are similarly stable, mostly because I was able to leave the card’s fan slope at PowerColor's factory settings.

From a cooling value perspective, I might have gone overkill on the new $1600 PC. It certainly didn’t need a $75 cooler to run at this low overclocked voltage. Conversely, I couldn’t have known how much voltage it would need until after I received the processor! Oh, the real-world issues we face as PC builders.

The new build starts out 2.2% more efficient than Q1’s $2400 machine and 11% better than its second-quarter incarnation. Because it was overclocked at relatively low voltage, its class-leading efficiency improves with the added frequency.

19. Is There Such Thing As A Perfect $1600 PC?

We saw big performance gains from this quarter’s build compared to the previous machine, even though it was designed as a direct replacement with the same graphics and RAM. Intel’s higher-clocked Core i7-4790K was one of its key advantages, but we also remember that the previous build’s Core i7-4770K was from a bad batch, wouldn't overclock well at all, and served as a detriment in our final evaluation.

The chart above is one of the cleanest performance-per-dollar comparisons I've ever put together, and I can only hope that the gains made by today’s machine can help it in System Builder Marathon Day 4’s face-off between all three new systems.

Let’s not forget why I spent so much money on graphics way back in Q1, though. That machine was designed to address the complaints of gamers with money to spend. And if gaming is all you really care about, its extra cost might be money well-spent.

I personally find the $1600 machines fast enough to game at 5760x1080. I don’t use 3D displays, I don’t own a pair of 3D glasses, and I doubt that experiments with 4k displays will happen for me this year. In the end, today’s $1600 machine is the first SBM build in recent history that I can recommend almost universally to performance enthusiasts. It’s cheap enough to catch the top of the mainstream market and still fast enough to serve the needs of most high-end users. That applies to both games and applications. It’s too bad I’m forced to give this one away.

That’s not to say I wouldn’t reconsider my purchase of a giant CPU cooler though. I might not have been able to find a reasonably-priced closed-loop cooler when I spec’d this machine, but Newegg now carries several models for less than I paid for big air. Corsair’s reputable H60 is now only $65, for example, achieving a $10 savings in a cooler that will also make the system easier to move around and less likely to damage the motherboard if mishandled.

But how do I know I won’t need a bigger cooler to handle more voltage on a different CPU sample? Rated at 4.4 GHz with a single core active via Turbo Boost, the Core i7-4790K only need a 200 MHz bump to reach the 4.60 GHz clock I achieved today. My somewhat facetious experience with Haswell cores suggests that your chances of reaching 4.8 GHz at safe voltage levels are lower than your chances of meeting your own doppelganger. With so little room to play, the best reason to get a big cooler is to reduce the system’s acoustic footprint. By the way, that was 35 decibels maximum at full GPU / CPU combined load in this build. Now that you know why I’m completely happy with my Q3 effort, let’s get ready for tomorrow’s value rumble!