Asus has announced its first-ever ROG Apex motherboard for AMD CPUs. The ROG Crosshair X870E Apex is Asus' all-new extreme overclocking-focused motherboard, aimed at CPU overclocking and extreme DDR5 memory overclocking. Overclocker Safedisk has already achieved eight world records with the Apex by overclocking two G.Skill DDR5 memory sticks to over DDR5-10000.

The new board has an E-ATX form factor and supports all AM5 Ryzen-based CPUs: the Ryzen 7000 series, Ryzen 8000 series (APUs), and Ryzen 9000 series. The defining trait of the Apex series is the DIMM slot count, which drops from four to just two. This maximizes DDR5 overclocking potential, as fewer DIMM slots enable board makers to improve the traces connecting each slot to the CPU.

Asus advertises a "minimum" overclocking window of "8200+ MT/s" for Ryzen 9000 CPUs, "9600+MT/s" for Ryzen 8000 CPUs, and "8000+ MT/s" for Ryzen 7000 CPUs for CPUs and memory kits listed in the board's CPU/Memory support list.

The board features all the bells and whistles of an Asus extreme overclocking motherboard featuring 18 110A, two 110A, and two 80A power stages with MicroFine alloy chokes and premium SMD capacitors. Asus' plethora of CPU overclocking-focused technologies, such as Dynamic OC Switcher, Core Flex, and Asynchronous clock, are also present, on top of AMD's overclocking features courtesy of Precision Boost Overdrive.

Operating on the X870E chipset, the new Crosshair Apex has a robust assortment of I/O and PCIe connectivity. The board comprises dual PCIe 5.0 slots, one PCIe 4.0 x4 slot, and one PCIe 3.0 x1 slot. Three M.2 slots are on the board, running at PCIe 5.0 speeds. I/O consists of dual USB4 (40Gbps) Type-C ports, six USB 10Gbps ports, five Type-A connectors, and one Type-C connector. Rounding out the rear I/O USB ports are two USB Type-A 5Gbps ports.

The front I/O is almost as numerous, consisting of dual USB 20 Gbps ports, one of which features 60W power delivery and Quick Charge 4+ support, two USB 5Gbps headers supporting four additional USB 5 Gbps ports, and two USB 2.0 headers supporting up to four additional 2.0 ports. There is also a Wi-Fi 7 card (with Bluetooth 5.4 support) for internet connectivity and a single Realtek 5 Gbps Ethernet port.

Like other ROG boards, the Apex has a black finish with RGB LED accents. Covers cover most of the board, including the PCIe slots, M.2 slots, and the rear I/O area.

Professional overclocker Safedisk achieved eight new records with the Crosshair X870E Apex paired to a Ryzen 9 9950X and a 48GB DDR5 G.Skill memory kit overclocked to DDR5-10200 on air cooling. Safedisk scored 17,0757 points in Geekbench 3 multi-core, 324,364 MIPS in 7-Zip, 9,372 points in R15, 23,585 points in R20, 60,868 points in R23, a time of 17s 498ms in GPUPI for CPU 1B, 316.969 FPS in HWBot x265 benchmark at 1080p, and 78.319 FPS in HWBot x265 benchmark at 4k.

Crosshair X870E Apex will be available in the first quarter of this year. Pricing, however, remains a mystery.

Asus’ ROG Maximus Z790 Apex is a board dedicated to overclockers. You get one of – if not the – most powerful VRM solutions in Z790, optimized memory configuration (read: two DIMM slots instead of four), and a slew of integrated buttons and switches, dubbed the Overclocker’s Toolkit, to help push compatible processors to their limits. Priced at $649.99 on the Asus website, you not only get arguably the best made-for-overclocking motherboard with the AI Tools (DIMM Flex, Ai OC, Differential Sensing, and Condensation Detection). It’s also good at gaming and productivity, with a high-quality audio solution, ample storage options (including five M.2 sockets), and the premium ROG appearance. Overall, it’s a well-rounded board that will do well in almost any situation.

While this board can do everything, if overclocking to the CPU limits isn’t something you’ll do, there are plenty of other Raptor Lake refresh boards up the product stack and below. If you need 14th-gen support out of the box, there’s the Z790 Dark Hero ($649.99), the Z790 Formula ($749.99), our Z790 Apex Encore, and some mid-range offerings like the ROG Strix Z790-E Gaming Wi-Fi II ($499.99) and Z790-F Gaming Wi-Fi II ($429.99). None of the MicroATX or ITX-size boards seem to get updates. However, several have the BIOS Flashback feature and could already be using a BIOS with support for the new processors.

The Z790 Apex Encore performance was average to above average in most tests, just like the Z790 Formula we recently reviewed. The increased power limits/duration let the processor run free (power-wise) for a time before the system limits the processor to around 250W. You’re usually limited to what your cooling system allows, anyway.

Below, we’ll dig into the details of the board and see how it performs against other refreshed Z790 motherboards. While it won’t make the best motherboards list, this board is still easy to recommend for the hardcore overclocker, especially those planning sub-ambient cooling methods. But before we get into those details, we’ll start by listing the specifications from Asus.

Specifications: Asus ROG Maximus Z790 Apex Encore

Inside the Box of the Asus ROG Maximus Z790 Apex Encore

Along with the motherboard, Asus includes many accessories to help get you started. You get the basics, plus a DDR5 fan holder, and more. Below is a complete list of the included accessories.

• ROG DIMM.2 with Heatsink
• M.2 pad for ROG DIMM.2
• (2) M.2 screw package for ROG DIMM.2
• Thermal pad for M.2
• ROG Memory Fan Kit
• ASUS WiFi Q-Antenna
• Q-connector
• ROG VIP card
• (3) M.2 Q-Latch packages for M.2 backplate
• (3) M.2 backplate Rubber Packages
• USB drive with utilities and drivers 
• Quick start guide

Design of the Z790 Apex Encore

The Z790 Apex Encore’s appearance oozes premium ROG vibes, with black-on-black, oversized VRM heatsinks and tastefully done accents. The ROG symbol on the M.2 and chipset heatsink is unique, with the three different ‘finishes’ to make the whole. But if you don’t like the look, it will be covered by your video card anyway if it’s installed directly in the top slot, anyway. This standard-size ATX board may not have all of the shrouds similarly (or more expensive) priced boards do, but it’s made for overclocking, and shrouds can get in the way.

For those who want RGB lighting, you’re in luck, as just under the left heatsink is a RGB lighting area that glows through the base. It’s not terribly bright, but the colors are saturated, and it looks good. Either your BIOS or the Aura RGB application in the Armoury Crate software controls the patterns you see.

At the top of the board, in the left corner, we’re greeted by the two Dual Procool II 8-pin EPS connectors (one required) to power the processor. Surrounding them are two heatpipe-connected VRM heatsinks cooling the VRM bits below. On the left is the sole RGB element shining through the miter-cut heatsink. On top are the ROG letters and the board name. To the right, the socket area is clean using the flat (possibly Tantalum POSCAP) capacitors in this area, ready for sub-ambient insulation.

Just past this, we run into the two DRAM slots, reinforced with Asus’s Safe DIMM design. At the time of this writing, the website lists support up to 92GB. However, it actually supports 128GB, as this page wasn’t updated after the March announcement of increased capacity. Memory speeds are listed as incredibly fast to DDR5-8400 - one of the fastest on the Z790 platform. But, like any board, your mileage may vary. To push those speeds, pray your CPU’s integrated memory controller (IMC) can handle it as, currently, a tiny percentage can. The Apex Encore also comes with the ROG Memory Fan Kit designed to keep the warmer running (particularly when overclocked) DDR5 memory in its optimal temperature range.

To the right is Asus’s unique DIMM.2 slot, designed to be used with the included DIMM.2 add-in card for additional M.2 storage. The ROG DIMM.2 card supports two PCIe 4.0 x4 modules up to 80mm, adding to the three M.2 sockets in their more traditional locations on the motherboard. 

Above these slots on the board's top edge are two (of seven) 4-pin fan/pump headers. All headers support PWM and DC-controlled devices with the CPU and Chassis fans, each Q-Fan controlled. The FS and AIO headers all output up to 12W, while the W_PUMP+ is a whopping 36W - plenty for most aftermarket watercooling pumps. The 3-pin EF_FAN header is used with the ROG Memory Fan Kit.

Also above the DIMM slots on the top edge are two RGB headers - in this case, two (of three) 3-pin ARGB headers. Along the bottom edge, you’ll find another ARGB header and another 4-pin RGB header. Control over the integrated RGB feature and any devices attached to the headers is handled through the Aura RGB section in the BIOS or the Aura application in the Armoury Crate software.

Below these is a strip of seven ‘dummy’ LEDs (on or off). The first four on the left are your Q-LEDs. These LEDs, CPU/DRAM/VGA/Boot, light up during the POST. And if a problem arises, the offending area’s LED remains lit, showing users, generically, where the problem lies. When used with the 2-character Q-Code display below, these should help you better understand what went wrong. The three LEDs on the right, labeled CPU, DRAM, or PCIe, are part of the Condensation Detection for the extreme overclocker. When the sensors detect moisture in these areas, the corresponding LED remains lit, just like the Q-LEDs.

Moving down the right edge, we run into several buttons, switches, and a voltage read point area. There’s quite a bit of functionality here, but it’s most useful for those pushing the envelope using sub-ambient cooling methods. The first items we run into is the small Safe boot button below the large Start and Flex Key buttons, and a set of three switches for the Overclocker, including Pause and Slow Mode (the third is reserved for Asus). Also, for the hardcore overclocker, the BCLK +/- buttons to adjust BCLK on the fly, along with a retry button. The strip of dots along the edge is the ProbeIt section that allows users to connect a multimeter to read the voltages off several different areas. Last but not least for the hardcore overclocker is the LN2 mode jumper. Enabling this remedies the cold-boot bug during POST. Below that, we run into the standard 24-pin ATX power connector and, finally, the front panel USB 3.2 Gen 2x2 (20 Gbps) Type-C connector.

Power delivery on the Z790 Apex Encore consists of 26 phases, with 24 dedicated to Vcore. Power flows from the EPS connector(s) onto a 20-channel Renesas RAA229131 PWM controller. From there, it moves to the 24x 105A Renesas RAA22010540 SPS MOSFETs for the processor. The 2,520A available to the processor is one of the most we’ve seen on Z790, and will handle overclocked flagship-class processors like our Intel i9-14900K, even pushing the limits with sub-ambient cooling. You’ll run out of cooling capacity before the power delivery gets in the way. There’s nothing to worry about here!

On the bottom half of the board, we’ll start with the audio section on the left, mostly hidden under shrouds and heatsinks. Asus uses the premium Realtek ALC4080 codec, but unlike the Formula, it does not come with an integrated DAC. I like Asus’ choice to go with the modern audio codec. It’s not a flagship-class solution, but it still has high-quality audio that’s perfect for gaming or other uses.

In the middle of the board, we’ll start with the M.2 sockets. In this case, three are on the board, along with the two from the DIMM.2 module. Onboard, all integrated sockets support up to 80mm modules. Both M.2_1 and M.2_2 connect through the processor, while the top M.2_1 runs up to PCIe 5.0 x4 speeds (128 Gbps), while the lower slot runs up to PCIe 4.0 x4 (64 Gbps). M.2_3 connects through the chipset (as does the DIMM.2 device) and runs up to PCIe 4.0 x4 and SATA modes. If you want additional speed or redundancy, the Z790 Formula supports RAID0/1/5/10 modes for NVMe drives. Just note that when you use the PCIe 5.0-capable M.2, the second full-length PCIe slot is disabled, and the primary slot runs at PCIe 5.0 x8.

For PCIe slots, there are a total of four. Two full-length slots connect through the CPU, and two x4 size through the chipset, which is an update from the Z790 Apex. The top full-length slot, primary for graphics cards, uses Asus’ Q-slot reinforcement and sports the GPU-swapping-friendly Q-Release button. The bottom full-length reinforced slot connects through the CPU and runs up to PCIe 5.0 x8 speeds. But as noted above, you’ll lose this slot when M.2_1 is populated. With bandwidth the same between 5.0 x8 and 4.0 x16, you won’t lose performance on existing graphics cards. Finally, the two x4 slots above and below the full-length slots connect through the chipset with PCIe 4.0 x4 lanes available.

Moving to the edge of the board, a lot is happening here, too. You have the standard SATA ports (4) and front panel USB (3.2 Gen 1, 5 Gbps), but you also have an 8-pin PCIe power connector to support the quick charging (QC4+) capability. Just below is a fan header and a lightbar header for RGBs.

Across the bottom of the board are several exposed headers. You’ll find the usual, like additional USB ports, RGB headers, a header for temperature and water flow probes, and a BIOS switch, and a switch to adjust PCIe generations on the fly. Below is a complete list from left to right.

• Front panel audio
• 4-pin RGB header
• 3-pin ARGB headers
• (3) System Fan headers
• Reserved and V_Latch switches
• PCIe Alt Mode switch
• Thunderbolt header
• USB 3.2 gen 1 (5 Gbps) connector
• Water Flow in/out headers
• (2) USB 2.0 headers
• 2-pin temperature sensor
• Front panel header

The Rear IO plate on the Z790 Apex Encore comes preinstalled, as you’d expect from high-end boards these days. It uses a black background with white labels on the ports for easy reading. There are a total of 10 USB ports: One USB 3.2 Gen 2x2 (20 Gbps) Type-C, Five USB 3.2 Gen 2 (10 Gbps) ports in red, and five USB 3.2 Gen 1 (5 Gbps) ports in blue. On the left are the BIOS Flashback and CMOS reset buttons, next to legacy PS/2 ports that are especially useful for extreme overclocking. For networking, you have the 2.5 GbE port, along with the Wi-Fi 7 Q-Connector. Lastly, the audio stack comprises five 3.5mm analog plugs and an SPDIF output.

MORE: Best Motherboards

MORE: How To Choose A Motherboard

MORE: All Motherboard Content

Firmware

Asus’ BIOS on the Z790 Formula remains unchanged. The BIOS sports the familiar black, red, and yellow ROG theme that’s easy to read. Asus starts in an Easy Mode that displays high-level information, including CPU and memory clock speeds, temperatures, fan speeds, storage information, etc. Advanced Mode has several headers across the top that drop down additional options. The BIOS is one of my favorites, as almost everything you need isn’t buried deep within menus.

A unique feature of Asus BIOS is the integrated Memtest functionality. As the name implies, Memtest86, a memory testing application, is in the BIOS ready to use. This is most useful to the tweaker and overclocker, but even those who ‘just’ set XMP can find it useful.

Software

Armory Crate for the Formula is the same as other ROG-based boards and follows the ROG-inspired theme. Several applications exist for various functions, ranging from RGB lighting control, audio, system monitoring, overclocking, etc. We’ve captured several screenshots of the applications below. Here’s a look at Ai Suite 3, Armoury Crate, Sonic Studio, and the Realtek Audio application.

Also worth mentioning is the included software. When buying Asus motherboards, you get a year of AV software (McAfee, not a trial). You also get a full 1-year license of AIDA64 - a good application for stress and performance testing.

Test System / Comparison Products 

 We’ve updated our test system to Windows 11 64-bit OS with all updates applied as of mid-October 2023. We kept the same Asus TUF RTX 3070 video card from our previous testing platforms but have updated the driver to the latest, keeping our games, F1 22 and Far Cry 6, the same. Unless otherwise noted, we use the latest non-beta motherboard BIOS available to the public using ‘optimized default’ settings, except for the memory (XMP). The hardware and drivers we used are as follows: 

Test System Components

 Benchmark Settings 

MORE: Best Motherboards

MORE: How To Choose A Motherboard

MORE: All Motherboard Content

Benchmark Results and Final Analysis

Our standard benchmarks and power tests are performed using the CPU’s stock frequencies (including any default boost/turbo) with all power-saving features enabled. We set optimized defaults in the BIOS and the memory by enabling the XMP profile. The Windows power scheme is set to Balanced (default) for baseline testing so the PC idles appropriately.

Synthetic Benchmarks

Synthetics provide a great way to determine how a board runs, as identical settings should produce similar performance results. Turbo boost wattage and advanced memory timings are places where motherboard makers can still optimize for stability or performance, though, and those settings can impact some testing.

The Z790 Apex did well for the synthetic benchmarks, running slightly faster than average across several tests. The power policy allows the CPU to run free but is eventually limited to around 250W and your cooling capacity, like most boards. There is nothing out of the ordinary in this set of tests.

Timed Applications

Starting with Handbrake in the Timed tests, the Z790 Apex Encore was on the faster side of average, completing the tests in 82 and 229 seconds, respectively. The timed tests are good also!

3D Games and 3DMark

Starting with the launch of Zen 4, we shifted our test games from F1 21 to F1 22 while keeping Far Cry 6. We run the games at 1920x1080 resolution using the Ultra preset (details listed above). As the resolution goes up, the CPU tends to have less impact. The goal with these settings is to determine if there are differences in performance at the most commonly used (and CPU/system bound) resolution with settings most people use or strive for (Ultra). We expect the difference between boards in these tests to be minor, with most falling within the margin of error differences. We’ve also added a minimum FPS value, which can affect your gameplay and immersion experience.

For gamers, the Z790 Apex Encore did well in the synthetic tests and was average overall in games. It was tied for the slowest by around 7 Frames Per Second in Far Cry 6 (a few percent slower) but was on the average in F1 2022. It’s a competent gamer, though.

Overclocking

Generation after generation, overclocking headroom has been shrinking with both Intel and AMD processors, with motherboard partners pushing the limits to set themselves apart from the plethora of options available to the consumer. With the overclocking headroom all but gone now, we’ve left things at stock for cores, but will push the integrated memory controller (IMC) for faster RAM.

The Apex (and several Asus boards) offer AI Overclocking for those who don’t want to tinker manually. AI OC is a hardware-based design (it needs a microcontroller on the board) that’s specific to your system, reading the VF curve and silicon information, among other things, to gauge overclocking capability. You can engage this feature from the Armoury Crate or within the UEFI. Users can also define a temperature target. For example, if your fans or pump fail/degrade (or ambient rises), AiOC downclocks your overclock to meet the target, whereas manual overclocking leaves it in place throttling. Additionally, if you increase your cooling capacity, it adjusts up as well.  Asus claims you can get an additional 300-600Mhz on the P and E cores. But what you’ll gain varies wildly (depending on silicon quality, cooling, etc.). We stuck with pushing the RAM.

We tested all three of our memory kits, and all three worked, even our fastest DDR5-8000 kit from Klevv. Setting XMP worked on the DDR5-6000 and 7200 kits. However, we needed to raise the System Agent voltage a small amount (from 1.21V to 1.275V) in the BIOS to pass our 30-minute stability test. If you’re into pushing memory and the high-end ITX boards don’t tickle your fancy, the Apex is the way to go. Just ensure you have a CPU that runs those high memory speeds (rare) and a kit that’s on the memory QVL.

Power Consumption / VRM Temperatures

We used AIDA64’s System Stability Test with Stress CPU, FPU, Cache and Memory enabled for power testing, using the peak power consumption value. The wattage reading is from the wall via a Kill-A-Watt meter to capture the entire PC (minus the monitor). The only variable that changes is the motherboard; all other parts remain the same. Please note we moved to using only the stock power use/VRM temperature charts. Since the system uses every available degree Celsius, unless you’re using a sub-ambient cooling solution, you’ll use more power and generate more heat using default settings.

Idle power consumption on the Apex Encore was one of the lower values, settling at 59W at the wall. Load wattage peaked at a stunning 466W for the system during stress testing, the highest value we’ve recorded. However, this is just a peak, as the PL2 limit drops to 253W after some time. You can change this behavior in the UEFI through canned settings or adjust it yourself.

VRM temperatures are well within specification. The oversized, heatpipe-connected heatsinks keep the high-end MOSFETs running in their sweetspot. Like the X790 Forumula, the Apex Encore handled our Intel Core i9-14900K without issue using ambient cooling and will do so even when sub-ambient cooling methods are used.

Bottom Line

Asus’ Z790 Apex Encore is a curtain call for the original Apex. While there aren’t a lot of changes between them, the updated Encore presents users with the best Asus has to offer for the extreme overclocker. You get an updated, improved aesthetic, Wi-Fi 7 instead of Wi-Fi 6E, native 14th-gen support, and higher memory speed support (among other things). I also like the black-on-black look of the Encore better than the white. Priced at $649.99 on the Asus website (more expensive in many other places), you get the full gamut of features the Z790 platform offers, as well as all of the Asus exclusives, including DIMM Flex, AiOC, and the slew of “Q” features (Q-Latch/Release/Antenna/Code/LED). Asus does an excellent job of adding useful extras, which brings additional value to the product, and the Apex Encore is no different.

As far as the competition goes, that’s tough as there is only one direct competitor in the Z790 Tachyon X, which is also focused on overclocking. The Tachyon is rated higher with memory speeds but doesn’t have Wi-Fi 7, uses a last-gen (though still quality) audio solution, and is also E-ATX versus the Apex Encore at ATX. The other problem is finding a Tachyon at a reasonable price. No MSRP was listed, and a quick look around the web shows the updated Tachyon X version costs over $1,000 from a third-party seller. We imagine the MSRP is much lower, but that’s how much you’ll have to pay now.

In the end, we like what the Asus Apex Encore offers for those who want to push the limits of their processor and RAM, especially those using sub-ambient cooling methods like liquid nitrogen. Extreme overclocking aside, the board is also suitable for daily computing. Be it productivity, content creation, or gaming, the Apex can handle it with aplomb. It also looks like a premium motherboard. The black-on-black look with a tasteful RGB implementation looks good on a test bench or inside your case. If you don’t plan on pushing your RAM or processor to the limits, plenty of options within the Asus product stack cost less, perform similarly, and look as good without some of the bells and whistles for overclocking.

MORE: Best Motherboards

MORE: How To Choose A Motherboard

MORE: All Motherboard Content

Seby9123, one of the world's top professional overclockers, set a new DDR5 SDRAM data transfer record of 11,240 MT/s at G.Skill's overlocking competition at Computex last week. Using premium hardware, liquid nitrogen, and a lot of magic, he has managed to increase the stock speed of G.Skill's DDR5-7800 module by 44%.

The new world record of 11,240 MT/s set by Seby9123 is just slightly higher than the second-highest result of 11,233 MT/s set by Hicookie on the same day. Meanwhile, the third highest result is noticeably higher than 11,130 MT/s set by lupin_no_museme back in October 2022, so the progress made in a few months by DDR5 is more or less obvious.

To set his memory overclocking record, Seby9123 used Asustek's ROG Maximus X790 Apex motherboard, Intel's Core i9-13900KS processor with only two cores enabled (albeit working at 5.0 GHz), a lot of liquid nitrogen, and a 16 GB G.Skill Trident Z5 RGB DR5-7800 memory module in a single-channel mode. The actual frequency was 5620.1 MHz, and timings were set to CL62 126-126-127-127-2. 

Unfortunately, there is no word about DRAM voltage for the memory, but we suspect it is even higher than 1.45V, which is already over 30% higher compared to the standard DDR5 voltage set by JEDEC.

In addition to setting the DDR5 speed record, the world's top extreme overclockers assembled by G.Skill at its contest set several other overclocking records at Computex.

Sabrent's Rocket 4 Plus SSD is one of the best SSDs around. Now imagine glomming 21 of these PCIe 4.0 SSDs together, because that's precisely what Sabrent has planned for the company's next product. The upcoming Sabrent Apex X21 Destroyer will arrive with 21 Rocket 4 Plus 8TB drives to offer unprecedented storage capacity and transfer speeds.

Sabrent is still conducting in-house testing on the Apex X21 Destroyer. However, the company said it easily achieved sequential read and write speeds of 31 GB/s. Sabrent's results even surpassed Apex Storage's figures, which rate the X21 with a sequential performance of 30.5 GBps reads and 28.5 GBps writes.

If you look at the brand's past products, such as the Sabrent RocketQ Battleship, Sabrent Rocket 4 Plus Destroyer, or Sabrent Rocket 4 Plus Destroyer 2, it's easy to see that Sabrent has a sweet tooth for big storage solutions. However, the Sabrent Apex X21 Destroyer is on an entirely different level. Previous solutions saw Sabrent assembling a crew of eight SSDs. Now, the Sabrent Apex X21 Destroyer has enough landscape for 21 drives, offering almost three times the storage and significantly more performance.

Sabrent's Apex X21 Destroyer is based on the X21 from Apex Storage. While Sabrent's prior AICs leverage the HighPoint SSD7540 PCIe 4.0 x16 RAID card, the Apex Storage X21 doesn't feature a RAID controller. Instead, it utilizes a pair of unnamed PCIe 4.0 switches. This allows consumers more freedom to set up the AIC to suit their needs. Nonetheless, the Apex Storage X21 is bootable and supports RAID arrays through software or third-party hardware solutions, like the ones from GRAID.

The X21 features a passive cooling design because the riser card's primary target is the server market. However, when deployed inside a regular system or even a workstation, the X21 will require some form of active cooling assistance. This is because so many PCIe 4.0 SSDs packaged together will output a fair amount of heat. Apex Storage recommends at least 400 LFM of airflow for the X21. In addition, given the considerable number of drives on the X21, the AIC needs two 8-pin PCIe power connectors to feed the entire lot of SSDs. It's basically like a high-end graphics card when fully loaded, at least from the power requirements perspective.

It's worth pointing out that maximum performance ends up being limited by the x16 PCIe 4.0 connection to the card. A single Sabrent Rocket 4 Plus 8TB can hit peak transfer rates of just over 7,000 MB/s in our testing, so 21 of them in aggregate could theoretically deliver 147 GB/s. However, an x16 PCIe 4.0 slot tops out at 16 GT/s per lane, and with 128b/130b encoding you're limited to at most 31.5 GB/s of actual throughput.

Sabrent hasn't established a price tag for the Apex X21 Destroyer yet, and there's no estimated time frame for when the AIC will release. Considering that each Rocket 4 Plus 8TB SSD currently retails for $1,099 (down from an official MSRP of $1,999 at launch), it would be foolish to think that the Apex X21 Destroyer will sell for less than $23,000. Moreover, the X21 sells for $2,800, so we have to factor that into the pricing. Even if Apex Storage offers discounts on volume purchases, the X21 still costs an am and a leg.

Apex Storage, a newcomer to the storage game, has launched the X21, an AIC (add-in-card) that accommodates up to 21 PCIe 4.0 M.2 SSDs. With the X21, consumers can use the best SSDs to build configurations with capacities of up to 168TB and experience speeds up to 31 GBps.

Apex Storage has its headquarters in Utah and focuses on AIC products. Mike Spicer and Henry Hill appear to be the team behind Apex Storage. You may have heard of Spicer as he kickstarted the Storage Scaler expansion card in 2021 that houses up to 16 M.2 SATA drives. The X21, the only listed product on Apex Storage's website, seems to be the jacked-up version of the original Storage Scaler. But, strangely, Apex Storage opted for PCIe 4.0 on the X21 since PCIe 5.0 SSDs are already available on the retail market.

The X21 sticks to a double-width full-height full-length (FHFL) form factor with a single-slot PCI design. The AIC, which measures 274.2mm long, communicates through a standard PCIe 4.0 x16 expansion slot. It's backward compatible with PCIe 3.0, but performance will take a significant hit. Apex Storage essentially bonded two PCBs in a sandwich with the X21, which is why the AIC can handle up to 21 M.2 SSDs putting rival options like the Sabrent Rocket 4 Plus Destroyer 2 to shame. The concept takes us on a trip down memory lane, for sure. If you're old enough, it's the same concept Nvidia used for the prehistoric GeForce 7900 GX2.

There are 10 PCIe 4.0 M.2 slots and a massive heatsink that covers the unspecified controller in the X21's interior. We suspect a PCIe switch is under the heatsink, likely making the X21 tick. The remaining 11 PCIe 4.0 M.2 slots are located on the exterior of the PCBs.

Given the vast number of SSDs the X21 can support, the AIC can't get all the power from the single expansion slot. Therefore, Apex Storage implemented two regular 6-pin PCIe power connectors to provide auxiliary power. This configuration allows for up to 225W. Although the X21 features a passive cooling design, the manufacturer does recommend a minimum airflow of 400 LFM for optimal operation.

The X21 photographs show the AIC with Samsung 990 Pro SSDs and the card only supports M.2 2280 drives. It also accepts Intel's Optane M.2 SSDs, like the H20; not like there are a lot of Optane drives floating around now that Intel has axed its Optane business. Apex Storage doesn't reveal the inner details of the X21. However, the manufacturer confirmed that the X21 offers 100 PCIe lanes, suggesting the presence of a PCIe switch.

In a single-card configuration, the X21 delivers sequential read and write speeds up to 30.5 GBps and 28.5 GBps, respectively. The random performance on the AIC consists of 7.5 million IOPS reads and 6.2 million IOPS writes. The X21 shines even more in a multi-card setup. According to Apex Storage, consumers can enjoy up to 107 GBps sequential reads and 80 GBps sequential writes. The random performance also gets a substantial boost. The X21 flaunts over 20 million IOPS reads and 10 million IOPS writes in a multi-card arrangement. The AIC has an average read and write access latency of 79us and 52us, respectively.

With 8TB SSDs, such as the Corsair MP400 or the Sabrent Rocket Q, the X21 can supply up to 168TB on a single card. It also supports higher-capacity drives. Once future 16TB M.2 SSDs hit the market, consumers can have up to 336TB of storage on the X21. Regarding the feature set, the X21 supports RAID configurations in Windows and Linux environments. However, Apex Storage didn't expose the type of RAID arrays. The X21 also flaunts "enterprise-grade reliability," NVMe 2.0 support, advanced EEC, data protection, and error recovery.

Apex Storage didn't reveal the pricing or availability for the X21.

Rising inflation and a weakening global economy had forced many, if not most, chipmakers to lower aggressive expansion plans set when demand for chips exceeded supply. Nonetheless, even with the semiconductor industry's decreased capital expenditures (CapEx), the 2022 forecast stands at $181.7 billion, a new record high, according to IC Insights.

Capital expenditures of semiconductor manufacturers have been increasing since 2020 due to strong demand for everything digital, from smartphones and PCs to giant televisions and self-driving cars. The report claims that CapEx budgets increased by 35% in 2021 to $153.1 billion and are poised to grow 19% in 2022 to $181.7 billion.

But now that companies like Intel and Micron are reviewing their CapEx plans for 2023, IC Insights revises its forecasts and believes that the industry's spending will drop by 19% in 2023 to $146.6 billion. This is just 4.25% lower than the $153.1 billion in 2021 and is significantly above capital spending in 2020.

Meanwhile, IC Insights believes that the memory industry will suffer more than the logic industry as major producers will reduce their capital expenditures by 25%. Also, Chinese semiconductor companies will have to cut their budgets by around 30% as the U.S. sanctions against the People's Republic's chip sector will inevitably affect their ability to procure new production tools.

Anyhow, a 19% decline in capital spending year over year is the steepest decline since the global financial crisis in 2008 – 2009, notes IC Insights.

Indeed, semiconductor industry CapEx spending dropped 40% year-over-year in 2009 but only grew to 107% in 2010 when the global economy showed signs of a rebound. Given how quickly semiconductor CapEx budgets have increased in recent years, we would not expect them to grow by around two times from $146.6 billion in 2023. Meanwhile, virtually all makers of chips (including logic and memory chips) expect demand for their products to rebound in 2024 ~ 2025. As a result, they think they need expanded production capacities to meet that demand in the middle of the decade. As a result, spending on new fabs will grow significantly in 2024 ~ 2025.

Fabs already being constructed or equipped, including those in the U.S. by companies like Intel, Micron, Samsung Foundry, TSMC, and Texas Instruments, will come online on time, as delaying such projects is expensive.

Interestingly, IC Insights does not believe that funding U.S. semiconductor suppliers as part of the U.S. CHIPS and Science Act will give a significant boost to their spending in 2023 as they will use the money they are going to receive in grants in a bid to replace their own money they would have spent on their new fabs.

"In other words, the CHIPS and Science Act money is not expected to be 'additive' funding to planned semiconductor industry spending, but instead is likely to replace the money a semiconductor producer was going to budget if CHIPS and Science Act funding was unavailable," wrote Bill McClean, the president of IC Insights.

As demand for PCs and smartphones slumps, so does demand for 3D NAND and DRAM memory. As a result, companies like SK Hynix suffer from dropping revenue and dramatic declines of profits. According to reports via Reuters and Nikkei Asia, in a bid to balance the books, SK Hynix plans to halve its capital expenditures next year and focus on the manufacture of more expensive types of memory. The company also plans to assess the future of its fab in China.

SK Hynix Cuts CapEx Due to Major Profits Fall

Just like other semiconductor companies, the memory maker believes that demand for chips will be sluggish for several quarters and supply will exceed demand, which means tiny profits amid lowering revenue. To that end, SK Hynix will reduce its CapEx investment next year by more than 50% year-over-year. The company does not disclose how much will it spend on new fabs and tools in 2023, but only says that it would be 'at the upper range of 10 – 20 trillion won ($7 billion – $14 billion). 

SK Hynix has good reasons to cut down its expenses. This week the company posted a 10.983 trillion won ($7.741 billion) revenue, which was down 7% compared to the same quarter a year ago. A 7% YoY drop is hardly significant, but the company also reported operating profit of 1.656 trillion won ($1.167 billion) in operating profit (a 60% year-over-year decline), and net income of 1.103 trillion won ($777.396 million) in the second quarter of 2022 (a 67% YoY decrease).

Reduce Production, Focus on Premium Segments

In addition to reducing its CapEx, SK Hynix will increase production of its premium products, which includes 238-layer 3D NAND memory that can be used to power the fastest SSDs with a PCIe 5.0 x4 interface. Also, the company intends to boost production of memory using its 1a-nm fabrication technology in general as well as premium DDR5, LPDDR5, and HBM3 for high-end applications like servers, AI, and analytics accelerators. 

In addition, the company intends to lower production of commodity 3D NAND and DRAM memory in a bid to balance supply and demand on the market. Meanwhile, the company does not disclose how significantly it plans to cut its output.

Fab Dilemma

Back in September SK Hynix began to construct its $11 billion fab in South Korea that is set to come online in 2025. It remains to be seen whether the company will be able to build the new production facility on time with its reduced CapEx budget, but it looks like it might face a dilemma with its spendings on new manufacturing capacity. 

Building a brand new fab is by definition more expensive than upgrading existing production facilities. SK Hynix has fabs in China and South Korea. The Chinese fabs may be cheaper to operate (and the company needs to cut costs now) and can even be upgraded with new tools produced in the U.S. in the next 12 months as the company has a waiver to ship new equipment from America to China for a year. But if the company's partners fail to get an export license from the U.S. Department of Commerce to ship new tools and spare parts to SK Hynix's facilities in Wuxi and Dallian (which is formally controlled by SK Hynix's subsidiary Solidigm), these fabs will become obsolete in a couple of years (i.e., 2025 – 2026). 

Given the risks associated with its Chinese fabs, it might make sense for SK Hynix to focus on investments in its all-new fab in South Korea. At present the company is considering several contingency plans for the facilities, some of which include transferring equipment to South Korea, reports Nikkei. 

"As a contingency plan, we are considering selling the fab, selling the equipment or transferring the equipment to South Korea," said Kevin Noh, chief marketing officer for SK Hynix, at the earnings call this week. "It's a contingency plan. We want to [continue to] operate without facing this situation. We expect to extend it each year, but it is not for sure. It is very uncertain."

Earlier this year, SteelSeries launched the Apex Pro Mini and Apex Pro Mini Wireless — redesigned 60 percent mini keyboards featuring the company’s new OmniPoint 2.0 linear optical switches with adjustable actuation. The Apex Pro Mini/Apex Pro Mini Wireless did an excellent job of shoving every possible feature into the tiniest, most ultra-compact package possible. But even packed with features, a 60 percent layout is pretty extreme for anyone with a desk larger than an economy class airplane tray. 

Good news: SteelSeries now offers its Apex Pro line in a more accessible TKL (tenkeyless) format with the Apex Pro TKL and Apex Pro TKL Wireless. This keyboard is still not full-size, but I assume that’s coming — eventually.

The Apex Pro TKL/Apex Pro TKL Wireless sport the line’s OmniPoint 2.0 linear optical switches, which have both adjustable actuation (37 levels per-key) and the impressive-sounding (but tricky-to-use) dual-action actuation. The keyboards also pack plenty of premium features — aluminum alloy top plate, double-shot PBT keycaps, bright per-key RGB, and a detachable magnetic wrist rest. The Apex Pro TKL/Apex Pro TKL Wireless’ roomier layout allows for dedicated media keys and a customizable OLED smart display.

For this review, I tested the Apex Pro TKL Wireless, but the two keyboards are largely identical, differing mainly in connectivity: the Apex Pro TKL is wired, while the Apex Pro TKL Wireless adds both 2.4GHz “Quantum 2.0” wireless and Bluetooth 5.0. Both keyboards are available now, and they’re priced (unsurprisingly, as part of SteelSeries’ flagship lineup) on-par with the most expensive boards on our list of best gaming keyboards: $190 for the Apex Pro TKL and $250 for the Apex Pro TKL Wireless.

Design and Construction of the Apex Pro TKL Wireless

The Apex Pro TKL Wireless is a wireless keyboard with a TKL (tenkeyless) layout, which means it includes function, arrow, and navigation keys (but no numpad). In the upper right corner, it also has a customizable OLED smart display, a multi-function media key, and a textured, clickable metal volume roller.

Visually, the Apex Pro TKL Wireless is almost identical to the Apex 9 TKL — save for the OLED smart display. Both keyboards have the same matte black chassis, “aircraft-grade” aluminum alloy top plate, and matte black double-shot PBT keycaps with translucent legends. On the back, both keyboards have two pairs of flip-out plastic feet for adjusting the angle. The back of the Apex 9 TKL also has an orange keycap puller stored under a silicone flap; the Apex Pro TKL Wireless presumably needs this space for its battery (a keycap puller does come in the box, however).

The Apex Pro TKL Wireless comes with a detachable magnetic wrist rest, which fits perfectly along the keyboard’s bottom edge and has a slightly rubbery, soft-touch finish. The wrist rest isn’t particularly plush — it’s not soft and cushy, like the HyperX Wrist Rest — but its velvet-smooth surface looks and feels luxe, and its snug alignment offers excellent wrist and palm support. It only aligns perfectly when you’re not using the keyboard’s flip-out feet (but you shouldn’t be using those anyway, if ergonomics are important to you). The wrist rest also fits (and magnetically attaches to) the Apex 9 TKL.

In addition to the keycap puller and wrist rest, the Apex Pro TKL Wireless also comes with a 2.4GHz wireless USB-C dongle, a USB-C extender, and a braided, seven-foot USB-C to USB-A cable. The Apex Pro TKL Wireless has the same measurements as the Apex 9 TKL — 14 inches long by 5.04 inches deep and 1.65 inches thick.

While the Apex 9 TKL is listed at 1.4 pounds (635g) on its box, the Apex Pro TKL Wireless is listed at 2.4 pounds (1070g). This seemed suspect — the Apex Pro TKL Wireless’ battery and OLED screen don’t seem like they weigh an extra pound — so I weighed both keyboards: The Apex 9 TKL weighed 1.42 pounds (645g), and the Apex Pro TKL Wireless weighed 1.65 pounds (747g). The Apex Pro TKL Wireless and all of its accessories (wrist rest, cable, dongle, extender, and keycap puller) weighed approximately 2.33 pounds (1059g), however — so we can probably chalk this up to an early-production mistake.

Specs 

Typing and Gaming Experience on the Apex Pro TKL Wireless

The Apex Pro TKL Wireless comes with SteelSeries’ OmniPoint 2.0 linear optical switches. Optical switches technically have mechanical parts, but they’re not the same as mechanical switches — mechanical switches are actuated via physical force, while optical switches are actuated via light. The main advantage of optical switches is speed (light is, after all, faster than...everything else) — according to SteelSeries, its OmniPoint 2.0 switches have an actuation time 10x faster than “traditional mechanical key switches.”

The downside of optical switches is the overall feel and typing experience. Typing on the Apex Pro TKL Wireless wasn’t terrible — of course, I knew what to expect from my time with the Apex Pro Mini. The OmniPoint 2.0 switches are linear, with smooth, uniform keypresses — no tactile bump or audible feedback on the press, though I still appreciated the semi-hollow clacky sound I got with more aggressive typing. These weren’t bad for optical switches, but they did retain the slight sponginess on the bottom out that I don’t feel with mechanical switches (even linear ones).

The Apex Pro TKL Wireless may be just okay for typing, but it’s excellent for gaming. The linear optical switches are light and speedy — perfect for games that require a lot of quick successive keypresses. Plus, the fact that you can adjust each key’s actuation level individually means you can customize your gaming setup so that you can skirt over some keys without accidentally actuating others. I’m still not totally sold on the dual-bind feature, which allows you to bind two actions to one keypress (at different actuation levels) — but it’s more of a bonus, anyway, since the Apex Pro TKL Wireless already allows for extensive per-key customization.

Features and Software of the Apex Pro TKL Wireless

The Apex Pro TKL Wireless has a handful of settings that can be controlled directly from the keyboard (via default secondary keybind and/or the OLED smart display), without the need to download or install any software. These include brightness and actuation level, switching profiles, checking battery life, and recording macros. But the main selling point of Apex Pro keyboards is extensive, detailed customization. And you’ll need SteelSeries’ companion app, SteelSeries GG, to take full advantage.

You can do a lot when it comes to customizing the Apex Pro TKL Wireless in SteelSeries GG (arguably quite a bit more than you really need to be able to do). You can configure primary and secondary keybinds, adjust primary and secondary actuation points (0.2mm - 3.8mm, in increments of 0.1mm), and assign dual bindings (up to eight).

On the customization side, there’s an illumination section for configuring the keyboard’s bright, per-key RGB. You can also customize the image (128 x 40 pixels) that appears on the OLED screen when it’s not displaying other info (not very useful, but fun). In addition to your custom image and remaining battery life, the OLED screen can display information from apps and games (e.g. real-time hardware monitoring, Discord notifications) through SteelSeries GG’s app/game integrations.

Wireless Experience and Battery Life of the Apex Pro TKL Wireless

The Apex Pro TKL Wireless offers three ways to connect: 2.4GHz “Quantum 2.0” wireless for gaming, Bluetooth 5.0 for convenience, and wired (via USB-C). For gaming, the keyboard’s 2.4GHz wireless connection is solid, stable, and very low latency (not no latency, but virtually undetectable). While the keyboard’s wireless performance was excellent while in use, it did seem to lag a little when waking up after a keypress — I think this may have been related to one of the firmware updates/rollbacks I had to do during my testing (I was assured all firmware issues would be fixed upon launch, so it’s not a big issue — but it’s there).

SteelSeries says the Apex Pro TKL Wireless gets up to 40 hours of battery life — not a powerhouse, but decent considering the keyboard’s relatively small footprint and light weight. Plus, most of us aren’t toting TKLs around on a regular basis, so 40 hours should be plenty of juice for something that can be easily and frequently plugged in.

The Bottom Line

The Apex Pro TKL Wireless isn’t revolutionary or even surprising. It’s the Apex Pro Mini-redesign of the existing Apex Pro TKL, featuring the Apex Pro Mini’s sturdy, minimal, and more generic-looking chassis design and updated OmniPoint 2.0 switches. The new Apex Pro TKL Wireless retains its predecessor’s customizable OLED screen, media key/volume roller, and detachable magnetic wrist rest. It also comes in both wired and wireless versions, unlike its predecessor.

The Apex Pro TKL Wireless is an excellent, highly customizable gaming keyboard in a TKL layout, which is much less restrictive than the Apex Pro Mini’s 60 percent layout — but I’m still holding out for an update to the full-size SteelSeries Apex Pro.

My only minor issue with the Apex Pro TKL Wireless is that — despite being a very predictable next step in SteelSeries Apex Pro lineup, it felt hastily launched. I had a handful of issues while I was testing that were fixed and unfixed by some firmware updates and rollbacks. SteelSeries says all issues would be fixed by launch. While that may be accurate, I still want to double-check to make sure everything’s been ironed out.

MORE: Best Gaming Keyboards

MORE: How to Pick Keycaps for Your Mechanical Keyboard

MORE: How to Build a Custom Mechanical Keyboard

MORE: How to Make Your Mechanical Keyboard Feel and Sound Its Best

A couple of months ago, SteelSeries launched the Apex Pro Mini — an ultra-compact, highly-customizable gaming keyboard. And while it may be one of the best gaming keyboards on the market, it’s also, frankly, more than a little too compact and too customizable for the majority of gamers. The brand dials it back a little (but just a little) with the launch of its Apex 9 series.

The Apex 9 series comes in two layouts — TKL (tenkeyless) and Mini (60 percent); this review will focus on the Apex 9 TKL, but the keyboards are largely identical (minus layout). The Apex 9 TKL and Apex 9 Mini are both wired keyboards sporting SteelSeries’ new hot-swappable OptiPoint linear optical switches. These switches have 2-point actuation, which means they have two points of actuation (1mm and 1.5mm) — like the Apex Pro Mini’s adjustable actuation, only with two preset points that you can toggle between, instead of per-key adjustability to the nearest 0.1mm.

The SteelSeries Apex 9 TKL and Apex 9 Mini are available now, for $140 and $130, respectively; these keyboards aren’t nearly as customizable as the Apex Pro Mini, but they’re also not nearly as expensive. At the moment, the Apex 9 TKL and Apex 9 Mini are only available with linear switches; SteelSeries plans to offer both clicky and tactile versions of its OptiPoint switches in the future as accessory packs, but at the moment we have no information on how much these will cost or whether the company will offer the keyboards with clicky/tactile switches included in the future.

Specs

Design and Construction of the Apex 9 TKL

The Apex 9 TKL sports a tenkeyless layout, with navigation keys and arrow keys but no number pad (the Apex 9 Mini has a 60 percent layout — no number pad, navigation keys, or arrow keys). The keyboard is sturdily built, with a “Series 5000” aircraft-grade aluminum metal frame and top plate (though the chassis itself is plastic) and double-shot PBT keycaps.

The Apex 9 series has the same minimal, elegant chassis design we first saw on the Apex Pro Mini. In fact, the Apex 9 Mini is visually identical to the Apex Pro Mini.

The Apex 9 TKL has slightly more room and includes a multifunction media button and a textured, clickable metal volume roller in the upper right corner.

On the back of the Apex 9 TKL, you’ll find two pairs of flip-out feet for height adjustment, as well as an orange plastic SteelSeries keycap puller stored under a silicone flap. The keyboard is wired via detachable USB-C cable (it comes with a 7-foot braided USB-C to USB-A cable) and has a USB-C port in the upper left corner. I know SteelSeries is trying to keep the chassis streamlined, but it would be nice to see a passthrough port or two (or even multiple USB-C input ports for cable management).

Typing and Gaming Experience on the Apex 9 TKL

The Apex 9 comes with SteelSeries’ new OptiPoint optical switches, which are hot-swappable and feature 2-point actuation (not to be confused with dual-actuation, which is a feature that currently only exists on the Apex Pro Mini).

While the Apex 9 does have hot-swappable switches, it’s only compatible with optical — not mechanical — switches. The company also says the keyboard only officially supports its own OptiPoint switches; while third-party optical switches will fit in the board, those with black bottom housings absorb too much light to operate accurately (OptiPoint switches have white bottom housings).

The Apex 9’s switches feature 2-point actuation, which means they have adjustable actuation with two levels — 1mm (“gaming mode”) and 1.5mm (“typing mode”). You can toggle between these two modes using SteelSeries’ GG software or with a keyboard shortcut (the SteelSeries key + I/O). I was initially concerned that these two actuation points didn’t vary enough in distance, but using the keyboard in both modes (for both typing and gaming) put that to rest.

Typing on the Apex 9 feels pretty good, considering it has linear optical switches instead of my preferred clicky mechanical switches. The Apex 9’s switches have both a lighter actuation force (35g) and a shorter travel distance (3.4mm) compared to the standard Cherry MX Red mechanical switches, so typing on the Apex 9 does feel a little too fluttery, and will take some getting used to — especially if you’re coming from a heavier, non-linear, mechanical switch background.

Gaming on the Apex 9 felt better than typing — unsurprisingly, since linear switches, optical or not, usually offer a better gaming experience than typing experience. According to SteelSeries, the switches have a response time of 0.2ms and zero debounce, and they certainly felt very speedy, especially in gaming mode, which allows for very light, quick keypresses.

Features and Software of the Apex 9 TKL

The Apex 9 TKL works with SteelSeries Engine, which is part of the SteelSeries GG software suite. The keyboard comes with preset shortcuts, however, for switching between gaming and typing mode (SS key + I/O), switching between on-board profiles (SS key + F9), recording macros (SS key + F10) and adjusting the brightness of the per-key RGB (SS key + F11/F12), so the software isn’t critical.

Using SteelSeries Engine, you can remap the keyboard’s primary and secondary (SS key) bindings, and set up the keyboard’s five on-board profiles. Almost every key can be remapped, including the SteelSeries function key, except for — curiously — the multi-function media key and clickable volume roller. Also, you can’t adjust each key’s actuation point individually (like you can on the Apex Pro Mini);you can only toggle the entire keyboard between gaming mode and typing mode.

You can also change up the Apex 9’s lighting in SteelSeries Engine. Each key can be customized individually with different colors and effects, for three different layers (active, reactive, and idle). While you can save different lighting profiles in SteelSeries Engine, you can’t link them to onboard profiles.

The Bottom Line

The SteelSeries Apex 9 TKL is actually a pretty attractive offering for anyone interested in optical switches with adjustable actuation — but who doesn’t need the over-the-top 0.1mm per-key adjustment offered by the Apex Pro Mini. The Apex 9 TKL’s larger layout is definitely a plus for anyone who finds a 60 percent layout too small — for those who don’t, the Apex 9 Mini is also available. The keyboard’s hot-swappable PCB is interesting but not nearly as versatile as it sounds, given that it’s only truly compatible with SteelSeries’ own yet-to-be-released optical switches.

Both the Apex 9 TKL and the Apex 9 Mini are significantly cheaper than the Apex Pro Mini, but that doesn’t mean they’re cheap. If you don’t need the Apex 9’s 2-point actuation hot-swappable optical switches, you may want to look for something cheaper, such as the TKL Redragon Vishnu K596 or the 60 percent Corsair K65 RGB Mini.

A PCB diagram of a high-end AM5 Asus motherboard has been shared online, according to a report by TechPowerUp (originally shared in a Tweet by @harukaze5719, which has since been removed). This motherboard appears to be an AMD version of the high-end overclocking Apex boards, based on its distinctive twin slot memory setup and beefy power delivery system.

If this is indeed an AM5 version of Asus' Apex motherboards, it's the first AMD AM5 board we've seen — the Apex brand has been Intel-only since its inception.

The leaked PCB diagram shows an impressive motherboard setup, the most prominent feature being the power delivery system. The entire VRM set is large, and wraps around nearly half the entire board, with a length nearly as long as the DDR5 memory slots on the left side. 

But what's most impressive about this VRM is the number of phases on the board — 27. We can't be sure if Asus is using doublers, but the number is impressive regardless. For some perspective: the Crosshair X570 Extreme has "just" 20 phases, and that board is Asus' flagship AM4 overclocker.

As a result, we can expect this next generation AM5 Apex to pack some serious power under the hood — with enough power delivery to support any type of overvolting and overclocking done on AMD's upcoming Ryzen 7000 series processors. 

These chips are expected to be noticeably more power-hungry than their Zen 3 predecessors, with a rumored peak power target of up to 230W and a max TDP of 170W. So the extra phases will probably be required for high-end overclocking.

The defining trait of the PCB design is the dual-slot DDR5 memory configuration on an ultra high-end ATX motherboard — a trait only found on overclocking-exclusive boards. This design choice alone suggests this board is an Apex model, since all Apex boards feature this unique memory design.

Switching down to dual memory slots can improve the board's memory overclocking capabilities substantially in many cases. Bringing down the slot count from four to two reduces the complexity of the board's traces and shortens them which improves signal integrity.

We don't know how this will impact DDR5 memory performance, but it should enhance this board's overclocking capabilities compared to current DDR5 motherboards on the market. Asus' Z690 Hero peaks at a maximum official supported frequency of 6400MT/s. 

We could see this new Apex board featuring support well into the 7000MT/s range, if not 8000MT/s. Prototype DDR5 memory modules have been known to hit well beyond 10,000MT/s in preliminary tests.

Other features worth noting are the triple PCIe x16 slots visible on the motherboard — the top two will probably be Gen 5 models, as AMD's AM5 platform has been confirmed to support the Gen 5 standard. But the secondary slot can clearly be seen as wired to just 8 lanes, due to lane limitations, on the CPU. Finally the third slot can also be seen as a physical x16 slot, but wired for 8 lanes. This slot should be wired into the chipset, and will feature Gen 4 or Gen 3 speeds.

Speaking of the chipset, this board undoubtedly features AMD's higher-end X670 chipset with the dual physical dies present on the PCB layout. AMD's new X670 and X670E chipsets use a "CrossfireX/SLI" setup consisting of two smaller chipsets (instead of one bigger chipset like in previous designs). 

But we cannot confirm whether this board will use the mid-range X670 chipset or the flagship X670E chipset. We assume it will use X670E, since this board looks to be a flagship overclocking platform in Asus' lineup.

We don't know for sure if this is going to be an Apex board, but the evidence points towards it being an Apex board for AMD CPUs. Hopefully we'll get more news about this mysterious motherboard as we get closer to the release date of AMD's Ryzen 7000 processors.

Today at GameStop, the SteelSeries Apex 7 TKL Mechanical Keyboard is available for just $96. This gaming keyboard usually stays around $130 across multiple vendors making this the best discount we’ve seen for it since it was first released. With Amazon Prime Day right around the corner, this discount is just one of many to appear in the wave of early Prime Day deals.

Overall, this is a quality gaming keyboard that’s definitely worth a look, especially at a sub $100 price tag. We had the opportunity to review the full-size model of the Apex 7 mechanical keyboard but this offer applies only to the TKL version. Both have similar specifications apart from the numeric pad. When reviewing, we noted our appreciation for its RGB support and sturdy frame construction but felt its OLED display did not add anything significant to the experience. The price was also a point of concern but with today’s discount, the features are much more notable.

This discount applies to the edition with Red SteelSeries QX2 Mechanical RGB switches which are linear and quiet with a 2mm actuation point and 50 million keypress guarantee. There are other editions available with clicky switches but those models are not included in this discount. The keys support 100% anti-ghosting and have 84 N-Key rollover support.

The SteelSeries Apex 7 TKL mechanical keyboard has per-key RGB LED support which can be programmed using SteelSeries Engine, a proprietary application for customizing various SteelSeries products. There are dedicated media controls in the upper-right corner including a clickable metal roller. The wrist rest is detachable with magnetic connectors.

Visit the SteelSeries Apex 7 TKL mechanical keyboard product page at GameStop to get a closer look at this discount. As of writing, it’s not clear for how long it will be made available.

For more early Prime Day offers and hardware discounts, check out our Amazon Prime Day hub which features the best discounts we can find on hardware and components of all types. We also have dedicated pages for things like the best monitor deals, best CPU deals, and even the best gaming laptops deals for those looking for an all-in-one machine with the latest hardware.

Asus ROG has announced that one of its motherboards has been used to achieve the DDR5 memory frequency world record. Overclocker Lupin No Musume achieved DDR5-10552 memory speeds using a system based around an Asus ROG Maximus Z690 Apex motherboard.

In the video below you can see Lupin No Musume verify the world record memory clocks in CPU-Z. To the left of the monitor display, you'll see the overclocker artfully kept the liquid nitrogen topped up to keep the system from overheating. To the right, an oscilloscope adds a further layer of verification to the memory clock achieved. We have seen overclockers add this kind of video evidence recently, to prove there is no frequency over-reporting software bug being exploited.

With Lupin No Musume's DDR5-10552 achievement firmly at the top of the table on HWBot, the top three positions  with DDR5-10K+ scores, come from Asus, MSI and Gigabyte, respectively. Check the table below to see the respective system configurations used for the top three records.

It's cool to see that three major motherboard brands can all achieve record-breaking performance, and no particular brand is dominant here. We just need to see DDR5-10K+ from an ASRock motherboard based system for a full-house. (Though ASRock has an entry at position 8.)

Another interesting point to see in the tabulated results, is that each overclocking expert took a slightly different tack with their processor configuration. You can see that sometimes performance cores were used, while others used efficiency cores, and at a range of clock speeds.

The newest and highest scoring DDR5-10552 world record score saw Lupin No Musume use four active Alder Lake processor P-cores, running at 3692.25 MHz. The single Asus ROG 16GB memory stick had a JEDEC rating of DDR5-4800, a lot slower than the record achieved.

We previously reported on DDR5 world records by both Hicookie, and Koven Yang, both from April this year. Both have improved on those scores in the meantime, but Lupin No Musume enjoys quite a clear lead as of this writing.

If you're motivated to get some higher-performing RAM now, check out our list of the  best RAM for gaming.

SteelSeries recently introduced its first ultra-compact, 60-percent form factor keyboards, the Apex Pro Mini and the Apex Pro Mini Wireless. 

SteelSeries’ Apex Pro line notably features the brand’s OmniPoint adjustable actuation switches, and these new mini keyboards are the first to sport the next generation, OmniPoint 2.0. In addition to adjustable actuation, OmniPoint 2.0 introduces dual actuation — a feature that lets you set two actions, triggered at different actuation points, in a single keystroke. 

As innovative as this all might sound, switches with adjustable actuation and dual actuation aren’t unique to SteelSeries. Razer’s Analog Huntsman keyboards, including both the full-size Huntsman V2 Analog and the 60-percent Huntsman Mini Analog (which dropped in March 2022) have switches with adjustable actuation points and dual-step actuation. 

Still, since the Analog Huntsman keyboards are, well…analog, the Apex Pro Mini Wireless seems to be the first wireless keyboard with dual actuation. But it doesn’t come cheap: SteelSeries’ Apex Pro Mini costs $180, while the Apex Pro Mini Wireless costs $240 — more expensive than almost every keyboard on our list of best gaming keyboards. 

Specifications

Design and  Construction of SteelSeries Apex Pro Mini 

The Apex Pro Mini is a 60-percent keyboard, which is the smallest standard keyboard layout you’ll commonly encounter (smaller layouts, such as 40-percent aren’t mainstream). The 60-percent layout is actually fairly popular, despite lacking a number pad, navigation keys, a function row, and arrow keys. This ultra-compact layout leaves you just enough keys for basic typing, and is great for traveling as well as reclaiming desk space. 

The Apex Pro Mini has a simple, elegant design — maybe it’s a little too simple, but, hey, there’s not a lot of room for creativity on a 60-percent board. It has a dark gray (or maybe “light black”) chassis, plastic with an aluminum alloy backplate, and black double shot PBT keycaps. On the back are two sets of flip-out feet for adjusting tilt; each set adds approximately 4mm (0.16 inches) in height, though, so it’s not a large adjustment. 

The keyboard is 11.53 inches wide, 4.02 inches deep, and 1.59 inches tall (at its tallest), which makes it a little taller than, but otherwise almost identical in size to, Razer’s Huntsman Mini.

SteelSeries’ branding is nicely understated as usual, with the full logo on the back of the board in dark silver, and just a small SteelSeries logo on the front left corner (well, and the logo on the dedicated SS function key).

Both the wired and wireless versions have a USB-C port in the back left corner and come with a seven-foot-long detachable, braided USB-C to USB-A cable. The wireless version has a power switch that toggles between 2.4 GHz wireless and Bluetooth, and comes with a 2.4-GHz wireless USB-C dongle as well as a USB-C to USB-C adapter.

The back of the wired version features a silicone flap that lifts up to reveal an orange keycap puller. The wireless version also comes with a keycap puller — just no place to store it thanks to the 30-hour battery. Both versions of the keyboards are the same size, but the wired version is a touch heavier — 1.34 pounds versus 1.2 pounds.

Both the Apex Pro Mini and the Apex Pro Mini Wireless will come in US English, French, German, Japanese, Nordic, and UK English layouts, but the US English layout is the only layout available at launch.

Typing Experience on the Apex Pro Mini

The Apex Pro Mini is the first keyboard outfitted with the company’s new OmniPoint 2.0 switches, which — like the original OmniPoint switches — have a linear feel. Pressing a key is smooth and seamless, with no tactile bump or resistance. 

Although the switches themselves are silent, typing on this keyboard produces some pretty loud clacking, as well as some slightly hollow-sounding thunks as you bottom out against the aluminum backplate. This isn’t necessarily a bad thing — I kind of liked the way this keyboard sounded. Like many writers, I prefer switches with tactile feedback for typing. But what I really like are clicky switches. The Apex Pro Mini is nowhere near as loud as a keyboard with clicky switches, such as my current daily driver, the Razer BlackWidow v3 with Razer Green switches, but it’s loud enough and clicky enough that the sound alone nearly made up for the lack of tactile feedback (nearly). 

The switches’ adjustable actuation doesn’t change how the keys feel, but it does change the overall typing experience. You can adjust the actuation from 0.2mm, which is so shallow that a stiff wind will trigger a keypress (seriously — at 0.2mm, I was able to type with some well-aimed blasts of canned air), all the way up to 3.8mm. If you really want to get into customizing the experience, you can adjust each key’s actuation point individually. 

I liked typing on the Apex Pro Mini, but I didn’t love it. I liked the way the double shot PBT keycaps felt — their slight texture helped keep my fingers from stray too far from the keys. That said, linear switches will always feel a little spongy to me, and the lack of tactile feedback did affect my typing speed and accuracy.

For reference, I tend to average around 120 words per minute (wpm), with around 98% accuracy, on a keyboard with tactile feedback. On the Apex Pro Mini, my fastest typing speed was 106 wpm with 96.3% accuracy — slower than my norm, though not by a lot. These results were achieved when I had the actuation set to the lowest (0.2mm). 

I also tested the keyboard with the actuation set to 1mm (103 wpm, 97.1% accuracy), 2mm (104 wpm, 96.8% accuracy), and 3.8mm (92 wpm, 89% accuracy). As you can see, my speed and accuracy dropped at the highest actuation setting — it appears I don’t usually type with full, 3.8mm-deep keypresses, and so I kept missing letters when the actuation was set higher.

Gaming Experience on the Apex Pro Mini

The Apex Pro Mini might not be my first pick for typing, but it’s much more attractive for gaming — which is what it’s made for, anyway. Both the Apex Pro Mini and the Apex Pro Mini Wireless are compatible with Windows, PlayStation, Xbox, and Mac OSX (though not all software features are supported on OSX). 

I had trouble getting used to secondary keybinds while I was using the Apex Pro Mini for writing, because they just didn’t feel intuitive to me (and the legends on the sides of the keys were hard to read, especially with the board’s bright RGB). When I’m typing, I expect the arrow keys to be…roughly where the arrow keys are on a full-size keyboard, not as secondary keybinds on W, A, S, D. 

Gaming, however, was a totally different story. The secondary keybinds were much more intuitive-feeling to me when I was gaming. And it didn’t matter which game — I went through my Steam library and played everything from Subnautica: Below Zero to Euro Truck Simulator 2 to Elden Ring (though I personally think Elden Ring is still best played with a controller). As a keyboard for typing, the 60-percent layout of the Apex Pro Mini is, honestly, just a little too small. But as a keyboard for gaming, it’s pretty good — I even adapted pretty quickly while playing Elder Scrolls Online, and I basically need two full keyboards’ worth of keys to feel satisfied in an MMO.

SteelSeries is really pushing the dual actuation on its new OmniPoint 2.0 switches. And it’s an interesting feature, once you figure out…what it is. 

It sounds relatively simple: Dual actuation — or, rather, dual action keypresses — lets you set two different actions, which are activated at different actuation points, for a single keypress. SteelSeries says this will let you do things like “walk and then run” or “pull out a grenade and then throw it.” 

It’s not quite as intuitive as it sounds, though. Dual action keypresses are not the same as function keys/secondary keybinds — you can’t activate the second action without activating the first. They’re also not just macros (though you can assign macros to either — or both — actions), because they work in a very specific order and the actions influence each other. 

Basically, there will definitely be a bit of a learning curve as you try to figure out exactly how to best implement dual actuation. I tried implementing the feature using SteelSeries’ example: Walk and then run. In Elder Scrolls Online, you walk (okay, default speed is more of a jog) using W, A, S, D, and you sprint by holding down Shift at the same time. I first attempted to put a second action (Left Shift) on the W key, thinking I could lightly press W to ‘walk’ and then hold it down to ‘run.’ This didn’t work, though, because the two actions are separate — the first action (W) doesn’t stay pressed when the second action actuates.

SteelSeries does give you some options for how the second action works, though — instead of just a single keypress, you can set the second action to play several times, repeat while pressed, toggle auto-repeat, or toggle hold. I first tried “repeat while pressed,” which didn’t work because repeating a key isn’t the same as holding a key (and pressing Shift five times in a row tries to activate Sticky Keys). But it also didn’t work because the first action also actuates on key release. This makes sense now that I think about it, but it wasn’t intuitive. You can choose whether the second action actuates on the key press or the key release, but because you’re passing the first actuation point twice, you’ll always get two actions. 

This is probably a little confusing, so here’s a straightforward example: Let’s say I bind a second action (‘K’) to the ‘O’ key, with the default action type (play once, on key press). If I press the ‘O’ key lightly and only actuate the first action, I will type ‘o.’ If I press the ‘O’ key harder, activating both actions, I will type ‘oko.’ This is because I actuate the first action twice, once on the way down and again on the way back up. I can change the second action type to play once, on key release, and now if I press the ‘O’ key hard, I’ll type ‘ook’ — the first action still actuates twice. 

The good news is that SteelSeries’ dual actuation feature is extremely customizable — in addition to keypresses, you can also assign device functions, media buttons, mouse buttons, and macros as the second action on a dual actuation key. I appreciate that SteelSeries allows you to create macros using numpad keys, which work even without a numpad (this is surprisingly difficult to find on TKLs). The OmniPoint 2.0 switches are also precise enough that I was able to press lightly and only actuate the top action with decent accuracy.

Features and Software on the Apex Pro Mini

To customize and configure the Apex Pro Mini, you’ll use SteelSeries GG, the brand’s software suite for all of its products. was able to test out an official update to the software, which is robust — though not perfect — and allows for extremely detailed fine-tuning of the keyboard. 

The Apex Pro Mini has onboard memory that allows up to five different onboard profiles in which you can customize all 61 primary keybindings, 60 secondary keybindings, adjust the actuation and dual actuation of all 61 keys individually, and program dual actuation keybinds — of which a maximum of eight are permitted. The GG platform makes this all pretty easy, and even lets you live-preview things like actuation level (which works…most of the time). 

The 60-percent layout means that just about every key does double-duty, and most of the Apex Pro Mini’s keys have preset secondary (or “Meta”) keybinds that are printed on the side of the keycaps that faces the user. They’re printed in white but do not allow for shine-through, which probably looks better overall but makes it difficult to see the functions. All keys except for the dedicated SS function key can be assigned a custom secondary keybind. 

Pressing the SS function key causes the keys with assigned secondary keybinds to light up in SteelSeries’ signature orange — a nice touch…even if it’s less useful in practice, when almost all of the keys have secondary keybinds.

You can adjust the actuation point of each key individually in SteelSeries GG using a slider or by typing in your desired actuation level. There are 37 levels, from 0.2mm to 3.8mm. The Apex Pro Mini’s keys have a total travel distance of 4mm, so 3.8mm is basically a full keypress. For reference, all Cherry MX and Kailh switches (except low-profile versions) have an actuation point of 2mm. SteelSeries helpfully points out that the switches on Razer’s Huntsman Mini Analog keyboard only have an adjustable actuation range of 1.5 - 3.6mm.

The Apex Pro Mini’s wider range does allow for more flexible customization when it comes to dual actuation, however. You can adjust the second actuation level of each individual key in the SteelSeries GG software as well, although this isn’t that useful, because you can only program a maximum of eight dual actuation keybindings at once. You can program these dual actuation keybinds on any key except for the SS function key.

Actuation point can also be adjusted manually using the keyboard (the default secondary keybinds are I and O; I increases actuation distance, O decreases it). You can set 10 different actuation levels, which you’ll see on the keyboard — numbers 1 through 0 will light up to indicate actuation level (1 is the shortest, 0 is the longest). This method trumps custom per-key actuation levels and adjusts all keys at once.

SteelSeries GG also lets you customize the keyboard’s per-key RGB. The software allows you to adjust each key’s lighting individually, with custom effects using multiple colors and different animation patterns. The software includes several presets with names such as “Prism” (rainbow), “West Coast” (blue-yellow-orange), and my favorite “Vapor Dreams” (pink-purple-blue-teal). I’m a sucker for the vaporwave aesthetic.

Okay, what I should say is that the software will allow you to do all that…hopefully, in the future. The illumination section of SteelSeries GG is sort of broken at the moment — I was able to adjust some of the per-key RGB on the Apex Pro Mini Wireless before it froze and left me with a blank screen, and I wasn’t able to do anything except cycle through presets on the Apex Pro Mini. SteelSeries also says it will allow you to set reactive lighting in addition to active lighting, but this feature isn’t yet available.

Bottom Line

SteelSeries’ Apex Pro Mini packs a lot into a tiny package. A lot. Maybe too much. Do you really need to be able to adjust the dual actuation point for each individual key, especially when you can only program eight dual-action keypresses at a time? 

While I’m not a big fan of the 60-percent layout, I’ll be the first to admit that this keyboard surprised me with how relatively intuitive and easy-to-use it was to use for certain activities, such as gaming. Gaming on the Apex Pro Mini was so seamless, I didn’t miss my function keys or navigation keys at all. (I did miss the numpad — though not while gaming — because I have an unhealthy dependency on the Alt code for the em-dash.) 

The Apex Pro Mini definitely feels like a bigger keyboard than it is, and that’s pretty impressive considering it is very, very small. It’s not my favorite keyboard for typing, but that wasn’t a surprise — linear switches are never going to be the writer’s favorite switch — and the clacky sound and lightly-textured keycaps do make it feel better. 

But if you’re not feeling all of the Apex Pro Mini’s many, many features, there are some very similar small form-factor keyboards that are much cheaper. Razer’s Analog Huntsman Mini even has adjustable actuation switches with dual-action keypresses — and if you don’t care about changing your actuation point, the HyperX Alloy Origins 60 has a very similar look and build for less than half of what the Apex Pro Mini costs. If you want something customizable — but on the hardware end, the Mountain Everest 60 has hot-swappable switches and can be purchased with a detachable numpad.

All that said, my biggest criticism about the SteelSeries Apex Pro Mini is that it gives you too many options and too many layers of customization…and that doesn’t really seem all that bad on such a limited layout.

PC components stalwarts NZXT and Alphacool have launched new thermal pastes. NZXT's new paste — its first-ever thermal paste — is dull-but-descriptively named NZXT High-performance Thermal Paste ($9.99/3g, $19.99/15g). Alphacool claims its new thermal paste, called Alphacool Apex Thermal Paste (€14.99/4g), is the best non-liquid metal thermal paste. Both thermal pastes are electrically non-conductive and are touted as easy to apply, with long-term stability.

The thermal compound that you put between your CPU and cooler (or in re-pasting your GPU, or laptop) can make a significant difference to cooling performance, so it's worth considering carefully when you build or maintain a PC. We tested over 90 different pastes (thermal and non — we even tested toothpaste) in our exhaustive round-up of the best thermal pastes, published earlier today. However, NZXT and Alphacool's new products have somewhat spoiled the comprehensiveness of our coverage — we'll have to test them once we get our hands on them.

Over recent years there have been some esoteric pastes made with materials such as diamond and liquid metal (which can be hazardous due to electrical conductivity and chemical reactions with aluminum). These new pastes from NZXT and Alphacool are more traditional: non-electrically conductive, non-corrosive, and easy to apply and clean off.

For a better handle on the new thermal pastes and how they compare, take a look at the specs.

The new Alphacool Apex thermal paste has a much higher claimed thermal conductivity value, and its marketing material asserts it is "one of the best performing thermal pastes on the market," which is only beaten by liquid metal alternatives. Meanwhile, the slightly cheaper (much cheaper if you buy it in bulk) NZXT High-performance Thermal Paste has thermal conductivity on a par with PC-builder-favorite Arctic MX-5 — so its performance is by no means poor.

NZXT says its paste uses zinc oxide, liquid polymer, and aluminum material. Alphacool isn't as forthcoming with its ingredients, and instead says that its thermal paste is formulated using "a nano-powder with a very low thermal impedance." So, it looks like Alphacool's new paste relies on nanotechnology-boosted thermal conductance.

If you're not sold on NZXT's first foray into the market or Alphacool's nanotech-infused secret sauce, we suggest going for one of the picks in our best thermal pastes round-up (not toothpaste).

The SteelSeries Apex 3 RGB gaming keyboard offers plenty of options for gamers who like customizable features and multi-platform compatibility. This wired gaming keyboard has a variety of specs to take advantage of ranging from customizable LEDs to the ability to program and save custom macros into separate profiles. The going rate for this keyboard is $49, but right now it’s listed at just $34.

According to SteelSeries, the Apex 3 is an IP32 water-resistant keyboard which also gives it an extra level of protection against dust. The Apex 3 is full-sized with a numeric pad off to the side. While its keys aren't mechanical, they are rated to 20 million keypresses.

The Apex 3 gaming keyboard is capable of saving custom macro combinations to individual profiles and has a key dedicated to swapping between these profiles on the keyboard. There is also a dedicated multimedia button alongside a volume rotary encoder for media.

Rather than per-key lighting, the SteelSeries Apex 3 has 10-Zone RGB LED support. These zones can be programmed with custom colors or effects and saved to individual profiles. The brightness of the backlight LEDs can also be adjusted. It has a magnetic wrist rest that can be easily removed and reattached.

Visit the SteelSeries Apex 3 gaming keyboard product page at Amazon for more details and purchase options and remember to take a look at our list of Amazon promo codes. If you’d like to see what else the market has to offer right now, check out our list of Memorial Day deals and other various tech deals. We’ll be updating these lists as the day progresses so be sure to check back regularly to get an idea of what else is available throughout the holiday sale.

Being the world's largest contract chipmaker, TSMC is the largest beneficiary of the ongoing semiconductor boom. The company expects its sales to continue growing for many years to come, so it plans to raise its capital expenditures (CapEx) by 33% — 46% this year all the way to $40~$44 billion — to support continued growth. This level of spending will be particularly useful as the company prepares to start making chips using its N3 (3nm class) and N2 (2nm class) processes in the coming years.

Preparing N3 Ramp

TSMC's 2021 revenue reached an unprecedented $56.82 billion, up 24.1% year-over-year due to extraordinary demand for semiconductors in general, and chips made using leading-edge fabrication technologies in particular. In addition, TSMC projects demand for its services to accelerate going forward because the company is the largest foundry in the world, and it is the leading provider of mature, advanced (FinFET), and leading-edge (7 nm-class and below) nodes. Therefore, to boost its capacities for all types of manufacturing processes (including the upcoming N3), the company intends to increase its CapEx from $30 billion in 2021 to $40 – $44 billion in 2022. To put that number in perspective, TSMC's CapEx budget in 2019 totaled $14.9 billion. 

Approximately 70% – 80% of TSMC's $40 to $44 billion CapEx will be allocated to building new fabs and expanding capacities for the latest and upcoming nodes, including N2, N3, N5, and N7. For example, TSMC spent quite a fortune building up capacity for its N5 nodes (including N5, N5P, N4, N4P, N4X) last year. This year it will spend billions on building N3 and N2-capable fab(s).  

TSMC will also spend 10% to 20% on specialty technologies and another 10% on facilities for advanced packaging and mask making.  

Since the industry is in the midst of several major megatrends (5G, AI, HPC, digitization, etc.), TSMC and other foundries are enthusiastic about future growth. Moreover, since TSMC is arguably the most competitive contract maker of semiconductors, it is not surprising that the company is particularly optimistic about landing orders both from existing and new customers with its N5 and N3 nodes.

Leading Nodes Account for 50% of Revenue

TSMC's N7 and N5 nodes accounted for about 50% of TSMC's $15.74 billion wafer revenue in Q4 2021 (a 24.1% increase year-over-year and up 5.8% sequentially), based on the numbers provided by the company. The two fabrication technologies also accounted for 50% of TSMC's $56.82 billion wafer revenue for the whole year. In 2020, they accounted for 41% of TSMC's $45.51 billion revenue. 

As far as applications are concerned, sales of smartphone chips dropped from 51% of TSMC's revenue in Q4 2020 to 44% in Q4 2021. Meanwhile, revenue share of CPUs, GPUs, accelerators, and SoCs used in such applications as PCs, servers, supercomputers, and game consoles (which TSMC calls high-performance computing (HPC) applications) upticked from 31% in Q4 2020 to 37% in Q4 2021.

"Our fourth quarter business was supported by strong demand for our industry-leading 5-nanometer technology," said Wendell Huang, VP and Chief Financial Officer of TSMC. "Moving into first quarter 2022, we expect our business to be supported by HPC-related demand, continued recovery in the automotive segment, and a milder smartphone seasonality than in recent years." 

TSMC charges a premium for its N7 and N5 production because these technologies require advanced tools and provide substantial transistor density, power, and performance benefits. As fabless designers of chips adopt the new nodes, it is not surprising that the revenue share contributed by these nodes increases, just like TSMC's earnings in general. Meanwhile, TSMC's revenue growth is also fueled by price increases for the production of trailing nodes.  

TSMC expects its Q1 2022 revenue to be between $16.6 billion and $17.2 billion (up sequentially and year-over-year) and its gross profit margin to be between 53% and 55%.

More Orders for Leading-Edge Nodes Incoming

Last year, TSMC said it would spend $100 billion over the following three years on new manufacturing capacities and research and development (R & R&D). If the company keeps increasing its CapEx in 2023, then TSMC's investments in CapEx and R&D will exceed $100 billion in the 2021 – 2023 period quite significantly. But there are good reasons behind this move. 

While the share of advanced and leading-edge nodes in TSMC's revenue has been increasing in recent years, it is noteworthy that two large customers — Nvidia and Qualcomm — began to outsource a significant portion of their sophisticated chips to Samsung Foundry in 2020. Yet it does not look like either are particularly happy with the outcome, which is why both are rumored to be returning to TSMC in 2022 with their upcoming designs. 

Considering that Intel is also beginning to outsource some of its most advanced components to TSMC, it is evident that the foundry is on track to significantly increase production using its N5 fabrication process in 2022 and then N3 manufacturing technology in the 2023 – 2024 period. In addition, while there are rumors that AMD and MediaTek plan to adopt a multi-foundry/dual-source approach with some of their products going forward, both companies are already on long-term agreements (LTAs) with TSMC (with some prepayments made), so they are not reducing their orders to the No. 1 foundry in the foreseeable future. 

N5 will be a very long-lived node for TSMC both because of decent yields and ample capacity as well as various performance and die size-optimized versions of the process. Industry observers believe that N3 will be another win for the contract manufacturer as Samsung Foundry's plan to adopt gate-all-around transistors for its 3GAE/3GAP nodes adds uncertainties. It also remains to be seen how competitive Intel Foundry Services will be in 2024 – 2025.  

"The return of Nvidia Qualcomm to adopt N3 in 2023 – 2024 speaks volumes about TSMC as the only credible advanced node foundry, in our view," Szeho Ng, an analyst with China Renaissance, wrote in a note to clients. "We believe that Nvidia will start sole sourcing some of its N5 products (next-generation HPC server GPU – 'Hopper' H100 signifying its foray into MCM/multi-chip module design) from TSMC this year. […] The next generation gaming GPU (GeForce RTX 40-series) may, however, still be dual-sourced from TSMC and Samsung on N4/5 for diversification purposes, in our view."

Summary

While TSMC's plan to spend up to $44 billion on new capacities seems bold (since this is the largest sum ever spent on new fab capacity by one company in a year), TSMC has a lot of motivations to be optimistic about future demand to expand its capabilities. 

The ongoing megatrends (5G, AI, HPC, digitization, etc.) will drive demand for all types of semiconductors quite significantly in the coming years. Furthermore, there are three major customers — Intel, Nvidia, and Qualcomm — that intend to adopt TSMC's advanced and leading-edge nodes for their high-volume products in the 2022 – 2024 timeframe. Finally, TSMC's existing top customers (Apple, AMD, MediaTek) that have been increasing their market shares quite steadily in recent years will require more capacity in 2022 – 2023 to support their growth. 

Constant chatter with your friends during extensive gaming sessions or Zoom meetings calls for proper hydration, but every time we bring liquids to our desks, we take on a big risk of destroying our keyboards with one little spill. Fortunately, SteelSeries eliminates that risk with its new Apex 3 TKL gaming keyboard, which is IP32 water resistant.

This keyboard also features shockingly good Whisper-Quiet membrane switches, and it's rare to find a typing experience you love for just $45.

SteelSeries Apex 3 TKL Keyboard

Design

When I first sat down to review the Apex 3 TKL, I didn’t know what to expect given its $45 price tag. I thought it was either going to be hot garbage or, at best, good but not great. But the design of this keyboard proved to me that the best gaming keyboards don’t need to cost an arm and a leg.

This entire keyboard is composed of slightly glossy plastic, but it doesn’t attract fingerprints nearly as much as competitors like the Wooting Two HE, which is impressive given that it costs less than half as much as Wooting’s board.

The plastic build is a featherweight. With a weight of just 1.41 pounds (639 g), it’s a bit too light for its own good, as its lack of heft caused its small ten-keyless frame to occasionally move around on my desk during use.

There are cable routing trenches on the underside, which I really appreciated because the Apex’s USB-A cable is non-removable. The Apex 3 also has two flip-up feet on its underside to help give you a comfortable typing angle. I appreciated that these were coated in rubber, but while that does give them some extra grip, it isn’t enough to counteract the keyboard’s lightness.

At first, I was a bit bothered that the Apex 3 lacked a removable cable, but that may be because the keyboard is IP32 water-resistant, and a detachable cable could be another way for water to get in.

To test the keyboard’s water resistance, I filled a cup of water, and with the Apex 3 connected to my laptop, I spilled water on the keyboard. After giving the Apex 3 a quick wipe-down with a napkin, it worked as usual, which is very cool and can be very practical if you’re clumsy.

There isn’t a whole lot I don’t like about the Apex 3’s design, but I didn't appreciate that the company flipped some of the legends around. For example, the 1 is above the exclamation mark, where it typically is underneath it. 

Unlike too many gaming keyboards, we have dedicated media keys here. While they’re a bit awkward to operate, they are better than nothing. 

On the top right of the keyboard is a notched volume wheel and a little black square that controls your media. If you press the square once, it’ll pause your media; a double press will skip a track and a triple press will go back to the previous track.

I will say that the media “square” works well until you have to press it three times to go back to a previous track. I can deal with two presses, but three clicks felt very awkward.

Typing Experience on the Apex 3 TKL

The SteelSeries Apex 3 TKL features the company’s Whisper-Quiet gaming switches that are rated for 20 million keystrokes, which is a lot less than the other switches we test. But for $45, I think 20 million strokes is more than enough.

Like their name implies, these switches are silent, and that’s partially due to them being membrane, rather than mechanical. Before you jump ship, let me tell you that I actually really enjoyed these switches, and I do not usually enjoy tactile bumps that much.

Typing with the Apex 3 was relaxing; I am so used to fat tactile bumps and ASMR-like linear switches that the feeling of membrane switches feels foreign. Unlike most membrane switches, though, the Whisper-Quiet switches do have some tactility to them, which made the typing experience feel elegant and responsive.

The only unpleasant noise I experienced with this keyboard was with the stabilizers, and to be honest, they didn’t sound that bad. It’s just that, because the switches are so silent, the stabilizer rattle is more noticeable on this board than on most other boards. But for $45, what do you expect?

Gaming Experience on the Apex 3 TKL

SteelSeries has proven itself when it comes to gaming peripherals, so even though the Apex 3 TKL is $45, I had relatively high expectations when it came to in-game performance.

Funny thing is, I performed better with the Apex 3 in-game than I did on any other keyboard I have reviewed. On Call of Duty: Black Ops Cold War, I got to round 50 on Firebase Z, and to give some perspective, round 40 is where the extra powerful Wonder Weapons become barely more effective than a normal rifle. I’m sure my friend and I could’ve made it to round 55 at least, but it was over two hours and we hit our record so it was time to exfiltrate.

The Apex 3’s membrane nature didn’t take away from its capability on the battlefield even though I had become so accustomed to mechanical switches, which is evident in my performance in Zombies. If I were to compare the Whisper-Quiet switches to mechanical switches, I would say that they felt a bit similar to Zeal Zilents, which are silent tactile switches.

Software

SteelSeries’ Apex 3 TKL operates on its SteelSeries Engine, and right off the bat I noticed it looks almost too much like the Epic Games software.

This keyboard’s software is pretty similar to Razer’s Synapse software as it lets you add some apps to it like PrismSync, which can be used to match your Apex 3’s lighting to any of SteelSeries’ compatible peripherals and even some motherboards. Unfortunately, my motherboard wasn’t cool enough to work with PrismSync, but that’s alright.

The Apex 3 has six onboard macro-keys. I set up one of them to boot up OBS studio, which allowed me to conveniently start streaming the trek to round 50 on Cold War. 

You can adjust the RGB on this board, but it is not on a per-key basis. Instead, it’s on an eight-zone basis, like a pizza! There are eight segments on the Apex 3 where you can adjust the RGB, and I think this is fair and looks great.

Besides that, SteelSeries Engine lets you remap any key you want and has many other features, like System Monitor which provides real-time thermal readings but is more suited for higher-end boards like the Apex Pro.

Bottom Line

I was at Staples a few days ago to pick up over-priced printer ink, and on my way out I took a look at the prices of the basic membrane office keyboards they had. And a lot of them either matched the price of the Apex 3 or were more expensive. 

After looking at these keyboards, I realized that the Apex 3 would be a great keyboard to bring to work even if you’re not a gamer because it is extremely silent, dirt cheap and water-resistant.

The SteelSeries Apex 3 is ultimately a very humble gaming keyboard because it doesn’t try to do too much and it doesn’t bring anything groundbreaking to your battlestation. But it does do more than I anticipated, and I love it for that. The Whisper-Quiet membrane switches are very relaxing to type with, the plastic construction is very solid and it doesn’t creak like a door in a horror movie like in other boards I’ve used. 

If SteelSeries just adds a bit of weight to it and figure out a better media key implementation, then we could have the perfect budget gaming keyboard.

Motherboard partners make boards in all shapes, sizes, and purposes. From Mini-ITX to E-ATX and budget offerings to flagships, there is likely something for every build type and budget. Today we’re focusing on Intel Z590 boards designed primarily for overclocking. 

All of the tested boards include some of the most capable VRMs for the platform, only two slots for DRAM, and a slew of buttons and switches designed to make overclocking--especially when using sub-ambient cooling--easier. That said, these boards are also great for more general-purpose builds, but at a significant premium. Specifically, we’ll be looking at the ASRock Z590 OC Formula ($579.99), Asus’ ROG Maximus XIII Apex ($599.99), and the Gigabyte Aorus Tachyon ($529.99), pitting them against each other in our testing environment.

If you’re in the market for a premium motherboard with a focus on overclocking, you’ll want to look for high-quality power delivery, features (like buttons, switches and software) that make overclocking more manageable, and maybe even consider how easy it is to insulate the CPU socket area (for the sub-ambient crowd). 

The latest and greatest audio, fastest network hardware, loads of storage/USB options, and appearance tend to be further down the list of importance here. Remember, you’re after the highest CPU and memory clock speeds possible, be it for competitive benchmarking using Liquid Nitrogen, or maybe you just want the last bit of performance possible out of your system. These boards are made for speed.

That said, going this route isn’t without some sacrifices. A key difference is with RAM capacity. Since all three of these boards have just two DRAM slots, RAM capacity is limited to 64GB. While this is plenty for benchmarking, gaming and general use, it may not be enough for those with high-end productivity and creative aims. Outside of that, these boards are pretty well-rounded for a wide variety of users. But if your primary objective isn’t extracting every last bit of performance out of your machine, you’re overpaying for features you won’t use with these boards.

Overall, the three boards we tested performed well, easily mixing in with other Z590 options we’ve tested--and were slightly faster in many tests. In the end, none of the three overclocking boards was more than negligibly quicker than the others in our standard tests, and they traded punches throughout, including in games. Power use varied, with the Asus using the most, but the difference between the three wasn’t significant. 

There are also a lot of similar features in this board class. All three boards we tested here include at least three M.2 sockets (the Apex has four), six or more SATA ports, one USB 3.2 Gen2x2 Type-C port, two RAM slots, a high-end audio codec, Wi-Fi 6E and 2.5 GbE, along with a slew of overclocking specific hardware and features. With a  roughly $70 price difference between the contenders, we’ll sort through the details below to see if one board stands tall over the others.

Specifications - ASRock Z590 OC Formula 

ASRock’s Z590 OC Formula includes some accessories along, with the motherboard in hopes of getting you started without a trip to the store. The accessory stack is slim compared with the others (and for the price) but should get what you need. Below is a complete list of all the extras.

• Support CD / Quick installation Guide
• Wi-Fi Antenna
• (4) SATA cables
• (3) Screws/Standoff package for M.2 sockets
• ASRock screwdriver

After taking the OC Formula (OCF) out of the box, we see a matte-black, 12-layer server-grade PCB, along with light green highlights on all of the heatsinks. There’s a small color OLED screen in the middle of the board, where you typically find an M.2 socket, that displays status information. The M.2 socket normally in this area was shifted to the right of the DRAM, runing vertically on the board. To the right are the OC buttons used for overclocking on the fly. Unlike the other boards in the roundup, the OCF takes a minimalist approach to the heatsinks, which makes sense considering the overclocking nature of the board. The upper-right corner of this E-ATX size board has a notch cut out and an RGB-illuminated plastic shroud, displaying the OC Formula name. In addition, the chipset heatsink has bright RGB lighting below the heatsink, illuminating that area of the board.

Overall, the OCF isn’t here to look good, but to be fast. The color may not fit in some build themes, but it’s a unique style that some will enjoy. I can’t hold the looks against it, as the board's intent is function over form.

Focusing on the top half of the motherboard, we get a closer look at the oversize heatpipe-connected heatsink for the VRM, along with the green accent pieces on top. Actively cooled, the VRM uses two small fans on the top, finned heatsink, and a larger fan hidden in the left bank. At default speeds, the fans were audible but not imposing during our testing. Wedged between the two VRM heatsinks are two 8-pin EPS connectors to feed power to the CPU. The socket area is relatively clean, though the standard cylinder-shaped capacitors are always a pain to insulate when preparing for sub-ambient cooling runs.

To the right of the socket area are two DRAM slots, capable of supporting 64GB of DDR4 RAM. With the board designed for overclocking, most users aren’t going to find this ceiling too low. If you want to use this board as a daily driver, 2x16GB sticks are enough capacity for most tasks, and you can still upgrade to 2x32GB. In addition, the board lists support up to a whopping DDR4 6000+(OC), which is the highest of our three tested boards here. As always, your mileage may vary to reach these speeds. You’ll likely need a hand-picked CPU with a killer memory controller, a suitable memory kit, and perhaps some Liquid Nitrogen to reach that lofty goal.

To the right of the DRAM slots, things get busy. The CPU-connected PCIe 4.0 x4 M.2 SSD slot (remember you’ll need a Rocket Lake-based CPU for that bandwidth) is just to the right, running vertically while supporting up to 80mm PCIe-based modules. The first two RGB headers (ARGB and RGB) and the first two 4-pin fan/pump headers are found just above that. There’s plenty of power to dole out on these headers, with the CPU supporting up to 1A/12W, CPU_FAN2/WP_3A header outputting up to 3A/36W, and the chassis/water pump fan up to 2A/24W. There’s plenty of headers and capacity to run your watercooling from the board itself here.

Continuing right gets us to the OC area, where many will spend time on this board. There’s your basic power and reset buttons, rapid OC buttons (labeled + and -), and three buttons for Nick Shih’s Overclocking profiles. Additionally, there are switches for LN2 and Slow modes, a retry button and a set of DIP switches to disable PCIe slots on the fly. For those into sub-ambient cooling, these are the tools of the trade that save you from frequently booting into the bios to make some changes and helps push higher when you’re on the very edge of stability.

Below that, towards the middle of the board, is a front-panel USB 3.2 Gen2x2 Type-C, header along with two USB 3.2 Gen1 headers (one of which is oriented vertically).

For the Z590 OC Formula, ASRock chose a 16-phase (8-phase doubled) configuration for Vcore. Power comes from the EPS connectors to a Renesas ISL69269 12-channel (X+Y+Z=12) controller, onto several Renesas ISL661A phase doublers, and finally to the 16 90A ISL99390B Smart Power Stages. The 1440A available for the CPU is plenty for stock operation and sub-ambient overclocking using the Z590 platform’s flagship Core i9-11900K processor.

The bottom half of the board is, for all intents and purposes, fully exposed. On the left is the Realtek ALC1220 audio codec, exposed along with several audio caps (yellow) and four WIMA capacitors (red) for the front panel. Surprisingly, this board uses an ESS9218 SABRE DAC to improve audio for the headphones (auto-sense up to 600 Ohms). Since this is an overclocking-based motherboard, I can’t ding ASRock for using the last-generation codec, especially since it’s still suitable for most users.

Shifting focus over to the middle of the board, there are five PCIe slots, two M.2 sockets, and a small color OLED screen to cover. ASRock says the company added the OLED to give a cleaner and informative way to express the status of the motherboard. The display is configurable in the BIOS. I thought this was a bit extra for an overclocking motherboard. Still, in thinking about it, I often want to know what my voltage and temperatures are without having to jump into Windows or have additional software running. For the average user, it’s still an informative display that covers voltages, speeds and power use (amperage), among other configurable items.

For PCIe connectivity, ASRock has three full-length slots and two x1 size slots. The full-length slots are reinforced and connected through the CPU. The top slot (PCIe1) supports up to PCIe 4.0 x16 (when using a Rocket Lake processor), while the second and third slots (PCIe4/5) run at x8 and x4 speeds, respectively. This configuration supports up to three-way CrossfireX. SLI is not mentioned. The two x1-size slots are fed from the chipset lanes, running at PCIe 3.0 x1 speeds. Note that when you use any graphics card that is more than single-slot width, it covers the first x1, leaving only one available.

M.2 storage in this area consists of two chipset-attached sockets. Remember, the first is CPU-attached and runs vertically next to the DRAM slots. The first socket in this location, M2_2, runs at a maximum of PCIe 3.0 x4 and supports PCIe and SATA modules up to 80mm. The bottom socket, M2_3, supports PCIe and SATA as well, but up to 110mm devices fit. There is some lane sharing with these sockets. When M2_2 is in use, SATA ports 0/1 are disabled. When M2_3 is in use, SATA ports 3/5 are disabled. In other words, you can run up to three M.2-based devices and still have four SATA ports available. This should be plenty, especially for systems focused on overclocking.

Continuing our board tour, we skip over the chipset heatsink, illuminated from below with bright RGB LEDs, and land on the eight SATA ports. Six of these connect to the system via the chipset, while the other two use an ASMedia controller. Any SATA-connected device supports RAID0, 1, 5 and 10, while the M.2 side supports RAID0 and 1. Just below the SATA ports are two more 4-pin fan headers.

Across the bottom are several headers, ports, and switches. We list these below from left to right.

• Front panel audio
• Thunderbolt AIC
• RGB and ARGB headers
• (3) fan headers
• Voltage read points
• (2) USB 2.0 headers
• 2-character debug LED
• Front panel header

As we swing around to the pre-installed rear IO plate, you’ll notice that it’s mostly standard fare. There’s a total of eight USB ports (3x USB 3.2 Gen2, 1x USB 3.2 Gen2 Type-C, and four USB 3.2 Gen1 ports), all of which have ESD protection. Above the USB ports are the two Intel Ethernet ports. We also spy the Wi-Fi antenna ports, along with two legacy PS/2 ports (good for working with older OSs in the benchmarking world). If you plan on using the integrated video on the CPU, you’ll have to use a different board as there aren’t any video outputs here. Convenient for the overclocker is a CMOS reset button, BIOS flashback button and a BIOS switch to swap between BIOSs. Last but not least is the standard 5-plug analog + SPDIF audio stack. These plugs also have LEDs in them that match the port color. 

Firmware

The BIOS theme in the OC Formula is as unique as the accent colors on all of the heatsinks. And although the layout is the same as other Z590 motherboards, the color is a bit different. The writing is still white on black, with a light blue showing the active part. The difference comes in the background theme matching that zombie green color scheme. ASRock includes an Easy Mode for high-level monitoring and adjustments, along with an Advanced section. The BIOS is organized well, with many of the more commonly used functions accessible without drilling down multiple levels to find them. For example, here, you adjust the Memory, Voltage, and CPU details in separate sections, but it’s all on the first page of each section. In the end, the BIOS worked well and was easy to navigate and read.

Software

On the software side, ASRock includes a few utilities that cover overclocking and monitoring (Formula Drive), audio (Nahimic 3), software for updating drivers and downloading applications (App Shop), and of course, RGB control (Polychrome RGB). There’s also an OLED control application (ASRock OLED Control) to configure the small display. Again, we did not run into any issues in our limited use of the applications.

Before we get to the performance of this board and its competitors, we’ll detail the other two models as well. Next up is the Asus ROG Maximus XIII Apex.

MORE: Best Motherboards

MORE: How To Choose A Motherboard

MORE: All Motherboard Content

The $599 Asus ROG Maximus Apex is $20 more expensive than the OC Formula and $50 more than the Gigabyte Aorus Tachyon, the two competing boards that fill out this roundup. To determine if this Asus is worth the extra money, we’ll have to dig deep into the features. First, here’s the complete list of specs for the ROG board, direct from Asus.

Specifications - Asus ROG Maximus XIII Apex 

Along with the ROG Maximus Apex, Asus includes several accessories in the box. These range from stickers and keychains to extension cables, the DIMM.2 expander module and more. Overall it’s an robust set of accessories to get you started. Below is a complete list of all the included parts that come in the box with the board.

• ARGB/RGB extension cables
• (8) SATA cables
• (1) 3x1 thermistor cables
• ROG DIMM.2 / pads / screw package
• Wi-Fi Antenna
• M.2 screw package
• Q-connector
• ROG Key chain/stickers/thank you card
• ROG graphics card holder
• Support DVD
• User Manual

At first glance, the Maximus XIII Apex looks like most other boards. It has a matte-black PCB and large heatsinks covering the power delivery area, top M.2 socket and most of the board’s bottom. The heatsinks have large grooves to increase airflow and surface area. A U-shaped heatpipe connects the heatsink to distribute the heat across the entire array efficiently. The black and white styling gives way to RGB LEDs under the IO heatsink, the ROG symbol above the chipset and an acrylic piece with ROG letters stenciled in white. There’s also some coordinates (Longitude and Latitude) written in this space. When entered into Google Maps, it takes you to an Asus/Pegatron building just northwest of Taipei City.

Overall, I like the look of the Apex. It leaves those who want to use this in a chassis or open bench a lot of freedom to pick your theme, as there are no other accent colors outside of the minimal RGB lighting above the rear IO area and chipset. That said, for the extreme overclocker who plans to use a pot for dry ice or LN2, there’s not a lot of room around the socket area, and several capacitors to get in the way. But for a daily driver, this won’t matter and provides ample cooling for the VRMs.

As we look closely at the top half of the board, you can see through part of the VRM heatsink that houses the RGB lighting. This configuration yields a cool effect if you’re not looking straight on, with the light shining through the cutouts. The silver accent piece also sports ROG branding, while the Apex name graces the top of the M.2 cover. We spy the first of four 4-pin fan headers here (eight total) on the top edge. All fans output 1A/12W, except for the W_PUMP+ header capable of 3A/36W. All chassis and CPU headers are Q-Fan controlled, while the FS and PUMP headers are all full speed. Above the left VRM bank are two 8-pin EPS connectors (one required) to send power to the CPU. Between these and some fan headers is an LN2 mode jumper that expands the voltage range you’re able to use.

To the right of the socket area are two DRAM slots and the DIMM.2 slot. The Apex supports up to 64GB of DDR4 at speeds up to 5333+(OC). This is less than the OC Formula but still in the “YMMV” category, where you need a CPU with a great IMC and the right memory kit to reach those speeds (and more). The DIMM.2 socket to the right adds two M.2 sockets via a separate board, capable of running in SATA or PCIe modes. Just above are the three RGB headers (2 ARGB, 1 RGB).

Along the right edge, a lot is going on, especially for the overclocker. Working our way down from the top, we see the debug LED and the safe boot button. Below that are three more buttons, Start, Flex Key and Retry. The four switches here are enable/disable, V Latch, Slow Mode, Pause and one that is labeled RSRVD. Next to that is a set of holes used for voltage read points covering Vcore, VccSA, VDIMM, and other essential voltages -- a must for the extreme overclocker (and great for any enthusiast). Below that is the 24-pin ATX power connector for the motherboard, another fan header and finally, a front-panel USB 3.2 Gen2x2 Type-C header.

Designed for overclocking, the Apex (like all the boards in this roundup) has a beefy VRM capable of handling any compatible CPU and taking it to ambient and sub-ambient limits. Power comes from the 8-pin EPS connectors (one required) and is fed to a Renesas ISL69269 12-channel controller, then on to 18 ISL99390 90A Smart Power Stage MOSFETs setup in a teamed (no doublers) configuration. This yields 1620A available for the CPU, which is one of the higher values we’ve seen overall, and the highest in this bunch.

Moving on to the bottom half of the board, we find the audio section on the left exposed for all to see. Hidden under the shiny faraday cage is the Supreme FX (Realtek) ALC 4080 codec and several small Nichicon audio capacitors. In addition to the board and channel separation is a Savitech SV3H712 amplifier said to improve THD+N performance for improved audio quality. The 4080 codec is the current flagship from Realtek and provides more than adequate sound for most users. This is the only board in this roundup that uses the latest and greatest for integrated audio codecs.

Sliding to the middle of the board, you’ll find two M.2 sockets, as well as two full-length PCIe slots and a single x4 size slot. The M.2 sockets are connected through the CPU and offer up to PCIe 4.0 x4 bandwidth using a Rocket Lake-based processor. Both support up to 80mm modules. Note, though, that there is plenty of bandwidth sharing and bifurcation going on. Instead of typing it all out, we’ll let the image below from the user manual show you.

The two reinforced full-length PCIe slots both use the CPU lanes. The top socket is wired for a maximum of PCIe 4.0 x16, while the second is up to x8 speeds. With an M.2 module installed in the bottom socket, the second slot breaks down to PCIe 4.0 x4. At the same time, the primary GPU slot goes to PCIe 4.0 x8 speeds. This isn’t a big deal on the graphics performance front, and it yields additional flexibility for storage. The x4 slot connects to the chipset, sharing lanes with SATA ports 4/5. In total, you can run up to four M.2 modules and four SATA-based drives simultaneously here.

Moving right, we jump over the chipset heatsink sporting the ROG branding, to the right edge where there’s a notch cut out of the board. Here you find the eight SATA ports (six connected to the chipset, two from the ASMedia controller). Below that is a front panel USB 3.2 Gen1 port. 

Across the bottom of the board are a slew of headers and buttons from USB to temperature headers; a lot is going on here. Below we’ve listed everything here, from left to right.

• Front panel audio
• Auxiliary 6-pin PCIe connector
• ARGB header
• (3) USB 2.0 headers
• TPM header
• RSVD and FS Switches
• Water flow and temperature headers
• (3) fan headers
• Thermal sensor
• Front panel headers

Swinging around to the rear IO, we’re presented with a pre-installed IO plate that matches the Apex’s black color scheme. Out back, there’s a total of 10 USB ports (3.2 Gen2x2 Type-C, 5x 3.2 Gen2, 4x 3.2 Gen1) and the single 2.5 GbE above them. On the left are clear CMOS and BIOS Flashback buttons, while just to the right are two legacy PS/2 ports for a mouse and keyboard. Finally, on the right side is the Wi-Fi 6E antenna and the audio stack (5-plug plus SPDIF).

Firmware

Asus’s BIOS for Maximus XIII Apex looks the same as is found on the previous generation Apex. Asus uses a red background with white writing on these ROG boards -- a theme we’re familiar with. Asus starts in “EZ” mode, where it presents some high-level information and a couple of adjustable options such as enabling XMP and Boot priority. Advanced Mode has a slew of BIOS options, many of which most users will not touch (although if you aren’t going to dig deep here, you should probably choose another board). Asus organizes the BIOS well, and many of the common functions are easily accessible without drilling deep into the pages. Overall, Asus offers a user-friendly BIOS with enough options to keep the most advanced user busy.

Firmware

On the software side, Asus includes a few applications designed for several functions. This includes RGB lighting control, system monitoring and overclocking, audio and more. Below we’ve captured a few screenshots of Ai Suite, the Armoury Crate and Aura Creator.

Before we get to testing, we have one last board to detail: Gigabyte’s Z590 Aorus Tachyon, which we’ll tackle on the next page.

MORE: Best Motherboards

MORE: How To Choose A Motherboard

MORE: All Motherboard Content

At $529, the Gigabyte Z590 Aorus Tachyon is the least expensive option in our Overclocking roundup, by a not-insignifcant $50. But don’t let the price tag fool you, this is a very capable and feature-rich overclocking motherboard. We’ll see if Gigabyte’s offering stands out among the others. But first, here is the full list of the board’s specs. 

Specifications - Gigabyte Z590 Aorus Tachyon 

Starting with the accessories, the Tachyon includes more than the OC Formula, but less than Asus’ Apex. The goal here is to have enough of the right accessories to get the system up and running without a trip to the store and Gigabyte accomplished that. It’s worth pointing out, though, that our board didn’t come with a driver disk or USB stick. Unless that was an error for the sample we have, you’d need to get your drivers from the Tachyon website and put them on a disk. Below is a complete list of all included accessories.

• RGB extension cable
• (8) SATA cables
• (2) thermistor cables
• Wi-Fi Antenna
• (3) M.2 screw packages
• Microphone
• IO plate
• Q-connector
• Aorus sticker sheet
• User Manual / Installation Guide

The Z590 Tachyon is an E-ATX size board that, outside of the parts on the board with overclocking buttons, you’d never know was designed to be an extreme overclocker. The 12-layer PCB sports a large heatsink to cool both VRM banks that reach out over the rear IO. There’s a dark reflective strip in the middle with the Tachyon name stenciled in grey. Around the reinforced PCIe slots are heatsinks that cover the M.2 sockets and Z590 chipset. The heatsinks have lines running through them to increase exposed surface area and create an aesthetically pleasing pattern. On the RGB front, there is a single strip on the chipset heatsink, above the Aorus branding. Overall, the Z590 Tachyon is a good-looking motherboard and matches well with any build theme. However, it does have a more purpose-built look to it, like the Apex.

Focusing our attention on the top half of the board, we get to see closer up what makes the board tick. Starting in the upper-left corner, along the top edge, are three (of eight) 4-pin fan headers. All fan/pump headers output up to 2A/24W, which should be plenty for most users. Below the rear IO bits are the two shielded 8-pin EPS connectors designed to feed the CPU power. Most boards have these connectors on the top edge, but these are in a unique location.

Moving our focus to the right to the socket area, the Tachyon uses 100% full tantalum polymer capacitors, which supposedly have a better transient response and less mechanical interference with extreme overclocking. In other words, it’s easier to fit a pot on top as well as insulating the area. Along with any performance improvements that may be derived from using these caps, this is a great value-add for the sub-ambient crowd, even if it’s only for simplifying insulation installation.

Continuing right, you run into the two reinforced DRAM slots that support up to 64GB of RAM with speeds listed to DDR4 5333 on the memory compatibility chart. As always, your mileage may vary in reaching these speeds, as it depends on the quality of the CPU’s integrated memory controller and the kit used.

Just above these slots are two (of four) RGB headers: one 3-pin for ARGB and a 4-pin for RGB. There’s a fourth fan header, SYS_FAN2, located here as well. 

To the right of that is the Overclocking section. The fundamental power/reset and CMOS reset buttons and others include two switches that adjust the CPU ratio, two more to access stored profiles, a voltage measurement point, and several other switches. Here there’s an OC Trigger switch (lowers CPU frequency to cool down the CPU), LN2 and Enhanced LN2 mode switches (to boot at ultra-low temperatures), reserved switch, dual BIOS enabled disable a simple BIOS switch. As far as these tools go, they’re just as comprehensive and useful as the other overclocking boards covered here. Below all that is a pair of SATA ports and a USB port conveniently located for the extreme overclocker. Last in this area is a front-panel USB 3.2 Gen 2 port and another system fan header.

The Tachyon’s VRM consists of 11 direct phases for Vcore. However, don’t let the lower count fool you. Power is sent from the EPS connectors to a Renesas ISL69269 (X+Y+Z=12) controller directly to 11 Vishay SiC840A 100A DrMOS power stages (others use the newer Smart Power Stages). While this is an incredibly robust power delivery system offering 1100A to the CPU, that is the lowest amount compared to other boards in this article. That said, the VRMs won’t hold your ambient or sub-ambient overclocking goals back.

Moving on to the bottom half of the board, the left side has the Realtek ALC1220-VB chip along with all the other audio bits on full display. There are six yellow audio capacitors along with four red WIMA audio caps. Although the audio section isn’t shielded at all (no faraday cage for the codec IC), it still has the line separating the audio from the rest of the board. While this board also sports a last-generation audio flagship chip, remember the goal is extreme overclocking. And most users would still find the solution here more than sufficient for a gaming or productivity machine.

The middle of the board houses a total of four full-length PCIe slots and three M.2 sockets. Starting with the PCIe configuration, The top and bottom slots connect to the chipset and run at x4 and x1 speeds, respectively. The primary video card slot is the second slot from the top and is the only x16 (electrically) slot on the board. The secondary graphics slot is the third from the top and runs at x8 speeds. Both of these slots are CPU connected and, when using a Rocket Lake-based CPU, uses the PCIe 4.0 protocol. This configuration supports AMD Quad-GPU CrossFire and 2-Way CrossFire. While there are enough lanes to run SLI, support isn’t mentioned. Worth noting is that you’ll more than likely cover the bottom slot leaving only the top available for additional peripherals if you plan to use two graphics cards.

For M.2 storage, the Tachyon sports three M.2 sockets, all of which have a heatsink over the top. The top socket, M2A_CPU, is connected via the CPU running up to PCIe 4.0 x4 (64 Gbps), with room for up to 110mm modules. Like most other Z590 boards, this socket only works with a Rocket Lake-based CPU and will otherwise be disabled using a 10th generation processor. The middle sockets, M2P_SB and M2M_SB, connect via the chipset and support up to 110mm SATA and PCIe 3.0 x4 drives. When M2P_SB is using a SATA-based drive, SATA port 1 is disabled. M2M_SB disabled SATA ports 4/5 regardless of the module installed. So your worst-case scenario with this board is three M.2 sockets in use along with five SATA ports available.

As we continue right, we pass over the chipset heatsink and over to the right edge. Here we spy a shielded 90-degree-oriented 24-pin ATX connector to feed the motherboard power. Below are the six chipset-connected SATA ports. The Tachyon supports RAID0, 1, 5, and 10. 

Wedged between the front panel header is a noise sensor header and two temperature sensor headers for use with the included thermistors. You can attach these to key points in your chassis to adjust fan headers and water pumps through these external sensors. This isn’t terribly helpful for the extreme overclocker, but is an excellent value-add for those running a system 24/7. The last jumper is to clear the CMOS.

Flipping the board around to the rear IO, one of the first things you’ll notice is the IO plate is NOT pre-installed. The board does come with one, but most using this for extreme overclocking won’t use it. From left to right is the CMOS reset and Q-Flash+ buttons, two legacy PS/2 ports for mouse and keyboard, two stacks of USB ports (4x USB 3.2 Gen1, 2x USB 3.2 Gen2), and the HDMI (v2.0) output. 

Next up are the Wi-Fi antenna connectors for the Wi-Fi 6E, the 2.5 Gbps Intel LAN port, a USB 3.2 Gen 2 port (Q-Flash port), along with a USB 3.2 Gen2x2 Type-C port. Finally, the audio stack is standard, sporting five analog plugs and the digital SPDIF output. Again, there’s nothing to complain about back here as there are plenty of USB ports for those who need them.

Firmware 

The Z590 Tachyon’s BIOS theme doesn’t look different from its other Z590 siblings. The Aorus board still uses the black and orange theme we’re familiar with. Like other board partners, Gigabyte includes an Easy Mode for high-level monitoring and adjustments, along with an Advanced section. The BIOS is well organized, with many of the more commonly used functions easily accessible without drilling down multiple levels to find them. In the end, the BIOS works well and is easy to navigate and read (although I still wish Page Up/Down worked). We’ve captured a majority of the BIOS screens to share with you in the slideshow below.

ALBM biosimagegiga 1-15

Software

Gigabyte includes a few applications designed for various functions, including RGB lighting control, audio, system monitoring, and overclocking. Below, we’ve captured several screenshots of the App Center, @BIOS, SIV, RGB Fusion and Easy Tune.

With the features of our three Overclocking boards out of the way, let’s look at performance on the next page. 

MORE: Best Motherboards

MORE: How To Choose A Motherboard

MORE: All Motherboard Content

Test System 

As of March 2021, we’ve updated our test system to Windows 10 64-bit OS (20H2) with all threat mitigations applied. On the hardware front, we’ve switched to all PCIe 4.0 components. We upgraded our video card to an Asus RTX 3070 TUF Gaming and the storage device to a 2TB Phison PS5-18-E18 M.2. Along with the hardware changes, we’ve also updated the games to F1 2020 and Far Cry: New Dawn. We use the latest non-beta motherboard BIOS available to the public unless otherwise noted (typically during new platform launches). The hardware used is as follows: 

Benchmark Settings 

Benchmark Results 

Our standard benchmarks and power tests are performed using the CPU’s stock frequencies (including stock Thermal Velocity Boost), with all power-saving features enabled. Optimized defaults are set in the BIOS and the memory set using the XMP profiles. For this baseline testing, Windows is set to the balanced power plan. 

Synthetic Benchmarks 

Synthetics are a great tool to figure out if a board runs out of spec, as identical settings should produce similar performance results. Advanced memory timings are the one place where motherboard makers can still optimize for either stability or performance, though, and those settings can impact some testing.

As we look at the synthetic benchmarks, our three boards performed well compared to the other tested boards. It’s tough to see any one of the three boards we’re looking at here as a clear performer over the others, as they trade punches in each of our tests. PCMark 10/Office results were overall above average. In many cases, particularly in the MS Office tests, the three boards scored on the higher end or the highest compared to other Z590 boards we’ve tested.

Timed Applications

In our timed applications, LAME testing put the Tachyon ahead of the Apex and OC Formula, with one of the fastest results at 11.36 seconds. The Apex and Formula came in just slightly slower at 11.50 and 11.51, respectively, just above the average. Corona test results were also the fastest we’ve seen. The OC Formula completed this benchmark the fastest of all tested boards so far (81 seconds), while the Tachyon and Apex were a mere second behind and still ahead of the curve at 82 seconds.

3D Games and 3DMark

We’ve recently updated our game tests to F1 2020 and Far Cry: New Dawn. The games are run at 1920x1080 resolution using the Ultra preset. As the resolution goes up, the CPU tends to have less of an impact. The goal with these settings is to determine if there are differences in performance at the most commonly used resolution with settings most people use or at least strive for.

Our three overclocking boards performed right around the averages in both the synthetic and gaming tests. The OC Formula threw out a slightly slower average in Far Cry: New Dawn. But otherwise, you won’t be able to tell the difference between boards here, outside of the benchmarks.

Power Consumption / VRM Temperatures

We used AIDA64’s System Stability Test with Stress CPU, FPU and Cache enabled for power testing, using the peak power consumption value. With Rocket Lake, we need to disable AVX-512 instruction sets; otherwise, temperatures are out of control at stock speeds on some boards. The wattage reading is from the wall via a Kill-A-Watt meter to capture the entire ecosystem. The only variable that changes is the motherboard; all other parts are the same.

Power consumption on these overclocking boards was all over the map. The Tachyon used the least amount of power, averaging 160W (idle 63W, load 257W), while the Apex averaged 178W (idle 79W, load 277W). At the top of the charts using the most power is the OC Formula. In stock form, it averaged 196W, which is the highest tested so far by over 10W. The load was also the highest at 296W, and idle power use was quite high at 95W. Power use/efficiency isn’t generally a concern for overclocking-type motherboards, so we won’t hold anything against those using more or less. You don’t get upset that your super-car has poor gas mileage, you know? Similar situation here.

During stock operation, all three boards handled the flagship Intel Core i9-11900K without issue. The actively cooled OC Formula ran the coolest, peaking just over 42 degrees Celsius. The passively cooled Apex and Tachyon were close, both peaking just over 44 degrees Celsius. Compared to most other boards, these ran cooler than the average overall. During overclocked operations, All three boards were within 1.5 degrees Celsius of each other. Again, these temperatures are some of the coolest running out of all the boards we’ve tested. These made-for-overclocking motherboards and their VRMs and heatsinks do a great job at managing things under ambient cooling.

Overclocking

When overclocking the i9-11900K, we decided to do so by testing without AVX-512 instructions. When you unlock all of the power limits, you’re thermally limited, with the CPU reaching 100 degrees Celsius in several seconds. Since every board is different in following (or not following) Intel specifications, we’ve removed that variable. The bottom line is that if you need to use these instructions, you’ll need to set a significant offset compared to non-AVX-512 loads. To that end, we settled on a 5.1 GHz overclock at around 1.3V. Our CPU uses around 225W in this configuration, which lands between stock power use (195W) and AVX-512 power use (~265W) and is the end of the line due to thermal constraints.

Overclocking on these boards through typical means (the BIOS) went as easily as any other capable motherboard we’ve tested. A few tweaks here and there, and we were on our way. Often when pushing the limits, the process of booting the machine to windows can prove fatal, even though you’re ‘stable enough’ to run a few benchmarks. To that end, All boards reached 5.2 GHz while running our memory at DDR4 4000 and ran Cinebench and POVRay, but temperatures were too high to sustain a lengthy stress test. The overclocking tools, specifically the hardware buttons to adjust the overclock in Windows, proved particularly useful to get that last 100 Mhz or so. 

Sadly, I didn’t have the chance to use sub-ambient cooling as procuring LN2 around me (yes, I turned my nose up at Dry Ice) proved more complicated than it was in the past. That said, all it takes is a quick stroll over to Hwbot.org to check each board’s records (Links for each: OC Formula, Apex, Tachyon). There you’ll find CPU frequencies above 7 GHz and memory frequencies well over DDR4 6000 (3000 Mhz-plus). With the proper cooling and in the right hands, these boards are the tools used to break records.

Final Thoughts

After taking the overclocking boards for a spin, I have to admit I miss the rush of using sub-ambient cooling methods and the competition scene. And although these motherboards are strongly focused on extreme overclocking abilities, they’re still competent, albeit pricey, daily drivers as well. 

All of these overclocking boards sport incredibly robust power delivery, two DRAM slots, as well as onboard buttons to get every last available Hz out of the system. On top of that, they offer plenty of storage options, premium audio solutions, fast integrated LAN and Wi-Fi 6E, along with a slew of fast USB ports. And last but not least, most have a premium appearance that fits in with your build theme. 

All three of these overclocking-centric motherboards performed well overall, just above average across most tests. Some results showed some slightly slower numbers, while most ran faster than the average—some the fastest we’ve recorded. I wouldn’t hold too much value in the stock results though, as a system running one of these boards is more than likely going to be heavily overclocked. For the most part, you’d be hard-pressed to find a significant difference using ambient cooling.

As far as complaints go, there isn’t much here. Every consumer wants a lower price, but for those who use these features, they’re considered priceless when in pursuit of Hwbot records or even overclocking for the fun of it. Outside of that, I wish the OC Formula didn’t have the ‘zombie’ green accent color and gave off a more premium vibe, but we should keep things in perspective. Aesthetics isn’t the primary concern with these boards and it too would do well in most build themes, especially considering the large number of RGB lighting elements.

Outside of that, the difference between these boards comes down to price, (the Aorus Tachyon is the least expensive at $529.99) overclocking tools (this is toss-up), storage (all are good for 3x M.2 and at least 3x SATA concurrently), audio (only the Apex has the latest and greatest), and power delivery (all will take your chip to the limits of whatever cooling method used). If you’re looking to display the board and not just beat on it with LN2, the Apex has a more premium appearance, followed by the Tachyon. Given everything the above, plus the Gigabyte board's significantly lower price, we give the nod to the Aorus Tachyon. It delivers good looks, the important overclocking features of its competitors, and will save you some money you can spend elsewhere.

That said, any of these boards are well equipped with the tools of the overclocking trade trade. I’d take any on an an extreme overclocking adventure and get great results, but prefer to work in the Asus BIOS versus the others. But for a daily driver with an eye on some kind of sub-ambient overclocking, the Tachyon would be my weapon of choice. It’s the least-expensive of the bunch, but still offers plenty of features for the average user and the extreme overclocker.

MORE: Best Motherboards

MORE: How To Choose A Motherboard

MORE: All Motherboard Content

Samsung spent tens of billions of dollars annually on its various chip businesses for several years now. On Thursday the company announced that its renewed semiconductor investment plan includes spending $150 billion on R&D and CapEx for its LSI and foundry businesses by 2030.  

$15 Billion per Year 

About two years ago Samsung announced plans to spend $115 billion through 2030 (i.e., $9.51 billion a year) on its LSI and foundry businesses, which includes R&D of new process technologies and materials, chip design, and actual fabs. $9.51 billion a year is a lot of money, but from now on the company is committed to spending $15 billion a year on its LSI and semiconductor production operations. The increase is motivated by the growing demand for chips and semiconductor fabrication services driven by megatrends like 5G, AI, and HPC.

Samsung Electronics has been outspending all other semiconductor producers in the world in recent years. Just last year it invested around $28.1 billion in production facilities that are used by its Samsung Foundry and memory (3D NAND, DRAM) businesses. With an increased commitment to foundry business, Samsung will almost certainly increase the general CapEx of its semiconductor division to well beyond $30 billion a year.

“The entire semiconductor industry is facing a watershed moment and now is the time to chart out a plan for long-term strategy and investment,” said Dr. Kinam Kim, Vice Chairman and Head of Device Solutions Division at Samsung Electronics. 

Samsung P3: A Fab for DRAM and Logic 

In an interesting turn of events, Samsung also announced that it had begun construction of its new P3 fab near Pyeongtaek, South Korea, that will be used to produce both DRAM and logic chips using process technologies that rely on extreme ultraviolet (EUV) lithography. Normally, Samsung uses different fabs to make DRAM and logic chips, but it looks like with next-generation EUV tools it makes sense to use one fab for both types of products.

Samsung's P3 is expected to be completed in the second half of 2022. Initially, it will produce chips using Samsung's 14 nm technology for DRAMs as well as 5 nm node for SoCs. Looks like leading-edge fabrication technologies, including 4nm and 3nm, will continue to be initially used at Samsung Foundry's V1 fab near Hwaseong, South Korea.

“For the memory business, where Samsung has maintained its undisputed leadership position, the company will continue to make preemptive investments to lead the industry,” said Dr. Kinam Kim. 

TSMC this week published its financial results for the first quarter and shared some of its expectations regarding the ongoing chip shortage. The company believes that chip shortages will continue well into 2022 and will only ease in 2023. Meanwhile, TSMC will provide some additional allocation for automakers shortly to boost supply to the clients that are important for the economies of counties like the U.S. and Germany. Also, TSMC intends to increase its capital expenditures beyond the planned $28 billion in 2020.

Supply Crisis to Ease in 2023

TSMC operated at an "over 100% utilization" during the quarter, though it could not fulfill all of its customers' orders. When asked about expectations when the shortages were going to end, the management responded that it expected the deficit to persist throughout 2021 and into 2022. It is impossible for TSMC to quickly respond to growing demand as it has to buy land, prepare the sites, construct new fabs, procure manufacturing equipment, install these tools, and ramp up the facilities. All of these steps take a long time, and TSMC expects its supply abilities to improve tangibly only in 2023. 

"We see the demand continue to be high," C. C. Wei, CEO of TSMC, told analysts and investors during the company's earning's call. "We have acquired land and equipment, and started construction of new facilities. We are hiring thousands of employees and expanding our capacity at multiple sites. […] In 2023, I hope we can offer more capacity to support our customers. At that time, we'll start to see the supply chain tightness release a little bit." 

To support the expansion, TSMC increased its CapEx budget from the previously planned $25 – $28 billion to around $30 billion in 2020. 

Revenues Hit $12.92 Billion, Demand for HPC Up

TSMC reported revenue of NT$362.41 billion ($12.92 billion), up 25.4% year-over-year, and net income of NT$139.69 billion ($4.932 billion) for the first quarter that ended on March 31, 2021.  

TSMC's N7 lineup of manufacturing technologies accounted for 35% of the company's earnings, whereas the N5 node and N16 family of processes accounted for 14% each of TSMC's revenue. N28 technologies continued to be fairly popular, so the node was responsible for 11% of TSMC's revenue. 

It's noteworthy that N5 dropped from 20% of TSMC's consolidated revenue in Q4 2020 to 14% in Q1 2021, whereas N7 increased from 29% to 35%. This happened because Apple lowered orders for its 5 nm chips for smartphones and tablets, whereas companies from PC and server spaces (which TSMC collectively calls HPC) increased their orders for APUs, CPUs, GPUs, network processors, SoCs, and other components. 

Multiple Megatrends Ahead, Automakers Get a Preference

TSMC has multiple growth opportunities ahead of it. Obviously, demand for PCs and other client devices is rising, so traditional customers like AMD, Broadcom, Nvidia, and Qualcomm are increasing their orders to the foundry. Also, there are impending 5G, AI, HPC, IoT, and edge computing megatrends that will further spur demand for semiconductors as all of these applications require brand-new chips produced using fairly advanced process technologies.  

Cars are among the applications set to get smarter, as they now use rather sophisticated chips and are set to use even more complex SoCs as they become autonomous. Automotive applications accounted for about 4% of TSMC's revenue in Q1 2021, perhaps not a big deal for the company. But the absence of chips made by TSMC essentially stops car manufacturing in Europe and the U.S., which hits not only automakers but also hundreds of their subcontractors and economies in general. Given the importance of automakers, TSMC vowed to increase its allocation for automotive chips next quarter.

Taiwan Semiconductor Manufacturing Co. via Yahoo Finance / Bloomberg today said it would invest $100 billion over the next three years in new manufacturing capacities as well as research and development (R&D). The investments in advanced fabs will increase TSMC's lead over rivals and will ensure that the company introduces new fabrication technologies in a predictable cadence. 

"TSMC is entering a period of higher growth as the multiyear megatrends of 5G and HPC are expected to fuel strong demand for our semiconductor technologies in the next several years," a statement by TSMC reads. "In addition, the COVID-19 pandemic also accelerates digitalization in every aspect. In order to keep up with demand, TSMC expects to invest $100 billion over the next three years to increase capacity to support the manufacturing and R&D of advanced semiconductor technologies. TSMC is working closely with our customers to address their needs in a sustainable manner."

Earlier this year TSMC announced plans to radically increase its capital expenditure (CapEx) budget to $25 ~ $28 billion, an increase of 45% ~ 62% year-over-year from $17.2 billion in 2020. Last year the company spent about $3.72 billion on R&D, or about 8.2% of its revenue. 

Gradual increases of R&D and CapEx budgets at leading makers of semiconductors are generally in line with expectations. New process technologies are getting more expensive to develop, whereas new fabs are getting larger and more expensive to build. For example, a 3nm/5nm-capable fab costs around $20 billion. Therefore, announcing a three-year $100 billion investment plan is not particularly surprising for TSMC as in the coming years it will have to build a 2nm-capable fab (or two). 

TSMC says that demand for chips is so high that it has been running its fabs at full load for about a year and supply is still falling short of the increased demand, which is why it needs additional fabs and new employees. To support its capacity expansion, the company intends to suspend wafer price cuts for a year from the start of 2022, reports Bloomberg citing a letter from TSMC's CEO to the company’s clients.  

Meanwhile TSMC is announcing its intention to pour in $100 billion in chip development and manufacturing just a week after Intel outlined its Integrated Device Manufacturing 2.0 plan under which the largest CPU maker in the world plans to become a competitor for TSMC. So far, Intel has announced plans to spend $20 billion on two production facilities in Arizona and said it would invest in expansion of chip manufacturing in other parts of the world.

It is well known that Taiwan Semiconductor Manufacturing Co. and Samsung Foundry dominate the market of contract chip production. They are the only companies to offer leading-edge process technologies and have the largest capacities. Meanwhile, TSMC and Samsung Foundry are on track to become the dominant manufacturers of advanced chips as nobody, including Intel, can match their capital expenditures.

TSMC: Big Can Only Get Bigger

Founded in 1987, TSMC was the world's first pure play foundry that manufactured chips for others. In 34 years of its history, the company has grown from a small entity with one fab to a multi-billion corporation with five 300mm fabs, seven 20mm fabs, and one 150mm production facility. Having developed dozens of process technologies throughout its history and having installed vast production capacities, TSMC can offer services to almost any fabless chip designer with almost any requirements. At present, TSMC serves over 460 customers.

As the demand for leading-edge fabrication processes and volumes from its large customers (such as Apple, HiSilicon, Qualcomm, Nvidia, and AMD) have grown in recent years, TSMC intensified building of new GigaFabs — production facilities with a capacity of more than 100,000 300-mm wafer starts per month (WSPM). Each costs around $20 billion, and TSMC also increased its research and development (R&D) budgets. The strategy has paid off and today TSMC has not only left Intel and Samsung Foundry behind with its manufacturing technologies, but it also has more leading-edge capacity than other makers of semiconductors. This is largely because because it serves virtually all fabless designers that require advanced technologies. 

This year the company decided to radically increase its capital expenditure (CapEx) spending to $25 billion ~ $28 billion, an increase of 45% ~ 62% year-over-year from $17.2 billion in 2020. IC Insights believes that TSMC will "begin what is likely to be a huge multi-year ramp of spending," and expects the company to boost its CapEx budget in 2022 and 2023 once again. 

Being the leading maker of semiconductors both in terms of volumes and in terms of technology leadership has its advantages. First, it's easier to get the fab tools when you buy them in high volumes. Second, it's easier to set up your own production and supply chain standards, something that is tremendously important in an industry that is all about standardization. 

Samsung Foundry: Closing the Gap with TSMC, Widening the Gap with Intel?

Samsung Electronics has been the world's largest maker of dynamic random access memory (DRAM) and NAND flash for quite a while and has been in the semiconductor business for decades. Furthermore, it has produced various chips for its own needs. The company started to offer foundry services in mid-2000s, as it realized that only the largest chipmakers will survive in the long term. Samsung Foundry has been trying to catch up with TSMC for years, and while the gap is closing, it is still not quite there yet.

Samsung Foundry's largest customer is still its parent company Samsung, which strives to make the world's best smartphones, televisions, PCs, displays and other electronics. To that end, SF's design decisions at times resemble those of an integrated device manufacturer (IDM) that makes money on actual products rather than on manufacturing services.

Samsung realized early enough that demand for chips (all chips, including DRAM, 3D NAND, SoCs, etc.) will only grow, so its corporate semiconductor CapEx spending exceeded $10 billion for the first time in 2010. Having spent $93.2 billion on expanding production capacities over the 2017–2020 period, the company significantly closed the gap with TSMC from a capacity point of view.

Samsung Foundry is still about three times smaller than TSMC in terms of wafer starts per month (and also in terms of the number of nodes it offers), but the gap between the two has been closing. So far, Samsung has not unveiled its 2021 semiconductor CapEx budget, but analysts believe that it could spend at least as much as it spent last year — around $28.1 billion.

Cumulative CapEx of Samsung and TSMC will total approximately $55.5 billion this year, according to IC Insights. A significant part Samsung's funds will of course be used to buy equipment for Samsung's memory businesses, but these two companies will be able to influence development of fab production tools and supply chains.

Should Intel Worry?

Intel traditionally spends tens of billions of dollars on CapEx (it spent about $14.3 billion last year), so it will remain a leading maker of processors. Yet, its spending on fabs will be about half that of Samsung and TSMC this year. Furthermore, since Intel will not start production of chips using a node that relies on EUV, it will not have an immediate significant influence on development of the industry and supply chains.

Historically, Intel had several competitive advantages that set it apart from all of its direct and indirect rivals:

1. Intel's CPUs were the fastest in the industry. 
2. Intel's microarchitectures and CPU designs were scalable for all market segments.
3. Intel had enough power to ensure that its architectural innovations were supported by software makers.
4. Intel had the best process technologies, which could offset certain imperfections of its microarchitectures or design.
5. Intel could produce CPUs in volumes unachievable by any of its competitors.
6. Since Intel was the de facto leader of the semiconductor market both financially and technologically, it set standards for the rest of the industry, which further ensured its leadership position.
7. While Intel competed against most companies in the semiconductor industry, it could build alliances or partnerships that strengthened it (e.g., with Microsoft, Dell, HP, Apple, and ATI Technologies) and helped it to better compete.
8. Intel spent hundreds of millions of dollars on marketing and advertising, usually more than all of its rivals combined. 

So far, Intel has lost at least three out of eight advantages. These days Intel's CPUs are not the undisputed leaders, and in many cases competing products from AMD are unchallenged. While Intel's 2nd generation and 3rd generation 10nm fabrication technologies are competitive against TSMC's N7, the company's nodes cannot offer the same transistor density as TSMC's N5. Finally, Intel no longer spends as much as its rivals on fabs and no longer has technological leadership.

If/when AMD becomes TSMC's second largest customer, it could ask its production partner to customize the nodes it uses in a bid to gain performance and/or lower power consumption. Meanwhile, we still know nothing about Intel's outsourcing plans other than the fact that some of its products will be made at TSMC in 2022.

Intel remains a driving force behind many industry initiatives, and no technology can get widespread in the PC world without Intel's support. Yet, there are no more Wintel-like initiatives and Intel is no longer an exclusive CPU supplier for companies like Apple.

Meanwhile, Intel has extremely capable x86 CPU architectures that offer higher single-thread performance when compared to those from AMD. Intel also produces more processors than any other maker, and it can supply its partners with volumes of chips not available from anyone else. Given Intel's market share and volume leadership, virtually all of its initiatives are supported by the software industry. Furthermore, the company knows how to advertise its products and promote its brand.

In general, Intel has many things to worry about, as it no longer can compete against all of its rivals on all fronts successfully. Hopefully, the company's new CEO will shed some light on the chip giant's future plans next week in a live chat.

Could Countries Compete Against Dominant Makers of Semiconductors?

Now that TSMC and Samsung spend around $28 billion each on manufacturing facilities and billions on R&D, it is extremely hard for a commercial company to catch up with these chipmakers. Even Apple, with its massive earnings and cash reserves, is hardly willing to invest tens of billions on chip manufacturing. In the recent years, the governments of the EU, US, and China started to talk about local semiconductor production industries and expressed willingness to assist chipmakers. 

IC Insights deems that it is close to impossible to catch up with TSMC and Samsung. Keeping in mind the two leading makers of semiconductors are way ahead of the industry both in terms of R&D and CapEx, analysts believe that "governments would need to spend at least $30 billion per year for a minimum of five years to have any reasonable chance of success." The Chinese corporation SMIC has received a lot of help both from local authorities and Chinese government over the years, but the company is still about five years behind GlobalFoundries, Samsung Foundry, and TSMC. 

Summary

Both TSMC and Samsung Foundry started to use EUV tools to produce chips using their leading-edge process technologies several years before Intel, so they have been gaining experience with new tools and supply chains for quite a while now. 

Both TSMC and Samsung will invest two times more in their production facilities than Intel will in 2021. Arguably, Intel does not need to spend as much as TSMC and Samsung on CapEx since it only produces chips for itself, whereas its peers offer foundry services. Yet, previously Intel's technological leadership was enabled by massive spending on fabs and R&D.

In theory, governments could stimulate development of the local semiconductor industry using direct help, tax breaks, and incentives. However, their total spending over the next five years would need to exceed $150 billion, and chances of success are not high.

Asus has listed the ROG Maximus XIII Apex on its website, implying that the successor to the ROG Maximus XII Apex may be closer than we think. The new iteration to the Apex series has been engineered to tame Intel's 11th Generation Rocket Lake processors.

Built around the new Z590 chipset and existing LGA1200 socket, the ROG Maximus XIII Apex comes equipped with an 18-phase power delivery subsystem. Each power stage, which can manage up to 90 amps, is accompanied by a MicroFine Alloy choke that can do 45 amps. Asus revamped the power design on the ROG Maximus XIII Apex completely by getting rid of the phase doublers. The motherboard also employs 10K Japanese black metallic capacitors that can take a beating. The VRM area is properly cooled with thick, aluminum passive heatsinks. The ROG Maximus XIII Apex feeds Rocket Lake chips through a pair of 8-pin EPS power connectors.

The overclocking toolkit on the ROG Maximus XIII Apex includes a double-digit debug LED, voltage read points, and a plethora of buttons and switches to aid in overclocking. There are also three condensation sensors that are placed strategically across the motherboard to notify you when condensation occurs around the processor, memory or PCIe slot. In total, the ROG Maximus XIII Apex has five temperature sensors, five 4-pin fan headers, two full-speed fan headers, and an assortment of headers for watercooling setups.

Like previous Apex motherboards, the ROG Maximus XIII Apex only provides two DDR4 memory slots. While memory capacity is limited to 64GB, the motherboard supports memory frequencies above DDR4-5000 with ease. The ROG Maximus XIII Apex sports Asus' OptiMem III technology, featuring an optimized memory tracing layout to improve memory overclocking.

The ROG Maximus XIII Apex offers numerous options for storage. It provides eight SATA III ports and up to four M.2 slots. The M.2 slots on the motherboard are PCIe 4.0 ready and come armed with an aluminium heatsink and embedded backplates to provide passive cooling. The other two M.2 slots reside on Asus' ROG DIMM.2 module that connects to the motherboard through a DDR4-type interface beside the memory slots. The DIMM.2 module accommodates M.2 drives with lengths up to 110mm.

The expansion slots on the ROG Maximus XIII Apex consist of two PCIe x16 slots and one PCIe x8 slot. Wired and wireless networking come in the shape of a 2.5 Gigabit Ethernet port and Wi-Fi 6E connectivity with support for up to 6GHz bands. The audio system on the ROG Maximus XIII Apex uses Realtek's ALC4080 audio codec complemented with a Savitech SV3H712 amplifier and high-end Nichicon audio capacitors.

In regards to USB ports, the ROG Maximus XIII Apex has four USB 3.2 Gen 1 ports, five USB 3.2 Gen 2 ports and one USB 3.2 Gen 2x2 Type-C port at the rear panel. There's an additional USB 3.2 Gen 2x2 header on the motherboard. The ROG Maximus XIII Apex doesn't supply any display outputs so it's mandatory to pair it with a discrete graphics card.

ROG motherboards have a very rich software suite. On this iteration, Asus has directly implemented MemTest86 into the ROG Maximus XIII Apex's firmware so overclockers can test memory stability without any hassles. Additionally, a one-year AIDA64 Extreme subscription is also included.

The pricing for the ROG Maximus XIII Apex is currently unknown. The previous Z490 version retailed for $356.99, so we can expect the Z590 followup to be price around that range if not a little bit higher.

If you're a fan (*cough*) of cool hardware, the Apexgaming AC-120Ti should be right up your alley. The AC-120Ti isn't your ordinary case fan as it actually features what the brand calls a "nano photocatalyst air purification" system that purifies the air that goes into your computer case.

Apexgaming's description of the AC-120Ti is a bit overwhelming for those of us that didn't pay more attention in our chemistry classes. The AC120Ti comes with a front filter, which basically acts as a dust filter, to help block large dust and other external materials from entering the fan. According to Apexgaming, the filter mesh comes with a coat of titanium dioxide (TiO2), whose purpose is to serve as a photo-oxidation catalyst. The filter can be easily cleaned by just washing it with water.

But the purification doesn't stop there. The fan includes an ultraviolet lamp that's the key component in the entire purification process. It essentially shoots ultraviolet rays at the catalyst to form hydroxyl radicals that absorb stuff like bacteria, odorous cells, and pollutants to decompose them into carbon dioxide and water molecules. The final product, at least according to the company, is air that's free of unpleasant smells, formaldehyde, germs, viruses, and allergens.

Apexgaming has all the certifications that prove that its UV LED is safe for human eyes and skin. The company is also touting an impressive baterial elimination rate up to 95.56%. For optimal operation, Apexgaming recommends installing the AC-120Ti in the front of your case as an intake fan.

The fan checks in with dimensions of 120x120x25mm and draws power from a standard 4-pin connector. Its sleeve bearing can spin at up to 1,000 RPM.  The fan boasts an airflow and air pressure up to 34 CFM and 0.9 mmH2O, respectively. The maximum noise level is around 16 dBA. Apexgaming rates the AC-120Ti with a lifespan of 40,000 hours at 25 degress Celsius.

Apexgaming sells the AC-120Ti for $29.99 via the company's online store. We doubt the AC120Ti is going to help you get more FPS out of your games, but if you want to turn your PC into an air purifier, the AC120Ti is the fan for the job. At the very least, it might keep your PC from catching a virus.

But please: No one get any ideas about building an SFF with one of these fans and strapping it to your face. If you want to mix your love of gaming with the ability to feel a bit safer around others, maybe hold out to see if Razer ever brings a version of Project Hazel mask to market. 

Taiwan Semiconductor Manufacturing Co. said this week that it would significantly increase its capital expenditures this year as it needs to build and equip new fabs to ramp up its existing and next-generation leading-edge process technologies. Demand for advanced fabrication nodes is on the rise. In Q4 2020, revenue for N5 (5 nm) and N7 (7 nm) technologies accounted for nearly half of TSMC's revenue, so it makes a great sense for the manufacturer to continue investing in leading-edge capacities. Furthermore, as Apple and Intel plan to increase orders to TSMC, the foundry will also need more capacity.

TSMC to Spend Up to $28 Billion on Production Facilities in 2021

TSMC's management plans to spend from $25 billion to $28 billion on capital expenditures in 2021, the world's largest contract maker of semiconductor announced this week. 

If TSMC's board approves all the investments proposed by the management, then TSMC's CapEx will rise by 45% ~ 62% year-over-year from $17.2 billion in 2020. The lion's share of TSMC's capital budget will be spent on expanding production capacities for the company's N5 and N7 nodes as well as equipping a leading-edge fab for its N3 (3 nm) node that will be used starting from the second half of 2022. 

"Out of the $25 to $28 billion CapEx for 2021, about 80% of the capital budget will be allocated for advanced process technologies, including 3 nm, 5 nm and 7 nm," said Wendell Huang, chief financial officer at TSMC, during this week's earnings call with analysts and investors. "About 10% will be spent for advanced packaging and mask making and about 10% will be spent for speciality technologies." 

"The 2021 CapEx is largely for N5 expansion (at least 2x its current 55-60k WPM capacity), and N3 build is limited to pilot line/initial production (+10-15k WPM, though we expect N3’s capacity to be similar to N5’s upon full ramp-up)," Szeho Ng, an analyst with China Renaissance Securities, wrote in a note to clients on Friday.

 Production of semiconductors is an extremely capital-intensive business and as process technologies get more advanced and manufacturing tools get more expensive, chipmakers have to increase their CapEx budgets to build fabs that cost as much as $20 billion or even more. Because of increasingly high R&D and fab costs, numerous contract makers of semiconductors had to pull out from the leading-edge and focus on specialized technologies in the recent years.

N5 & N7 Accounted for 49% of TSMC's Revenue in Q4 2020

But with a number of rivals out of the game, TSMC now has more clients than ever. And considering the fact that there are a number of ongoing megatrends — such as 5G, AI, high-performance computing (HPC), and edge computing — that require advanced chips produced using the most sophisticated technologies possible, demand for TSMC's higher-end production services is on the rise today and will continue growing in the coming years.   

In the fourth quarter, sales of wafers processed using N5 accounted for 20% of TSMC's total wafer revenue, whereas sales of N7 and 12 nm/16 nm accounted for 29% and 13% respectively. In general, leading-edge N5 and N7 nodes commanded 49% of TSMC's wafer revenue, whereas advanced nodes (N5, N7, N12/N16) accounted for 62% of the company's total wafer revenue. 

TSMC's N5 process technology in Q4 2020 was primarily used by Apple, which uses it for its A14 system-on-chips for smartphones and tablets as well as M1 SoCs for its latest PCs. The company's N7 is used more widely by companies like AMD, Apple, Nvidia, Qualcomm, and Xilinx, whereas N12/N16 is used by dozens of companies for all kinds of applications.  

"Our fourth quarter business was supported by strong demand for our industry-leading 5 nm technology, driven by 5G smartphone launches and HPC-related applications," said Mr. Huang. "Moving into first quarter 2021, we expect our business to be supported by HPC-related demand, recovery in the automotive segment, and a milder smartphone seasonality than in recent years." 

TSMC's N5 is a major improvement over N7 node that enables a 15% higher performance (at the same power and complexity) or a 30% lower power consumption (at the same frequency and complexity), and an up to 1.8X higher transistor density (not for all structures). In addition, N5 uses extreme ultraviolet (EUV) lithography on more than 10 layers to reduce mask count to 81 as well as reduce or even avoid usage of multipatterning, which optimizes yields and costs. Considering all the advantages that N5 has over N7, the majority of TSMC's clients that use N7 now will migrate to N5 over time. 

Later this year TSMC will offer its customers a performance-enhanced version of N5 that is called N5P. The latter is projected to enable a 5% performance gain and/or a 10% power reduction over N5 while staying generally compatible. Yet, N5P will be used by select chip designers only as next year TSMC will offer N4 that will provide further PPA (power, performance, area) enhancements while being compatible with N5 on design rules, IPs, and SPICE levels. TSMC's fabs that currently process wafers using its N5 technology will also be able to use N5P as well as N4, so any investments going there will be used for years.

N3 Could Be More Successful Than N5 and N7

TSMC's next major step after N5 will be N3 (3 nm). This technology promises an up to 15% performance gain (at the same power and transistor count), or an up to 30% power reduction (at the same speed and transistor count), along with an up to 20% higher SRAM density as well as an up to 70% higher logic density. This technology is still in development and will be used starting from the second half of 2022.

TSMC says that it sees more interest in N3 from potential clients in the HPC (TSMC considers 'HPC' all CPUs, GPUs, FPGAs, non-mobile SoCs, special-purpose accelerators, etc.) and smartphone spaces than it did with N5 and N7 when they were at a similar stage of development. 

"We are seeing a much higher level of customer engagement for both HPC and smartphone application at N3 as compared with N5 and N7 at a similar stage," said C. C. Wei, Chief Executive Officer of TSMC. "Risk production is scheduled in 2021, and volume production is targeted in the second half of 2022." 

TSMC does not disclose any actual numbers that would quantify interest for N3 from its clients, but since many new applications are going to emerge in the coming years, it is not surprising that more players with more designs will be opting for a leading-edge process technology in 2022 ~ 2025 than they did several years ago. Furthermore, since N3 continues to rely on FinFET transistor structures with predictable performance and costs, the fabrication process will be a very long node as its successors will move to GAAFETs (gate-all-around field effect transistors), which is going to have a significant impact on design costs. For example, TSMC's 28 nm node that relies on planar transistors accounted for 11% of its wafer revenue in Q4 2020, nine years after the fabrication process was introduced.  

"We note the bold expansion reflects N3-5 as long- lasting big nodes with expected robust future demand beyond the initial wave from key clients," Szeho Ng, an analyst with China Renaissance Securities, wrote in a note to clients on Friday. 

While eventually dozens of companies will use TSMC's N3 and N5, there are two customers that are going to have a significant impact on TSMC's CapEx budgets in the coming years: Apple and Intel.

Apple: More Chips Needed

Traditionally, the bulk of Apple's orders to TSMC have been its SoCs for smartphones and tablets. Many of its suppliers produce their chips at TSMC as well, but those semiconductors are not made under Apple's orders and are not made exclusively for the company. As Apple transits its PCs to its own SoCs, starts to use its own modems inside its smartphones and tablets, and increases sales of its accessories that rely on its own system-in-packages (SiPs), it will significantly increase its orders to TSMC.

Apple is the world's No. 4 PC supplier that shipped 23.102 million computers last year and commanded about 7.6% of the PC market. Sales of PCs in 2020 totalled 302.6 million units, according to IDC. Once the technology giant completes transition of Macs from Intel CPUs to its Apple Silicon SoCs, it will need at least 23.102 million fairly big chips (the M1 is at 120 sq. mm, which is 26 sq. mm smaller compared to Intel's Tiger Lake-U) every year assuming that the market stays flat and Apple stops gaining market share (something that it has been doing for years). 

Apple was also the third largest supplier of smartphones in 2019 with 191 million units shipped, based on IDC's data. When the company ceases to use third-party 5G modems (which happen to be made using a leading-edge process technology) and transits to its own, it will increase its orders to TSMC by another ~200 million of chips every year.  

Apple's transitions to its PC SoCs and modems will be completed within the next couple of years. While it is hard to quantify Apple's orders to TSMC in terms of wafers today, it is safe to say that the addition of ~25 million PC SoCs and ~200 million modems to around ~200 million smartphone SoCs that TSMC produces today will significantly increase the number of wafers that the foundry processes for Apple.

Intel: A Wildcard

Intel already outsources manufacturing of multiple products to TSMC and has confirmed officially and unofficially that it would produce Atom and Xeon SoCs as well as GPUs at TSMC. What nobody knows at this point is whether Intel intends to make client and server CPUs at TSMC and if it has such plans, what percentage of its CPUs will be made at TSMC. 

Since Intel's outsourcing plans are unclear and perhaps are even undecided, it is not  good business to make too many assumptions at the moment. Yet, there are a couple of things to keep in mind. Intel sells hundreds of millions of processors and chipsets for client PCs every year and even if the company decides to outsource 10% ~ 15% of its client CPU/chipset output to TSMC, it will still be tens of millions of units added to Intel's existing orders to the foundry. 

Intel traditionally uses advanced nodes, so to land its orders, which are going to be significant, TSMC will need to have enough capacity to serve the CPU giant along with other customers.

Summary

 TSMC's management plans to increase its 2021 CapEx budget by 45% ~ 62% year-over-year to record $25 billion ~ $28 billion in a bid to expand its leading-edge production capacities and prepare to ramp up production using its N3 process technology in the second half of 2022. 

The company sees the ongoing megatrends like 5G, artificial intelligence, high-performance computing (HPC), and edge computing as the main drivers for advanced semiconductors demand in the coming years. 

Apple, which is already TSMC's main customer for advanced nodes, will likely significantly increase its orders to the foundry in the coming years as it transits to its own Apple Silicon SoCs for PCs and in-house designed modems, which will further drive demand for TSMC's leading-edge manufacturing technologies. 

Intel is also going to outsource more of its products to TSMC in the coming years, which will be another factor that will increase demand for the foundry's sophisticated fabrication processes. 

Additional features on the ROG Maximus XII Apex include several fan headers, water flow and temperature sensor headers, three M.2 sockets, 10 USB ports, 2.5 GbE and Intel AX201 Wi-Fi, and more. Below is a complete list of specifications from Asus. Below, we’ll walk through the many features of this unique board.

Asus ROG Maximus XII Apex is an extreme overclocking-capable motherboard that fits in the upper mid-range of Z490 motherboard pricing, currently listed at $356.99 at Newegg. Asus says the Apex has an optimized layout with enhanced power delivery and control and a unique design with several customization options. Additionally, the company has crammed in enough buttons, switches and BIOS options to handle anything in the sub-ambient overclocking world. Not only is it designed for overclocking, but it also looks good and easily handles most any other use case for a PC.

Although there are many capable boards around its price point, none focus on extreme overclocking like the Apex. If pushing things to the limit using LN2 or any other sub-ambient cooling method, this is one of a few boards created with that in mind (does anyone else miss the ASRock OC Formula?). Features include switches to enable “LN2” mode, Slow mode, a Safe boot button, voltage read points and more. But it isn’t all about overclocking. The Apex is plenty capable of gaming or productivity. It includes a premium audio codec, ample storage options, an inclusive cooling ecosystem, and more.

Performance on the Apex was right around most other boards, particularly those with a bit more freedom from the Intel specification. Gaming performance was on par or better than most boards, and the other data sets showed similar results. As far as overclocking, we had zero issues pushing our Intel Core i9-10900K CPU to 5.2 GHz. RAM overclocking was painless, easily handling our DDR4 4000 kit. However, the board set the System Agent (VccSA) and IO (VccIO) voltage higher than most.

Specifications

Starting with the accessories, Asus includes quite a bit here. In the box along with the motherboard is the usual list of SATA cables, Support DVD, M.2 screws, RGB extension cables and more. Below is a complete list of what’s in the box with the board.

●      Support DVD

●      ROG big sticker

●      Q-Connector

●      M.2 screw kit (long screw and mount)

●      (2) M.2 screw kit (short screw and mount)

●      Wi-Fi Antenna(s)

●      Extension cable for RGB strips (80cm)

●      Extension cable for Addressable RGB strips (80cm)

●      User Manual

●      ROG coaster

●      Customizable Name Plate accessory pack

●      ROG Thank you card

●      ROG DIMM.2 with heatsink

●      M.2 pad for ROG DIMM.2

One of the first things you’ll likely notice on the Maximus XII Apex is the cutout on the right edge. This ‘notch’ removed from the board not only gives the Apex a unique look, it also helps with cable management for the SATA ports. Outside of that, the PCB itself is jet black with all-black slots/connectors. There are large heatsinks for cooling the VRM (the IO shroud is part of this), connecting to the chipset heatsink. The ribbed pattern increases surface area and extends from the chipset area and its customizable plates up through the rear IO.

On the RGB front, the Apex takes a subtle approach, with a few lights under the heatsink/shroud for the rear IO area, the chipset heatsink (the ROG symbol is illuminated), as well as a couple more found underneath the board by the SATA ports. The lights are bright and the colors saturated. If what is included isn’t enough, there are plenty of headers to add on. For the most part, the Maximus XII Apex looks the part of a premium Z490 motherboard and will fit in with most build themes.

Focusing on the top half of the board, we get a close look at the beefy heatsinks that surround the socket area, as well as some of the buttons, switches, slots and LEDs. Starting in the upper-left corner, we see two reinforced 8-pin EPS connectors (one required) for powering the CPU. To the left of these is the first (of nine) 4-pin fan/pump headers. All of these headers support a maximum current of 12W/1A, except for the water pump header able to output 36W/3A. The W_Pump+, AIO_Pump and FS_Fan1/2 run full speed, while the chassis and CPU fans are Q-Fan-controlled, supporting both PWM and DC models. There are plenty of headers for fan and pump control here. Between this and the water flow/temperature headers, you’re likely able to run your whole air or water cooling ecosystem from the board.

To the right of the socket, we see a lot going on. From the DRAM and DIMM.2 slots (which can house risers for adding more M.2 drives) to all the buttons and switches, it’s pretty busy here. First, we see the Apex uses only two DRAM slots, where most other boards have four. With the Apex focused on overclocking, the thought here is that copious amounts of RAM aren’t the norm in this world. That said, the board will still support up to 64GB, which is plenty for an overwhelming majority of users. Supported memory speeds on the webpage list up to DDR4 4800(OC), which is up there with the other high-end motherboards.

The DIMM.2 slot to the right fits the included DIMM.2 add-in card, which supports two M.2 devices, one on each side and both under the heatsink. The DIMM.2 module supports PCIe and SATA-based M.2 modules up to 110mm, with a maximum PCIe 3.0 x4 bandwidth on each. The DIMM.2 slot also sports heatsinks for each module to help keep some of these hot running drives cooler. I have to say, I like the DIMM.2 slot implementation as it gets most of the M.2 devices out from under the video card/PCIe area, so they don’t have to deal with the additional heat.

Just above these slots is another fan header, two-character debug-LED, and two RGB headers (3-pin and 4-pin). Zooming in close to the upper right-hand corner, we see several LEDs between the start button and DIMM.2 slot. The first set of four LEDs on the left light up to show where the board is in POST. If the device doesn’t POST, one of these lights (CPU, VGA, RAM, Boot) stays lit and tells you where the issue is. Combined with the 2-character LED, you’re able to see details on why the system is hanging before reaching the OS. The other three lights are condensation sensors for the CPU, DRAM and PCIe slot. The debug and condensation sensors are important to the extreme overclocker as condensation and boot errors when dialing in an overclock can be a huge problem, especially when using sub-ambient cooling methods.

On the far right edge is a slew of buttons and switches. Starting at the top is a start button to power the system on/off, a configurable flex key button that reboots the PC by default. This is also configurable and used to turn Aura lighting on or off, safe boot and other functionality.

Below these are additional switches for extreme overclocking. From top to bottom are controls for LN2 mode, Slow mode, and pause functions, with the bottom two switches reserved for future use. Just below these are two simple buttons; a retry button and safe boot. The list of tools for the overclocker keeps growing!

Just below this is the 24-pin EATX power lead for the motherboard, another fan header (CHA_FAN2), and a front USB 3.2 Gen2 Type-C port.

If a board was designed for extreme overclocking, you have to imagine power delivery needs to be up to snuff to handle the power going through it. To that end, Asus uses a 16-phase VRM with 70A MOSFETs (total of 1,120A for the CPU). Managing the power is the company’s own Digi+ ASP1405I (X+Y=8) controller, which then sends it to the ‘teamed’ VRM configuration. The power stages themselves consist of 70A Infineon TDA21462 chips on the Vcore phases. This setup is plenty to drive any current Intel CPU, even under the additional stresses of extreme overclocking. You won’t be held back by this VRM here, that’s for sure.

Like the top half, the bottom is also busy. housing the audio section, PCIe slots, as well as additional switches and buttons across the bottom. Starting with audio, hidden under the SupermeFX EMI shield is the SupremeFX S1220A codec. This Asus-modified Realtek codec supports 10 DAC channels for simultaneous 7.1 and 2.0-channel playback. The Apex uses switching MOSFETs for the front or rear port’s impedance sensing capabilities. While this may not have the fancy DACs the Hero or Extreme have, we do see the popular Nichicon audio caps. In the end, most users will be pleased with this audio solution.

In the middle of the board are four PCIe slots. There are two full-length reinforced slots and two smaller (x1 and x4) slots. The two primary GPU slots run at x16/x0 or x8/x8. This configuration supports AMD 2-Way CrossfireX and Nvidia 2-Way SLI. The other two slots source bandwidth from the chipset, with the larger slot running at PCIe 3.0 x4 speeds, the smaller PCIe 3.0 x1.

We usually talk about all the M.2 sockets in this area, but the Apex only has one here. This socket supports up to 80mm devices and only supports PCIe 3.0 x4 mode modules. Between the DIMM.2 module and this socket, there is some lane sharing with the SATA ports. If DIMM.2_1 is running in SATA mode, the SATA port 2 is disabled. When DIMM.2_2 is populated, SATA ports 5/6 are disabled.

On the right edge of the board where the notch is are the eight SATA ports. Six of these are sourced from the chipset, while two come from an ASMedia 1061 chip. Both sets of ports support RAID0, 1 and 5. Last in this area is a front-panel USB 3.2 Gen1 header.

The board’s bottom edge holds several headers, buttons and switches, including more USB ports, temperature sensor headers, and more. Below is a list of all headers/buttons/switches on the bottom in order from left to right.

●      Front panel audio

●      Node header

●      Molex power plug

●      (3) USB 2.0 headers

●      AURA Addressable Gen 2 header

●      FS Mode header

●      Thunderbolt header

●      Water In/Out/Flow headers

●      (3) fan headers

●      Temperature sensor header

●      Front panel

●      BIOS switch

The integrated rear plate on the Apex is full of IO, including dual PS/2 ports (keyboard and mouse), BIOS switch and Clear CMOS button, Wi-Fi antenna headers, a 5-plug audio stack plus SPDIF output, the 2.5 GbE RJ-45 port, and 10 USB ports (6x USB 3.2 Gen2, including one Type-C, four USB 3.2 Gen1 ports). 10 ports should be plenty for most users. We don’t see a USB 3.2 Gen2x2 Type-C port, but this board targets the hardcore overclocker, not power users who require ultra-fast USB support.

Software

Asus has several applications designed for various functions, ranging from RGB lighting control, audio, system monitoring, overclocking and more. Instead of plodding through each application as if it changes for each review, moving forward, we’ll capture several screenshots of a few major utilities. In this case, here’s a look at Ai Suite 3, Armory Crate, and the Realtek Audio application.

Firmware

Like the software section above, we’ve shortened the Firmware section as well. Instead of describing each section, we’ve gathered screenshots covering the vast majority of the bios screens. If there is anything of interest, we’ll make a note of it.

I’ve been a longtime fan of the Asus UEFIs as they are easy to get around, have a lot of options, and the most frequently accessed items are not buried deep within the menus. The BIOS is high-contrast and easy to read, too. The EZ Mode provides enough information and options to be useful, while the Advanced part of the BIOS has everything you need, and more, to tweak your motherboard.

Worth noting on these Asus boards: On the first boot or after a CPU change, a prompt asks you to choose or avoid Intel specifications by pressing F1, which locks the BIOS down to Intel specifications (for power limit, etc.). The other option, F3, loosens up the stock limits, allowing the CPU to stretch its legs. We use the F3 option to match other board partners who by in large do not conform to the Intel specifications by default.

Test System

Our test system uses Windows 10 64-bit OS (1909) with all threat mitigations applied. The motherboard BIOS used is the latest non-beta available to the public, unless otherwise noted. The hardware used is as follows:

Benchmark Settings

MORE: Best Motherboards

MORE: How To Choose A Motherboard

MORE: All Motherboard Content 

Benchmark Results and Final Analysis

All standard benchmarks and power tests are performed using the CPU’s stock frequencies (including stock Thermal Velocity Boost), with all power-saving features enabled. Optimized defaults are set in the BIOS and the memory is set using the XMP profiles. For this baseline testing, Windows is set to High Performance, except during power testing, where we switch to Balanced so the PC idles appropriately.

**The red bar represents the board we are reviewing, not the fastest/slowest result for a given chart.

Synthetic Benchmarks

Synthetics are a great tool to figure out if a board runs out of spec, as identical settings should produce remarkably similar performance results. Advanced memory timings are the one place where motherboard makers can still optimize for either stability or performance, though, and those settings can impact some testing.

The Apex did well across most tests here, due to using its ‘optimized’ mode as opposed to the stock Intel limits. Overall, the board was as fast as or faster than most other boards tested. We saw impressive results in 7Zip, Cinebench, POV-Ray and most PCMark10 applications. ALBUM - image3/7, 15/16

Results from our timed applications were similar to the synthetics. In Handbrake, for example, we saw the fastest time in the x264 test and a close second in the x265 portion. Our LAME results continued this trend and are also the fastest we’ve tested so far (by a minuscule margin). The Corona results saw similarly good results for Asus.

3D Games and 3DMark

We’ve recently updated our game tests to The Division 2 and Forza Horizon 4. The games were run at 1920x1080 resolution using the Ultra preset. As the resolution goes up, the CPU tends to have less of an impact on most games. The goal with these settings is to determine if there are differences in performance at the most commonly used resolution with settings most people use or at least strive for.

When looking at gaming, we found the Maximus XII Apex to be just as fast as the other boards as expected. No worries here either!

Power Consumption / VRM Temperatures

For power testing with our current-hungry Core i9 CPU, we used AIDA64’s System Stability Test with Stress CPU, FPU and Cache enabled using peak power consumption value within the first minute. The wattage reading is from the wall via a Kill-A-Watt meter to capture the entire ecosystem (minus the monitor). The only variable that changes in these reviews is the motherboard; all other parts are the same.

Starting with idle power, our setup idled at 55W when using the balanced power plan. This value sits around the mean of the results. During stress testing, the Apex peaked at 269W, again landing in the middle of the other motherboards. The average power use of 162W is on the lower side of things, matching its big brother, the ROG Maximus Extreme.

VRM temperatures on our Apex were well within operational parameters of the VRM, peaking around 48 degrees Celsius at stock speeds and 54 degrees Celsius when overclocked. Compared to the Aorus Extreme, MSI Godlike and the Asus Extreme, its temperatures were right in the other high-end boards’ ballpark.

Overclocking

In short, we had no significant issues overclocking the Apex. Our Intel Core i9-1900K CPU ran at 1.35V and 5.2 GHz without any problems. The only changes we made outside of voltage and multiplier was to adjust Load Line Calibration (LLC). Asus recommends using Level 4. However, this setting yielded a fair amount of vdroop. Setting the LLC up to Level 7 minimized the voltage sag. The sixteen 70A MOSFETs had no issues driving this power-hungry processor.

Switching to memory speeds, the Apex ran our DD4 4000 kit just by setting XMP and rebooting. The concern here is how high the System Agent and VccIO voltage was set. In this case, SA voltage hit 1.5V and IO set to 1.42V. These values are WAY too high for what we needed. Though setting this voltage higher in favor of stability is expected, those voltages are incredibly high. I set these values back to 1.2 and 1.1V and we completed the tests without issue.

Final Thoughts

If you’re looking for a board designed for extreme overclocking that can do just about everything else well, look no further than the Asus ROG Maximus XII Apex. Currently priced at $356.99 on Newegg, it isn’t cheap and there are some limits, specifically on the RAM capacity side. If you need more than 64GB, you’ll have to look somewhere else. While other motherboards may have similarly capable VRM configurations, most don’t have the additional features that help with extreme overclocking (condensation sensors, LN2/Pause modes, etc.).

Outside of that, the robust VRM, AX201 integrated Wi-Fi, premium audio codec, eight SATA ports and three M.2 modules (two via the DIMM.2 daughter board) help to support the price point. To that end, This board competes with the ASRock Z490 Taichi ($369.99), Asus ROG Maximus XII Hero ($393.99), Gigabyte Z490 Aorus Master ($329.99), and the MSI MEG Z490 Ace ($399.99). These boards will handle overclocking and even sub-ambient overclocking, but none are explicitly made for that or have as many features for that purpose.

The Apex’s unique styling with the cutouts on the board should also a positive for most. It will blend in well with any build theme, though if additional shrouds are your thing, the other boards listed above do have a more ‘clean’ appearance.

In the end, the Asus ROG Maximus XII Apex performed well out of the box with the optimized settings, has plenty of overclocking-specific hardware and functionality, and looks like an upper mid-range Z490 part. There are plenty of USB ports on the rear IO (10), though I would like to have seen a USB 3.2 Gen2x2 Type-C port out back. If you don’t plan on using sub-ambient cooling methods, the Maximus XII Apex is still a well-balanced board, though the price may be off-putting, since you’d be paying for features you won’t use. If you’re looking for a motherboard around $350 and heavy/extreme overclocking is a priority, the Apex should be on your shortlist.

MORE: Best Motherboards

MORE: How To Choose A Motherboard

MORE: All Motherboard Content 

Maingear’s Vybe high-end desktop impressed us last year when we reviewed it both as a fully configured desktop and as a PC case to build your own system inside. But now that Intel has released the tough-to-cool Core i9-10900K, Maingear sent an updated Vybe our way, with liquid cooling for both the CPU and graphics card (an Nvidia GeForce RTX 2080 Ti in this configuration), and the company’s stunning custom Apex Cooling option. 

Is the new Vybe one of the best gaming PCs? Well, that depends on what your budget is (sure, it starts at $749, but as configured, this system is $5,402), and how much you like Intel. Because Maingear will also sell you a version of the Vybe with an AMD Ryzen system instead, and unless you really care about gaming at high frame rates and 1080p, that may be a better option. 

Maingear Vybe Design

There’s nothing new about the Vybe’s design at this point, but that certainly isn’t a bad thing. At 18 x 8.5 x 18.7 inches (HWD), it’s a mid-tower system that’s far from small, but it’s not massive either. Maingear sent us the black model, but you can choose white or gray when configuring the system at no extra charge. And if none of those excite you, you can choose an automotive paint option in just about any color you want -- for an additional $399 - $699, depending on the paint.

The case features a tempered-glass side to show off your components, removable filters on both the top and bottom, and an ample selection of front-panel ports. On the left edge, near the front, you’ll find three USB 3 Type-A ports, one Type-C, as well as separate headphone and mic jacks and a square power button outlined by a ring that lights up when the system is powered up. 

Ports around back depend on which motherboard you choose, and Maingear currently has four offerings on the Z490 front -- ours came with the Asus ROG Maximus XII Hero (Wi-Fi). With that board, you get both Intel 1Gb Ethernet and a 5Gb port from Marvell, along with WiFi 6 (AX201). USB ports here total 10, with four being USB 3.2 Gen 2 (three USB-A, one USB-C), four USB 3.3 Gen 1 Type-A, and a pair of USB 2.0. There’s also the standard allotment of audio jacks and an HDMI port that you’ll generally want to avoid in favor of the ports on the RTX 2080 Ti. On the card below the board, you’ll find three DisplayPorts, one HDMI, and a VirtualLink (Type-C) port for connecting displays. 

Maingear Vybe 2020 Specifications

Components and Upgradeability

Inside our review $5,402 review configuration lives a new 10-core Core i9-10900K processor and an RTX 2080 Ti ⁠— both chilled by the company’s Apex liquid cooling (a pretty and powerful, but expensive, $500 addition). Memory is handled by HyperX, in the form of a 32GB kit clocked at 3200 MHz, and storage is a combination of a 1TB Intel 665P SSD and a 4TB WD Black hard drive.

In particular, the storage setup here seems a little odd. It’s nice to have the spaciousness of a spinning-platter drive, but the Intel 665P boot drive is both cramped for a system this expensive, and not all that fast. It’s also a QLC drive, which will give some qualms about endurance. But my bigger qualm is its combination of speed and capacity in a system this expensive. The rated 2,000/1,925 sequential read/write speeds are decent, but if you manage to fill up the cache (which we calculated at about 140GB in our review testing of the drive), performance drops way down to about 185 MBps. Granted, you’d have to be writing a lot of data from a similarly speedy drive to make that happen, but that’s not a situation you should have to worry about on a system you’re spending this much on. And the fact that the drive is only 1TB means you’ll have to install most of your game library on the much slower HDD. 

When it comes down to it, if you’re spending over $5,000 on a system, you should probably spend more than about $140 (the current rough going rate for the 665p) on the boot drive. Thankfully, this is a configurable system and Maingear also offers speedier, roomier drives as options from both Samsung and WD. We’d choose one of those, even if it meant spending less on flashy things like lighting. Skipping the RGB fan upgrade alone (not really a necessity given an interior RGB light strip comes standard) will save you $200 that would be much better spent on roomier, speedier storage.

There’s plenty of room for expansion here, though. Two of the four RAM slots are empty, as are two of the three M.2 slots on the Asus motherboard. Note, though, that you’ll have to remove the graphics card to get at one of these, which is a much more daunting task given the hard tubing full of coolant. 

There are also five empty SATA ports for adding even more storage. The board has three spare x1 PCIe slots and two spare x16 slots for adding expansion cars. But again, all of this hinges on your board choice, so choose wisely from the available options if configuring a Vybe yourself.

Gaming and Graphics

The Core i9-10900K in the Vybe may be new, but its RTX 2080 Ti is going on two years old. Given that, we don’t expect the Vybe to perform substantially higher than other high-end gaming desktops we’ve tested in the last year or so, particularly at resolutions above 1080p. Our competing systems are HP’s Omen Obelisk (Core i9-9900K, RTX 2080 Ti), Alienware’s Aurora R10 (Ryzen 9 3950X, RTX 2080 Ti), and the CLX Ra (Ryzen 3900X, RTX 2080 Ti). 

In Shadow of the Tomb Raider (highest settings), the liquid-cooled 10th Generation Core i9 CPU helped the Vybe stomp the competition at 1080p, with its 124 fps beating everything else by at least 20 frames. But bumping the resolution up to 4K, things were much closer. The Vybe’s score of 45 fps was second best, just behind the CLX Ra’s 46.

On Far Cry New Dawn (ultra), the Vybe’s 1080p score of 112 fps was again well ahead of competing machines. And even at 4K, its score of 80 fps was 9 frames ahead of the closest competitors.

On the Grand Theft Auto V benchmark (very high settings), the 1080p trend continued, with the Vybe delivering a class-leading 137 fps. And once again, at 4K things were much closer. The Vybe’s 45 fps was a few frames behind the Ra’s 48 fps, but better than the other two competitors.

We don’t have a full set of test numbers for Red Dead Redemption 2 (medium settings), but compared to the Alienware Aurora, the Vybe led at 1080p (88 fps versus 80 from the Alienware), but fell behind at 4K (33 fps for the Vybe compared to 42 on the Alienware). 

We also subjected the Vybe to our Metro Exodus gauntlet, in which we run the benchmark at the RTX preset 15 times to simulate roughly half an hour of gaming. On the test, the Maingear machine ran the game at an average of 89.2 frames per second (fps), with little variation. The system started out the test at 89.7 on the first run, and only dipped to 88.97 once, on the 12th run before edging back up.

During the Metro Exodus runs, the CPU ran at an average clock speed of 4.9 GHz and an average temperature of 64.3 degrees Celsius (147.7 degrees Fahrenheit). The GPU’s average clock speed was 1.8 GHz, with an average temperature of 54.7 degrees Celsius (130.5 degrees Fahrenheit).

Throughout the test run, CPU temps were almost always well below 70 degrees C, save for a handful of short spikes. And as a result, we also saw cores occasionally spike to the CPU’s rated top Turbo speed of 5.3 GHz. Maingear’s Apex cooling system is doing its job keeping the Core i9-10900K cool enough to achieve its peak performance at stock settings, and the GPU performed consistently as well. 

Productivity Performance

The components in our configuration of the Vybe make for a powerful productivity machine as well as gaming, though if you can make use of lots of cores, it’s easier to recommend an AMD-based configuration. As we’ll see below, the 12 cores of Ryzen-3950X-based Alienware Aurora ($3,629.99 at time of testing) often push it ahead of the $5,402 10-core 10900K-based Vybe. 

On Geekbench 4.3, the Vybe edged past the HP Omen Obelisk and AMD-based CLX Ra from last year, but the extra cores of the Aurora give it a clear advantage.

That carried over to our Handbrake video transcoding test, where the Vybe took 5 minutes and 35 seconds to convert a 4K video to 1080p. That’s neck-and-neck with most of the competition, but the Aurora R10 is a full two minutes faster.

Lastly, our comments earlier about the speed of the Intel 665P SSD are proven true in our file transfer test. It took the Vybe 6 seconds to transfer 4.97 GB of files, which translates to 848 MBps. That’s not awful, but it’s easily the poorest showing among the competing machines in our group, and just about half the speed of what HP’s Omen Obelisk delivered. 

Software and Warranty

Maingear sells the Vybe with a one-year full-coverage warranty as standard, though you can extend that up to three years for $100 a year ⁠— an easy recommendation if you’re buying something this expensive and complicated.

There’s also nothing to mention in terms of bloatware. Maingear ships the Vybe (and all its systems) with a zero bloatware guarantee. The only thing we noticed installed on the system when we took it out of the box and booted it up was Asus’ Aura Sync software for controlling the lighting.

Cost Considerations and Bottom Line

Aside from the Apex cooling option, pretty much all of the Vybe’s components and parts are off the shelf, save for Maingear’s exclusive Apex cooling setup. But it’s still hard to put a price on some things. For instance, the Core i9-10900K is still hard to find in stock at or close to the MSRP. Still, at a rough estimate, the core parts here would set you back about $3,000. That’s not including the custom Bitspower hard tubing and fittings, the Apex cooling pump/reservoir, or the CPU and GPU waterblocks. It also neglects the system’s six RGB fans and custom sleeved cabling.

All of these obviously add to the cost. And the work Maingear does assembling the whole system and tuning the fans and cooling for surprisingly quiet operation under load add a great deal to the overall enjoyment of the system. Bottom line on the cost: You could easily build a system to deliver this much performance (and with a better boot drive) for well more than  $1,000 less than our config of the Vybe. But it wouldn’t be this polished or pretty. And chances are if you’re considering shelling out this much for a custom desktop, you aren’t likely to seriously consider building your own system anway. A multi-thousand-dollar custom rig is usually at least as much about the experience as it is about anything else. And on that front, Maingear definitely delivers with the Vybe. 

MORE: Best Gaming PCs

MORE: How to Build A PC

MORE: All PC Builds Content

If you’re after a high-end gaming experience that rivals the upcoming Xbox and PlayStation while keeping things compact and living room-friendly, Maingear’s brand-new Turbo desktop might be just what you’re looking for.

At 12.3 x 6.7 x 14.4 inches (HxWxD), the Turbo should fit on just about any desk and won’t dominate your living room space either. Designed in partnership with AMD, the desktop will ship exclusively with Ryzen CPUs (sorry Intel fans), including the brand-new Ryzen 3000XT models, topping out with the 12-core Ryzen 9 3900XT. But if you need even more CPU cores, the new desktop can also support the 16-core Ryzen 9 3950X, alongside AMD graphics cards, like the RX 5700 XT, or Nvidia cards up to the Titan RTX. 

Cooling happens vertically in this compact chassis, with cool air coming in at the bottom and expelled out the top. The company’s gorgeous Apex liquid cooling system is also available with the Turbo, which chills the processor and graphics card with the help of a reservoir carved out of a solid block of acrylic, custom tubing and a quiet pump. There’s also a PWM fan controller on board to make sure the spinners don’t make any more noise than they need to.

As this is a custom system, there are loads of component choices available with the Turbo, including up to 16TB of storage and 64GB of RAM, as well as Windows 10 Home or Pro (with a “zero bloatware guarantee”). 

Customizable artwork is also an option with this system, although you’ll have to pay more than the $1,499 starting price if you want to get fancy with the exterior. 

The Maingear Turbo compact desktop is available today directly from Maingear. 

Intel's Comet Lake recently crashed into the market with the beastly $488 Core i9-10900K leading the charge, showing us that its aging Skylake architecture and hyper-refined 14nm++ process still has the gumption to keep the company's gaming performance crown. But the ten-core 20-thread processor brought some baggage along with it, like high power consumption that requires exotic accommodations to get the best performance possible, pricing it out of consideration for the majority of enthusiasts. 

However, Intel's clever price trimming on its Core i5 series, which comes via the addition of Hyper-Threading at no extra cost, has always looked like a possible addition to our Best CPUs for gaming list. After our testing, the $262 Core i5-10600K delivers and is just as impressive as you would expect from a six-core 12-thread Comet Lake processor clocked at a 4.1 GHz base and 4.8 GHz boost. Intel also offers the chip for $237 if you grab the graphics-less KF model. 

In either case, the 10600K comes with more threads than the previous-gen Core i7 models, so the implication is clear: You get the lion's share of the performance of last year's $370 Core i7-9700K, but for $267/$234 with the Core i5-10600K. 

Intel's pricing puts AMD's ~$245 Ryzen 5 3600X squarely in the crosshairs. While the ~$175 Ryzen 5 3600 is still tempting with its six cores and 12 threads, Intel hopes the 10600K's gaming supremacy, not to mention overclocking prowess, persuades enthusiasts to drop the cash on a new LGA1200 platform to support the chip. 

And Intel's Core i5 series could definitely use some help -- the third-gen Ryzen processors, for all of their headline-stealing high core counts in the higher-end models, did the most damage in Intel's mid-range due to their superior value and ultra-competitive gaming performance. The Core i5 and Ryzen 5 segments comprise the bulk of sales to the ever-growing cadre of gamers and enthusiasts looking for the best bang for the buck, so success here is key. 

The Core i5-10600K's combination of a higher thread count at similar pricing to the previous-gen, high stock clock frequencies, palatable power consumption, and agile overclockability cooks up a winner for the gaming and enthusiast crowds. Intel even reduced the gap in threaded workloads like productivity tasks. For gamers looking for the edge of performance and enthusiasts who like to tune their processors without hideously-priced supporting components, the Core i5-10600K slots in as the new mainstream champ. 

Intel Core i5-10600K Specifications and Pricing

Intel re-aligned its Core i5 series, which came with six cores and threads with the previous-gen chips, to six cores with 12 threads for the Comet Lake models. That comes by simply enabling Hyper-Threading, but it boosts the Core i5-10600K's thread count beyond that of the previous-gen eight-core/thread i7-9700K, setting the stage for impressive performance improvements. 

And the new chip didn't disappoint. Our testing found that the 10600K still trails the 9700K in application performance, but it's close enough in gaming to call it a wash. That's a huge step forward on the value front. Pricing is competitive, too, especially with the $25 discount for picking the graphics-less 10600KF model. If you're not interested in buying a discrete graphics card, the 10600K's UHD Graphics 630 engine remains the same compared to the prior gen, right down to the clock rates. That gives Intel a leg up over AMD, which doesn't offer integrated graphics in this price range. Just don't expect to do any meaningful gaming on the Intel chip without a dedicated graphics card.

The six-core, twelve thread Core i5-10600K ticks at a 4.1 GHz and 4.8 GHz boost, and unlike the Core i9 models, Intel doesn't include it's complicated matrix of TurboBoost 3.0 Max and Thermal Velocity Boost (TVB) clock rates, making this a simple case of what you see is what you get. 

As you can see in the table above, Intel has steadily bumped up its multi-core clock rates with each new generation of chips, with the 10600K marking yet another jump over the previous-gen Core i5 chips. The 10600K still trails the Core i7-9700K's single- and all-core boosts, but offsets it with a much higher 4.1 GHz base (and accompanying TDP rating). That means the 10600K won't entirely match the previous-gen i7 in all applications, but our testing shows it comes close. 

The 10600K comes with a 125W TDP rating, a new high for Intel's Core i5 series as the company turned up the power dial on its base clock rates from 3.7 GHz with the previous-gen Core i5-9600K to 4.1 GHz. Still, it's important to remember the TDP rating doesn't have a one-to-one relation to power consumption, and it only applies to stock settings at the base frequency (PL1). Intel recommends a maximum of 220W at PL2 (peak power consumption during boost), but as always, motherboard vendors are free to ignore those recommendations, so board selection is important. 

The 10600K comes with 12MB of L3 cache, an increase of 3MB over the Core i5-9600K. Intel increased memory support up to DDR4-2933, which lags AMD's DDR4-3200 memory transfer rates. Like the chip's other ratio multipliers, Intel fully enables memory overclocking on Z-series motherboards, but you'll lose that functionality on the B- and H-series. Be sure to price in a Z-series board and a cooling solution, preferably liquid, if you're off to the overclocking races. Meanwhile, AMD allows overclocking on all but its A-Series motherboards. 

All of Intel's Comet Lake K-Series models come with a thicker copper integrated heat spreader (IHS), a thinner die, and solder TIM (sTIM) to improve cooling and overclockability, which we covered here. The 10600K doesn't come with a boxed cooler, but mid-range air coolers should be sufficient for stock operation, and even some overclocking. Serious overclockers should plan on a 240mm or greater all-in-one cooler, but the overclocking results we'll outline on the following pages are pretty impressive. 

Intel Core i5-10600K Overclocking Settings

Intel has simplified the overclocking process over the years, and motherboard vendors have done a great job with adaptive Load Line Calibration (LLC) features. We dialed in a 5.0 GHz overclock with a 1.34V vCore and left the motherboard to its own devices for LLC control (auto), giving us a steady 1.368V under heavy load. We did have to settle for a -1 AVX offset though, meaning our chip operated at 4.9 GHz during ultra-taxing AVX workloads. 

As we did with the Core i9-10900K, we dialed the ring bus multiplier to 48 to improve stability and performance, but that's an optional tweak. We also ran our overclocked configuration with memory at DDR4-3600 with 18-18-18-36 timings. 

Our overclock challenged the Corsair H115i during a Prime95 stress test, topping out at 93C, but that stress test isn't indicative of what we see during the real-world AVX workloads we recorded above. As you can see above, thermals were good with the Corsair H115i, albeit with the fans cranking at full speed, during a string of real-world stress tests that includes multiple instances of the Corona ray-tracing benchmark, x265 HandBrake rendering tests, POV-Ray multi-threaded benchmarks, Cinebench R20 runs, and finally five iterations of the AVX-intensive threaded y-cruncher. 

This real-world stress test yielded peak temps of 88C, but those peaks were brief. Power consumption weighed in at a peak of 208W, which, while a new high in recent memory for Core i5, isn't excessively hard to cool with off-the-shelf cooling solutions. 

Intel did a good job here of reducing thermal output by using a thicker IHS and thinner die. The Core i5-10600K, despite its comparatively high power consumption relative to the Ryzen 3000 series processors, is surprisingly easy to cool. Our Corsair H115i yielded solid overclocking results, but you could get away with a capable high-end air cooler, like the Noctua NH-D15, as long as you set your all-core overclock sights a bit lower (in the 4.8 GHz range). 

Overall, the Core i5-10600K is an easy overclocker, and we have little doubt that a bit more tuning could unlock more performance. In fact, we kept our overclock in check to match what you could reasonably sustain with an AIO, but we think we could get to 5.1 GHz with a 5.0 GHz AVX offset and higher voltages with a better cooler. 

Intel Core i5-10600K Turbo Boost and Thermals

Switching over to the stock settings on the Gigabyte Aorus Z490 Master motherboard, we ran the same series of stress tests to measure the chip's power consumption, thermals, and boost capabilities. As per normal for any of our testing of stock settings, we disabled the MCE (Multi-Core Enhancement) feature that amounts to automatic overclocking (Enhanced Multi-Core Performance in Gigabyte's parlance). 

During our multi-core frequency test (first and second slides) with our stress test script, we see that the chip boosts to 4.8 GHz during pauses in the workloads, but maintains a flat 4.5 GHz all-core frequency for most of the test. The chip peaks at 65C and 130W of power consumption during the latter stages of this test. A quick run with Prime95 measured a peak of 146W. 

Slides three and four detail performance during our standard frequency test for lightly-threaded workloads (methodology here). Here we see much more frequent boost activity to 4.8 GHz, and power consumption and thermals weren't really an issue during the test. 

The Corsair H115i handles the Core i5-10600K with little effort during our stress tests, Prime95 included, signaling that this is an easy-to-cool chip, and air cooling is certainly an option. 

Intel Core i5-10600K Power Consumption and Efficiency

We logged 107W of power consumption for the Core i5-10600K at stock settings during the AVX-intensive multi-threaded y-cruncher workload, landing well above the previous-gen 9600K, but surprisingly 47W shy of the Core i7-9700K. However, kicking up the 10600K's clocks to 5.0 GHz yielded a 175W reading, which outstrips the 9700K at 5.1 GHz. As always, take power readings during overclocking with a grain of salt, as fine-tuning and motherboard firmwares have a huge impact here. The Core i7-9700K also benefits from a much more mature platform, so keep that in mind.

As expected, the Core i5-10600K consumes much more power than the competing Ryzen 3000 chips, but that isn't surprising given AMD's fast transition to a denser and more power-efficient 7nm node. That advantage really shines through when we take a look at power consumption during our HandBrake x264 and x265 workloads. Still, basic power consumption measurements never tell the full story – the Ryzen processors also dominate the renders-per-day efficiency metrics.

Here we take a slightly different look at power consumption by calculating the cumulative amount of energy required to perform an x264 and x265 HandBrake workload, respectively. We plot this 'task energy' value in Kilojoules on the left side of the chart.

These workloads are comprised of a fixed amount of work, so we can plot the task energy against the time required to finish the job (bottom axis), thus generating a really useful power chart. Bear in mind that faster compute times, and lower task energy requirements, are ideal. That means processors that fall the closest to the bottom left corner of the chart are best.

The Ryzen processors are pretty hard to beat at this metric. Still, the Core i5-9400F, with its fixed clock rate that is pretty low on the voltage/frequency curve (i.e., closer to the efficiency range), is surprisingly competitive with the Ryzen 3 3300X. That highlights that Intel tunes the Comet Lake chips more for performance than efficiency, but the Core i5-10600K does take a solid step forward over the Core i5-9600K. Meanwhile, the Ryzen 7 3700X is impressive with the best spot on the chart, and the Ryzen 5 3600(X) models put up a good show, too. 

MORE: Best CPUs

MORE: Intel and AMD Processor Hierarchy Comparisons

MORE: All CPUs Content

All test results marked with "PBO" reflect configurations tested with AMD's auto-overclocking Precision Boost Overdrive feature. For overclocking, we tuned our memory to DDR4-3600 for both Intel and AMD platforms.

We have a few interesting comparisons to keep an eye on as we work our way through the tests. Intel's previous-gen Core i7-9700K retails for $370, but comes with eight cores and threads while the Core i5-10600K lands at $262 with six cores and twelve threads, setting the stage for great gaming performance at a low price point. AMD's Ryzen 7 3700X retails for $290, representing the upper level of AMD performance in this general price range, while the $240 Ryzen 5 3600X and $175 3600 are in the mix with the Core i5-10600K. AMD's Ryzen 3 3300X slots in at a mere $120, but sets a high bar in gaming for the lower end of the market.

This set of tests focuses narrowly on this price band, but we have comparisons to the Core i9-10900K in gaming and application work on the final page of the review. 

VRMark and 3DMark on Intel Core i5-10600K

Synthetic gaming benchmarks often aren't generally indicative of real-world performance, but the 3DMark DX11 and DX12 tests are interesting because they measure the amount of raw computational horsepower exposed to the game engine. For now, most of today's game engines don't scale as linearly with additional compute resources, but these tests help us gauge how games could exploit processing resources as the engines become more sophisticated.

In the synthetic world of the Fire Strike benchmark, the Ryzen 7 3700X dominates due to its eight cores and 16 threads, while the Ryzen 5 3600X also carves out a lead over the stock Core i5-10600K. The stock Ryzen 5 3600 also nearly matches the 10600K, which is impressive given its $175 price tag. Overclocking propels the 10600K into the lead over its similarly priced competitors.

Flipping over to Time Spy finds the Core i5-10600K nearly matching the previous-gen Core i5-9700K at both stock and overclocked settings.

VRMark's test values per-core performance (a mixture of frequency and IPC), and it obviously prefers physical cores and lots of L3 cache. The Coffee Lake Refresh Core i5-9600K benefits from its lack of hyperthreading in this title, but it takes a healthy overclock to boost it over the 10600K. 

Civilization VI AI and Stockfish on Intel Core i5-10600K

Civilization VI's AI engine values per-core performance, which benefits Intel's higher clock rates. Here we can see the Intel chips without threading once again benefiting from the program executing on physical cores, but the stock Core i5-9600K lags due to its lower base and boost frequencies. 

Stockfish, an open-source chess engine, is designed to extract the utmost performance from many-core chips by scaling well up to 512 cores. The Ryzen 5 3600X's six cores and twelve threads prove powerful in this test, but the overclocked Core i5-10600K profits from its higher clock rates. The Ryzen 5 3600 also puts up a strong showing at stock settings, edging out the 10600K. 

Ashes of the Singularity: Escalation on Intel Core i5-10600K

Ashes of the Singularity: Escalation loves cores and threads, but clock rates also play a role. The stock Core i5-10600K vies with the Ryzen 5 3600X at stock settings, but its higher overclocking headroom grants it the overall win as it edges out the higher-clocked Core i7-9700K. It's noteworthy that the stock 10600K edges past the heavily-overclocked 9600K. Meanwhile, the Ryzen 3 3300X slips past the Core i7-9700K at stock settings, reminding us that AMD's Ryzen 3 has completely upset the value tier. We can't wait to see how Intel's Core i3 lineup stacks up. 

Civilization VI Graphics Test on Intel Core i5-10600K

Most gaming software is still optimized for low thread counts and high clock rates, and some game engines even prefer single-threaded cores. That doesn't stop the threaded 10600K from exerting itself in these tests, though, as it leads all but the overclocked Core i7-9700K.

Dawn of War III on Intel Core i5-10600K

The Warhammer 40,000 benchmark responds well to threading, but it's clear that clock speed and IPC also matter. The stock Ryzen processors fall down the charts in this title, though the Ryzen 3 3300X is particularly impressive due to its use of a single CCX – it even beats out its own more-expensive brethren and the stock 10600K. After overclocking, the Core i5-10600K takes the lead, albeit by the slimmest of margins over the Core i5-9700K. 

Far Cry 5 on Intel Core i5-10600K

Far Cry 5 tends to run well on Intel architectures, and the overclocked Core i3-9350KF reminds us that Intel's overclockability can be a powerful asset. Unfortunately, this chip represents the last overclockable Core i3 model in Intel's stable; Intel inexplicably neglected to release an overclockable Comet Lake i3 model. 

Regardless, the top of the chart belongs to Intel, and the 10600K once again delivers nearly the same performance as the previous-gen Core i7-9700K at stock settings and superior performance after overclocking. 

Final Fantasy XV on Intel Core i5-10600K

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 with the higher-resolution setting. The stock Core i5-10600K grapples with the mid-range Ryzen processors, and even the Ryzen 3 3300X, at stock settings, but overclocking propels it into second place. 

The Core i5-9400F, Intel's previous-gen locked i5 that was known for its solid performance, doesn't look nearly as impressive when stacked up against the latest models.  

Grand Theft Auto V on Intel Core i5-10600K

Here we see a familiar trend of the Core i5-10600K basically offering the same level of performance as the previous-gen unlocked Core i7 model, highlighting that industry competition has benefitted the consumer with big gen-on-gen value improvements. 

Hitman 2 on Intel Core i5-10600K

The Core i7-9700K carves out a substantial lead at both stock and overclocked settings compared to the 10600K here, but the latter clings to a comfortable lead over price-comparable Ryzen processors.

Project Cars 2 on Intel Core i5-10600K

Although Project CARS 2 is purportedly optimized for threading, clock rates obviously affect this title's frame rates. Intel's overclocked Core i3-9350KF pops up near the top of the chart, showing that clock rates are also important in this title. It even tops the stock Core i5-10600K, but overclocking propels the 10600K into a comfortable lead. 

World of Tanks enCore on Intel Core i5-10600K

The World of Tanks benchmark closes out our game testing with the same general trend we've seen throughout this series of tests: On the strength of its high single- and multi-core clock rates paired with 12 threads, the Core i5-10600K is a potent gaming chip, even at stock settings. Overclocking also yields significant gains that place the chip far ahead of its AMD competitors. 

MORE: Best CPUs

MORE: Intel and AMD Processor Benchmark Hierarchy

MORE: All CPUs Content

Rendering on Intel Core i5-10600K

Ryzen processors are apex predators in the realm of threaded workloads, but the Core i5-10600K marks a massive improvement for Intel's Core i5 series in these types of heavy applications. The 10600K scores big gains with its twelve threads relative to the 9600K's six threads, often beating the overclocked 9600K by substantial margins, even at stock settings. 

The Ryzen 5 3600X and 3600 still maintain their lead over the stock 10600K in threaded productivity applications, but the margins are far slimmer than we witnessed in the past. This time around, the Intel chip takes the lead after overclocking, marking a shift in the productivity landscape in this price range. The Core i5-10600K largely lags the Core i7-9700K, but again, by a slim margin. 

Moving on to the single-core rendering workloads, the Intel processors maintain a solid lead, particularly after overclocking. 

Encoding on Intel Core i5-10600K

The single-threaded LAME and FLAC encoding tests respond well to per-core performance, giving Intel convincing wins in these tests. 

The threaded HandBrake x264 and x265 tests really speak to the AVX performance improvements AMD made to the Zen 2 architecture. The stock Core i5-10600K lags the price-comparable Ryzen processors, but Intel has significantly reduced the deltas. Overclocking swings the HandBrake tests in Intel's favor. 

Web Browsing on Intel Core i5-10600K

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

Single-threaded performance still reigns supreme in these tests, making it hard to beat the overclocked Core i5-10600K. The chip is equally adept at stock settings, often beating its predecessor by solid margins. AMD's Ryzen processors are competitive, too, particularly in the Edge browser test that relies more on threading.

Office and Productivity on Intel Core i5-10600K

Clock rates benefit the majority of our GIMP image processing tests, but the Ryzen 3 3300X posts surprisingly strong performance borne of its design that leverages a single CCX. As expected, the Intel processors lead in the majority of the PCMark 10 tests, but the chips pick up a few wins at stock settings in the threaded rendering and visualization and video processing tasks. 

Compression and AVX on Intel Core i5-10600K

The 7zip and Zlib compression/decompression benchmarks also rely heavily upon threading and work directly from system memory, thus avoiding the traditional storage bottleneck in these types of tasks. At stock settings, the price-comparable Ryzen processors sweep the 10600K on the 7zip compression side of these tests, but overclocking improves Intel's standing considerably. 

The heavily-threaded y-cruncher benchmark, which computes pi using the taxing AVX instruction set, highlights the performance of the overclocked Core i5-10600K with its AVX offset clock rate of 4.9 GHz. The 10600K's extra helping of threads help boost performance significantly over the previous-gen 9600K in the threaded portion of the test. 

MORE: Best CPUs

MORE: Intel and AMD Processor Hierarchy Comparisons

MORE: All CPUs Content

AMD's Ryzen 3000 series processors upset the established pecking order in the high-volume mid-range, beating Intel's competing chips by huge margins in productivity-focused work while delivering similar, or slightly better, gaming performance at the most important price points. 

Intel's Comet Lake response helps reduce that imbalance with a heavier slathering of threads that improve its standing in productivity work and multi-tasking, accompanied by higher clock rates that carve out a perceptible lead in gaming workloads. In fact, dropping $262 on the Core i5-10600K gives you comparable performance to Intel's $370 Core i7-9700K that debuted last year. 

In the chart below, we plot gaming performance using average frame rates and a geometric mean of the 99th percentile frame times (a good indicator of smoothness), which we then convert into a frames-per-second measurement. Bear in mind that we tested with an Nvidia GeForce RTX 2080 Ti at 1920x1080 to alleviate graphics-imposed bottlenecks. Differences between our test subjects shrink at higher resolutions. We've included the Core i9-10900K in these charts to highlight its position relative to the Core i5-10600K.

At stock settings, the Core i5-10600K runs neck-and-neck with the Core i7-9700K throughout our gaming test suite, and overclocking grants it the slimmest of leads despite its 100 MHz deficit. Meanwhile, AMD's Ryzen 7 3700X puts up a good show with comparable performance at stock settings, but at a higher price point. Focusing on the price-comparable Ryzen 5 3600X and 3600, the Core i5-10600K enjoys a comfortable lead in both average fps and smoothness metrics, and overclocking opens up a tremendous gap. Remember, you can also opt for the graphics-less KF model for $237 and get identical performance. 

Overall, the Core i5-10600K convincingly offers up the best gaming performance in this price range. We expect gamers to pair a chip of this class with less-expensive GPUs, but that extra bit of performance will be useful if you plan on upgrading to next-gen GPUs like AMD's Big Navi and Nvidia's Ampere.

Flipping over to the geometric mean of our application testing suite, we see that Intel has made great strides in eliminating the massive disparity between its Core i5 series and AMD's Ryzen 5 in multi-threaded applications. The overclocked Core i5-9600K falls far behind the stock 10600K, and overclocking pushes the 10600K far ahead of the 9600K and AMD Ryzen 5 chips alike. A quick look at the stock 9600K's standing against Ryzen 5, not to mention the 10600K, highlights just how much improvement Intel made by enabling Hyper-Threading and increasing clock rates. 

As expected, the Intel Core i5-10600K maintains Intel's performance lead across the broad spate of our single-threaded tests, which will manifest as snappy performance in a wide variety of tasks. 

Intel's Core i9-10900K is an impressive piece of silicon, cementing the company's overall lead in gaming performance. But its high power consumption requires expensive accommodations, particularly on the cooling side. And the chip doesn't leave much overclocking headroom for enthusiasts with conventional cooling setups. 

In contrast, the Core i5-10600K consumes an acceptable amount of power given its performance and 14nm++ process, and you can top it with affordable off-the-shelf coolers and receive the full performance. That's due in part to Intel's thicker copper IHS and thinner die, which also helps unlock more frequency headroom from the Core i5-10600K. That equates to strong gains in gaming, meaning overclocking is still alive and well in this price range. Our overclocking efforts were a breeze, and you could carve out respectable overclocks with a brawny air cooler, or shoot for the edge with 240mm or greater AIOs.

AMD's competing chips come with bundled coolers and a lower price of entry and are attractive for productivity-focused users, helped by overclocking support with all but the A-series motherboards, and you can opt for a previous-gen motherboard that doesn't support PCIe 4.0. Support for the faster PCIe 4.0 interface remains one of AMD's advantages over competing Intel processors, so you do lose part of the value proposition if you take that route. 

Many of the high-end Intel's Z490 motherboards are pricey, largely due to in-built support for PCIe 4.0 that you won't be able to use until the next generation of chips, and the LGA1200 platform looks to be short-lived. You can opt for lesser boards at the cost of sacrificing overclockability. 

Overall, Intel's Core i5-10600K offers better gaming performance than the Ryzen 5 processors and minimizes disparities in multi-threaded workloads, earning our Editor's Choice award.

MORE: Best CPUs

MORE: Intel and AMD Processor Hierarchy Comparisons

MORE: All CPUs Content

Intel dialed up the power for its new halo $488 Comet Lake-S Core i9-10900K, quite literally, to bulk up its mainstream desktop flagship with two extra cores for a total of ten cores and 20 threads. Aside from the promise of world-leading gaming performance, Intel also lowered gen-on-gen pricing to help fend off AMD's Ryzen processors that currently hold sway over our list of best gaming processors and dominate our CPU Benchmark Hierarchy. 

Once again, Intel aims its flagship chip squarely at enthusiasts and power users, touting the Core i9-10900K as the fastest gaming processor in the world. That performance comes mostly via the processor's extra cores and the fact that it basically comes pre-overclocked to an all-core 4.9 GHz, not to mention the peak 5.3 GHz speeds.  

Intel's Core i9-10900K still doesn't match AMD's halo 16-core 32-thread Ryzen 9 3950X in terms of threaded performance. Instead, the 10900K competes with the 12-core 24-thread Ryzen 9 3900X in terms of both performance and price, but Intel's chip has the highest power consumption we've seen recently on the mainstream desktop. Intel pushes the 10900K's TDP envelope up to 125W (a 30W gen-on-gen increase), but that's only a measure of base power consumption. Intel rates the processor for 250W at peak performance, and we even measured peaks as high as 325W at out-of-the-box settings. Naturally, that results in a lot of heat. 

Intel does have a few tricks to deal with the resulting heat and improve overclocking, like thinning the die and using a thicker heat spreader, but cooling is still a significant challenge. The 10900K's high power consumption even overwhelmed our 280mm watercooler during some tasks, so you'll need a brawny cooler to handle the increased heat output. 

Intel also added a few new features to its halo product to satiate enthusiasts, like per-core hyper-threading adjustments and reworked overclocking software, among other new tunable knobs. However, because the Core i9-10900K is basically an overclocked processor right out of the box, there is precious little overclocking headroom with conventional cooling.

However, provided you have adequate cooling, the Core i9-10900K is unquestionably the fastest gaming processor on the market. It also offers incredibly snappy performance in lightly-threaded apps, and solid performance in heavily-threaded workloads. Despite its insatiable power consumption and high heat output, the 10900K will be sure to find some uptake from performance enthusiasts that are prepared to splurge on a robust motherboard, power supply, and cooler, but most users would be better served with cheaper AMD alternatives, like the Ryzen 9 3900X. 

Intel Comet Lake-S Core Specifications and Pricing

The Comet Lake-S lineup shores up the Core family with more cores or threads at every price point that equate to deep price cuts, which we recently covered in depth. Intel reduced its gen-on-gen Core i9 pricing by 20% per core and cut pricing up to 50% per thread on other downstream models. For the downstream models, Intel finally enables hyper-threading on its Core i7, i5, and i3 lineups. That creates some surprisingly strong value plays in the mid-range, at least on paper. 

For the Core i9-10900K, Intel essentially stretched out its existing 14nm++ die and the Skylake architecture's ring bus interconnect. The integrated UHD Graphics 630 engine remains unchanged and is available on all models, but you also have the option to buy graphics-less F-series variants to save some cash (you can shave off $16 the Core i9-10900KF). 

Intel Core i9-10900K Specifications and Pricing

The real story comes in the form of higher clock rates, extra cores, more performance, and all of the power consumption pain that entails. Intel has brought support for its Turbo Boost Max 3.0 technology from the high end desktop to its mainstream platforms, and also roped in its Thermal Velocity Boost (TVB) mechanism from its mobile chips. This creates a multi-tiered boost implementation that can be incredibly confusing, so here is a short summation of each boost and what it entails:

• Turbo Boost 2.0: Increased frequency if chip operates below power, current, and temperature specifications.
• Turbo Boost Max 3.0: Fastest cores are identified during binning, then the Windows scheduler targets the fastest two active cores (favored cores) with lightly-threaded applications. Chip must be below power, current, and temperature specifications.
• Single-Core Thermal Velocity Boost: Fastest active favored core can boost higher than Turbo Boost Max 3.0 if below a pre-defined temperature threshold (70C) and all other factors adhere to TB 3.0 conditions. 
• All-Core Thermal Velocity Boost: Increases all-core frequency when all cores are active and the chip is under 70C. 

The Core i9-10900K's ten cores and 20 threads operate at a 3.7 GHz base frequency when all cores are loaded, but that can improve to 4.8 GHz via the normal Turbo Boost, and up to 4.9 GHz via Intel's Thermal Velocity Boost (TVB) feature. TVB is yet another layer to Intel's boost mechanism, but according to Intel, it only engages if the processor remains below 70C. However, we found that the all-core TVB frequency routinely engages regardless of chip temperature, which means motherboard vendors are still free to largely ignore Intel's power recommendations (functionality will vary by board). These vastly improved all-core boosts give the 10-core 10900K a fighting chance to match the 3900X's twelve cores in threaded applications, but as we'll see, that comes with a heavy power burden. 

TVB also applies to single-core boosts, too. Intel touts a 5.3 GHz single-core boost if the chip remains under 70C and meets all of the other normal requirements for boost activity. The chip downshifts to a dual-core 5.2 GHz via Turbo Boost Max 3.0 if the workload spreads across two cores, and then falls to 5.1 GHz if the workload lands in a single 'normal' core. Intel's goal is to provide the utmost in single-core performance, which remains its key advantage against AMD's Ryzen processors in lightly-threaded tasks like gaming and standard desktop PC applications.

All of this means that, given appropriate cooling and power delivery, you can hit 5.3 GHz in lightly-threaded work and up to 4.9 GHz in demanding all-core workloads. Intel assigned the 10900K a 125W TDP envelope, but that only applies when the chip operates at its base frequency - remember, Intel doesn't guarantee that you'll ever see Turbo Boost frequencies during normal operation. The chip is free to jump to 250W during boost activity, and motherboard vendors can ignore those limits.

Unlike the competing Ryzen 9 3900X, the Core i9-1900K doesn't come with a cooler, but at least the requisite motherboards are compatible with existing 115x coolers. Surprisingly, Intel only recommends a cooler that meets the PCG 2015D specification, meaning a cheap cooler that comes as either a 120mm AIO or a standard heat sink and fan, both of which can only dissipate 130W of waste heat. While this may technically meet Intel's specification for the base TDP rating, you would likely would see little to no Turbo Boost activity with a cooler of that class. 

We tested with the beefier Noctua NH-D15 and could mostly satisfy cooling requirements in standard desktop PC applications, but you will lose out on performance in workloads that push the boundaries with AVX instructions. As such, you'll need a greater-than-280mm AIO cooler or a custom loop to unlock the best of the 10900K. You'll also need an enthusiast-class motherboard with beefy power circuitry, and also plan on some form of active cooling for the motherboard's power delivery subsystem. 

Intel's Core i9-10900K comes with a few other tweaks, like improved memory support that jumps from DDR4-2666 to DDR4-2933, but still lags AMD's standard DDR4-3200 support. Intel also increased the L3 cache to 20MB. As before, Intel supports the PCIe 3.0 interface, while AMD has long since moved on to PCIe 4.0 and its doubled bandwidth. 

You'll need to upgrade to an LGA1200 motherboard for the new Comet Lake chips, and many Z490 models do support the PCIe 4.0 interface, albeit at the cost of increased pricing. However, that feature exists for as-yet-officially-unnamed future chips (read; Rocket Lake), so you'll pay extra for flagship motherboards that come with a feature you won't be able to use unless you upgrade your processor in the future.   

We also see generally higher pricing due to the bulky power delivery subsystems needed to feed Comet Lake with enough clean current. AMD's X570 platform is also pricey, but you can opt to use older, more cost-conscious X470 motherboards instead. You don't have the option of using older motherboards with Intel's lineup, and B-series and H-series boards haven't arrived yet, so you'll have to budget in a new board. We covered the Z490 motherboard ecosystem in more detail in our accompanying article.

Intel Core i9-10900K All-Core Turbo Boost and Power Consumption

With AMD nipping at its heels, Intel decided to basically overclock its Core i9-10900K right out of the box. The resulting power consumption measurements are surprising, even for us jaded folks that test the world's fastest and hottest x86 processors for a living. 

Intel uncorked its peak power consumption limits, know as PL2 (Power Level 2), by discarding its old limit of 1.25x of the base TDP rating (PL1 - Power Level 1). This exposes more performance while maintaining a 'controllable' amount of thermal output, but it also conjures up images of AMD's infamous FX-9590 processor – especially when we see the impact. 

Intel set its official new PL2 watermark at a 250W TDP, which is double the 125W PL1 rating, and recommends that motherboard makers keep boost activity limited to 56 second bursts (Tau). However, motherboard vendors are free to exceed all of these limits at stock settings if they feel their product can handle the increased voltages, current, and heat.

While Intel designed its 250W limit to keep thermals 'manageable' with a wide variety of cooling solutions, most motherboard vendors feed the chip up to ~330W of power at stock settings, leading to hideous power consumption metrics during AVX stress tests. Feeding 330W to a stock processor on a mainstream motherboard is a bit nuts, but it enables higher all-core frequencies for longer durations, provided the motherboard and power supply can feed the chip enough current, and your cooler can extract enough heat. 

To find the power limit associated with our chip paired with the Gigabyte Aorus Z490 Master motherboard, we ran a few Prime95 tests with AVX enabled (small FFT). During those tests, we recorded up to 332W of power consumption when paired with either the Corsair H115i 280mm AIO watercooler or a Noctua NH-D15S air cooler. Yes, that's with the processor configured at stock settings. For perspective, our 18-core Core i9-10980XE drew 'only' 256W during an identical Prime95 test. 

As you can see in our chart above, which measures the maximum frequency and temperature from any given core at a per-second granularity (100ms polling), temperatures (thick red line, right axis) quickly accelerated to 100C during the Prime95 test, and stayed there. Meanwhile, the frequencies (left axis) of the hottest cores dropped from the all-core 4.9 GHz down to 4.7 or 4.8 GHz as the processor engaged its throttling mechanisms to protect itself. Naturally, the air cooled configuration suffered more.

It's noteworthy that Intel positions its all-core 4.9 GHz frequency as a new Thermal Velocity Boost that only engages if the processor is below 70C, but as we can see here, the motherboard completely ignores that temperature limit. In fact, even though we're only showing the highest measurement from each of the eight cores, our logs indicate that individual cores still boosted to 4.9 GHz even when they were at 99C. At times, they also remained at 5.0 GHz when the core reached 100C, too, albeit for very short periods. That's right at the Intel-specified temperature limit. 

It's surprising to see the sheer amount of power consumed (and heat generated) by the Core i9-10900K, but it's important to keep perspective. There's a stark difference between the power consumed during stress testing applications that serve as a power virus, like the Prime95 test we ran above, and the power consumption you'll see during everyday use with the majority of 'normal' applications. Those workloads include stressful multi-threaded applications like Blender and AVX-powered HandBrake distributions. 

In fact, we often don't include Prime95 power measurements in our standard CPU reviews, largely because there is a massive disconnect between this extremely rigorous stress test and the power consumption and thermal load generated by most real-world applications, as we'll show below. 

Here we ran two series of tests to measure power consumption, thermal output, and clock rates during heavy real-world all-core loads. First, we kicked off ten multi-threaded Cinebench R20 benchmarks in rapid succession. In the first two charts, we can see both the Corsair H115i and the Noctua NH-D15 handle the workload with no throttling (the downward spikes occur during idle periods between runs), and temps hover in the 85C range when the processor is under full load. These workloads consume ~250W (right at Intel's recommended PL2 value) during operation. That's about 80W lower than the Prime95 stress test. The underlying R20 engine supports AVX, AVX2, and AVX-512, but this benchmark doesn't hit the chip quite like other AVX-equipped applications, so we moved on to more demanding work. 

We swapped over to our 'stress suite,' which is an extended string of heavier applications that stress the chip more fully. Again, we topped the chip with the H115i and NH-D15, but stepped through multiple instances of the corona ray-tracing benchmark, x265 HandBrake rendering tests, POV-Ray multi-threaded benchmarks, Cinebench R20 runs, and finally five iterations of the AVX-intensive threaded y-cruncher. 

Overall, these are among the more demanding threaded tests in our suite, and temperatures peaked at ~92C with the H115i, and ~85C with the Noctua air cooler. Surprisingly, because of its lower overall temperatures, the air cooler appears to perform better in this test. That would be quite the feat considering that we recorded peak power consumption at ~280W during the test run. 

However, we plotted power over the course of the run (slides five and six) and noticed that the chip drew more power when paired with the H115i cooler (~293W peak), particularly during the five y-cruncher tests at the end of the test. That's despite the fact that neither setup reached the 100C temperature limit. We also recorded better y-cruncher scores with the H115i cooler (~20% delta), but it was the lone test that suffered from performance degradation.

Aside from swapping the coolers, all other configurable parameters remained consistent for these tests., including clock rates. That clearly shows the importance of using an extremely strong cooler to unlock the full performance of the chip. We're not sure if this condition will present itself when we test with other motherboards, but we'll dive deeper in the coming days. 

In the meantime, we ran another extended y-cruncher session with the DH-15, this time calculating more digits per run (last slide). Again, the chip remained comfortably below the throttle point, indicating that an air cooler could be somewhat feasible for the chip, provided you select an expensive and rather bulky model. However, you could pay the price of reduced performance in some workloads. We did run our suite of application benchmarks and games with the air cooler, and y-cruncher was the only benchmark that suffered significantly reduced performance. We're following up with more testing, stay tuned. 

Intel Core i9-10900K Lightly-Threaded Turbo Boost and Thermals

Intel's new Thermal Velocity Boost brings the promise of single-core 5.3 GHz boosts, so we ran our standard frequency test for lightly-threaded workloads (methodology here). We recorded several spikes to 5.3 GHz during the tail end of the test during the VRMark benchmark. These boosts only occured when the processor was under 70C, so it appears that while the 70C limit is ignored for all-core workloads, the feature works correctly for single-core boosts. 

Intel Core i9-10900K Power Consumption and Efficiency

In a surprise to absolutely no one, the Core i9-10900K notched high power consumption measurements and poor efficiency metrics. 

As we can see, the stock 10900K draws nearly the same amount of power as the overclocked 9900K during the multi-threaded y-cruncher test, and exceeds its predecessor during the HandBrake workloads. We actually don't see too much of an increase in power due to overclocking, largely because the Core i9-10900K is already overclocked at its stock settings. 

Ryzen's stellar power consumption and efficiency, especially given their high core counts relative to competing Intel models, is impressive. That's simply the benefit of having a smaller, denser 7nm TSMC node. In terms of efficiency, it's clear that Intel's Core i9-10900K isn't nearly as efficient as any of the Ryzen 3000 series processors. 

Here we take a slightly different look at power consumption by calculating the cumulative amount of energy required to perform an x264 and x265 HandBrake workload, respectively. We plot this 'task energy' value in Kilojoules on the left side of the chart. 

These workloads are comprised of a fixed amount of work, so we can plot the task energy against the time required to finish the job (bottom axis), thus generating a really useful power chart. 

Bear in mind that faster compute times, and lower task energy requirements, are ideal. That means processors that fall the closest to the bottom left corner of the chart are best. As we can see, Intel's bevy of 14nm chips ring the upper right (ie, undesirable) corner of the chart while the AMD processors distance themselves from all but the Core i9-10980XE.

Pay close attention to the Core i9-10980XE, though. That's what a 14nm chip dialed into the efficiency sweet spot of the voltage/frequency curve looks like in operation, while the Core i9-10900K represents the exact opposite - performance at all costs, regardless of power consumption. 

Overall, the Ryzen processors are faster at this test and consume much less cumulative power to complete the tasks. That speaks to the efficiency of the 7nm process, and one can only imagine how much more efficient the chips would be if the central 12nm I/O die were fabbed on the 7nm process, too. 

MORE: Best CPUs

MORE: Intel and AMD Processor Hierarchy Comparisons

MORE: All CPUs Content

Multi-Core Enhancement (MCE) Testing

This tech goes by many different names, but most motherboards have a default Multi-Core Enhancement (MCE) feature that allows overclockable processors to run at their maximum Turbo Boost bin on all cores, at all times. This setting modifies the CPU's clock rate and voltage to deliver higher performance, which is basically factory-sanctioned overclocking. Performance, power consumption, and heat are all affected, naturally. Unfortunately, the mechanisms have to account for the lowest common denominator in terms of chip quality, and that often requires excessive amounts of voltage to attain higher clock speeds. As we can see in the image above, that can lead to thermal throttling. 

We manually disable this feature for our stock CPU testing to best reflect Intel's specifications, but we did run a few tests with the feature (Enhanced Multi-Core Performance in Gigabyte's parlance) on the Z490 Aorus Master. With the MCE feature enabled, the cores operated at 5.1 GHz and hit 100C during our HandBrake test, which triggered throttling. As we outlined on the previous page, this same stress test ran at 4.9 GHz with normal stock settings and didn't exceed 92C.

Our CPU's vCore hit 1.42V during these tests, which feels excessive, but we did a bit better with manual tuning.

Intel Core i9-10900K Overclocking Settings

Our tests reinforced our feeling that our 280mm AIO watercooler is wholly insufficient for any serious overclocking attempts, so we switched over to a custom 720mm loop that is, admittedly, extreme. 

As with all overclocking, the silicon lottery applies. But Intel is obviously already pushing the Core i9-10900K to its limits at stock settings, particularly with the 5.3 GHz boost frequency. Much like the situation we find with AMD's Ryzen 3000 series processors, our overclocking efforts found that we couldn't sustain an all-core overclock that matches the single-core boost watermark, and feedback from several motherboard vendors indicates you would have to be extremely lucky to find a chip that can sustain 5.3 GHz on all cores and still run normal applications. 

The 10900K did boot with an all-core 5.2 GHz overclock, though, but the requisite 1.4V vCore didn't prove stable unless we engaged aggressive load line calibration settings that pushed the chip over 1.5V during heavy workloads. That obviously wouldn't be good for processor longevity, so we settled with dialing in an all-core 5.1 GHz overclock paired with a 1.3V vCore. 

We experimented with the Z490 Aorus Master's Load Line Calibration settings, but found that the auto setting provided the most stability. After a tremendous amount of trial and error, we settled for the LLC-imposed 1.4 voltage for our tests. We also dialed up the ring bus multiplier to 48 to improve stability and performance, but this is an optional tweak. Finally, we dialed the memory to DDR4-3600 with 16-18-18-36 timings. 

Unfortunately our 5.1 GHz all-core overclock resulted in reduced performance in some lightly-threaded workloads, just like we see with Ryzen processors, but provided extra performance in threaded workloads. 

Comet Lake Architecture and Overclocking Features

Intel made significant changes to enable stuffing more transistors into the same package. As you can see in the picture of the die, which WikiChip predicts measures ~199mm2 (Intel no longer shares die measurements), the new 10-core die is essentially an extended version of the eight-core die found on the previous-gen Core i9-9900K (~174mm2). 

With up to ten cores now cooking away under the integrated heat spreader (IHS), Intel decided to further refine its approach by thinning the die for K-series CPUs. Intel reduced the processor die z-height (thickness) by 300 micro millimeters (from 800 to 500) to improve thermal transfer efficiency. Think of this technique as similar to lapping the die itself, thus shaving away a thin layer of silicon that rests between the heat-generating transistors and the Solder TIM. 

Intel also bulked up the copper IHS. The company says copper is three times more efficient at transferring heat than silicon (which can actually act as an insulator), so more copper is better than more silicon. The thicker IHS also assures adherence to existing socket z-height dimensions, thus enabling compatibility between LGA 115x coolers and the LGA 1200 interface.    

Instead of pTIM (polymer TIM - a.k.a. thermal grease), the Core i9-10900K comes with Solder TIM (STIM) that improves heat transfer between the die and heat spreader, thus lowering temperatures and often unlocking higher overclocks. Lower temperatures also help during stock operation by enabling longer (and more frequent) turbo boost durations and reduced cooling requirements. 

Now you can also enable or disable hyper-threading (HT) on a per-core basis. This granular control could be useful for any number of scenarios, with overclockers benefitting the most. Tuners could identify 'weaker' cores in the system during overclocking and then selectively disable HT on those cores. Paired with targeted per-core overclocking, that could be a promising feature. You trigger the HT feature via Intel's eXtreme Tuning Utility (XTU) or in the motherboard firmware, but it requires a reboot before threads are enabled/disabled. 

Intel has also enabled fine-grained voltage/frequency (VF) curve adjustments in the motherboard firmware, XTU software, and third-party applications, which allows you to adjust the various points in the curve. Some of these points on the curve were not previously available (including idle, which makes this feature appealing to undervolters). Intel won't describe the various points in detail because they are proprietary, but the company says you can see the changes in XTU. 

Intel has also bulked up its PCIe overclocking functionality by creating a system that allows motherboard vendors to connect an external clock generator to the PCH, thus bypassing the fixed 100 MHz clock. Intel says this enables speeds (up to) ~104 to ~108 MHz, thus boosting PCIe and DMI throughput slightly. Intel positions this feature for extreme/professional overclockers. 

MORE: Best CPUs

MORE: Intel and AMD Processor Hierarchy Comparisons

MORE: All CPUs Content

All test results marked with "PBO" reflect configurations tested with AMD's auto-overclocking Precision Boost Overdrive feature. We tested the stock Core i9-10900K with the Corsair H115i cooler for all of the following tests, but switched over to a custom watercooling loop for testing performance with a 5.1 GHz overclock. 

VRMark and 3DMark on Intel Core i9-10900K 

Synthetic gaming benchmarks often aren't generally indicative of real-world performance, but the 3DMark DX11 and DX12 tests are interesting because they measure the amount of raw computational horsepower exposed to the game engine. For now, most of today's game engines don't scale as linearly with additional compute resources, but these tests help us gauge how games could exploit processing resources as the engines become more sophisticated.

In the synthetic world of the Fire Strike benchmark, the ten-core Core i9-10900K can only match the 12-core Ryzen 9 3900X if we bump up the frequency to 5.1 GHz, but we see a limited gain from overclocking. In either case, the stock 10900K easily beats the overclocked previous-gen Core i9-9900K. That's impressive given that both processors command a $488 tray price. Meanwhile, the Ryzen 9 3950X takes a healthy lead over the field of Intel stock processors.

Flipping over to Time Spy finds the Core i9-10900K taking a lead over the Ryzen 9 3900X at 'stock' settings, and the 5.1 GHz overclock takes the top of the chart. 

VRMark test values per-core performance (a mixture of frequency and IPC), and it obviously prefers physical cores and lots of L3 cache. The $979 Core i9-10980XE's 4.8 GHz overclock goes a long way, but the 10900K's higher stock clock frequencies result in a resounding win at both stock and overclocked settings. 

Civilization VI AI and Stockfish on Intel Core i9-10900K

Civilization VI's AI engine values per-core performance, seemingly placing it firmly in the 10900K's crosshairs. The stock Core i9-10900K lands in the middle of the pack at stock settings, beaten by its HEDT and non-hyperthreaded counterparts, but tuning grants Intel's latest a big win over the competing chips. 

Stockfish, an open-source chess engine, is designed to extract the utmost performance from many-core chips by scaling well up to 512 cores. The chips stack up based on core count, so the Core i9-10900K is at a disadvantage in this test compared to the 16-core 32-thread Ryzen 9 3950X and 12-core 24-thread Ryzen 9 3900X. However, the Core i9-10900K notches a considerable win over its heavily overclocked predecessor, the 9900K.

Ashes of the Singularity: Escalation on Intel Core i9-10900K

Ashes of the Singularity: Escalation loves cores and threads, but clock rates also play a role. The stock Core i9-10900K effectively ties the overclocked Ryzen 9 3950X, but the AMD processor doesn't benefit much from the auto-overclocking Precision Boost Overdrive (PBO) feature in this benchmark. 

Civilization VI Graphics Test on Intel Core i9-10900K

Most gaming software is still optimized for low thread counts and high clock rates, and some game engines even prefer single-threaded cores. The Ryzen 9 3950X is plenty impressive in this test, but the 10900K is a brutal competitor with its high single-core boost frequencies. Once again, the 10900K beats all the competing stock processors before we applied our overclock, and tuning extends the lead further. 

Given that the 10900K can boost to 5.3 GHz under the right conditions, you might be surprised to see the 5.1 GHz all-core overclock take the lead once again. This could be due to erratic or less impactful boost activity, but the fixed frequency also reduces dynamic clock adjustments that can hamper performance. We've also tuned the memory and ring bus frequencies, too, which also yields advantages. 

Meanwhile, the Ryzen 9 3900X lags the 10900K by 5.3 fps, which isn't noticeable at these lowered resolutions, and it actually yields a better 99th-percentile frame rate measurement, denoting a smoother gaming experience.

Dawn of War III on Intel Core i9-10900K

The Warhammer 40,000 benchmark responds well to threading, but it's clear that clock speed and IPC also matter. In what is becoming a repetitive theme, the stock 10900K edges past the heavily-overclocked Core i9-9900K on its way to a lead over the rest of the test pool. The Ryzen 9 3900X lags by 14.6 fps in this title, which generally isn't favorable to Ryzen processors. 

Far Cry 5 on Intel Core i9-10900K

Far Cry 5 also tends to run well on Intel architectures, and once again the 10900K takes a convincing lead. 

Final Fantasy XV on Intel Core i9-10900K

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 with the higher-resolution setting. Final Fantasy XV finds the Ryzen 9 3900X scoring its lone tie with the stock Core i9-10900K.  

Grand Theft Auto V on Intel Core i9-10900K

The Intel processors continue to lead in nearly every title, but if we strip out the overclocked configurations, you'll notice the stock 10900K separates itself from the pack by a decent margin. That does come with quite a bit of power consumption and our fans cranking away at high speed, but the 10900K is undeniably a powerful gaming chip. 

Hitman 2 on Intel Core i9-10900K

The $374 Core i7-9700K reminds us that while the halo parts are beastly performers, they come at much higher pricing and require more exotic accommodations than the decidedly more mainstream i7 series. The -9700K currently tops our list of Best Gaming CPUs, and we expect the new Core i7-10700K, which comes with twice the threads for the same recommended pricing, will be just as impressive. 

Project Cars 2 on Intel Core i9-10900K

Although Project CARS 2 is purportedly optimized for threading, clock rates obviously affect this title's frame rates. 

World of Tanks enCore on Intel Core i9-10900K

The World of Tanks benchmark caps our test suite with yet another win for the Core i9-10900K over the other stock processors.

MORE: Best CPUs

MORE: Intel and AMD Processor Benchmark Hierarchy

MORE: All CPUs Content

Intel's Core i9-9900K is a hit with some professionals due to its high clock rates, so we ran an extra series of tests to measure the Core i9-10900K's performance in workstation-class workloads. Some of these applications also make an appearance in our standard test suite, but those test configurations and benchmarks are focused on a typical desktop-class environment. In contrast, these tests are configured to stress the systems with workstation-class workloads. 

We loaded down our test platforms with 64GB of DDR4 memory spread across four modules to accommodate the expanded memory capacity required for several of these workstation-focused tasks. We also outfitted the test systems with PCIe 4.0 SSDs to factor in the platform-level advantage of AMD's support for the faster interface.

Puget Systems Benchmarks

Puget Systems is a boutique vendor that caters to professional users with custom-designed systems targeted at specific workloads. The company has developed a series of acclaimed benchmarks for Adobe software, which you can find here.

Adobe After Effects CC Render Node Benchmark

The After Effects render node benchmark leverages the in-built aerender application that splits the render engine across multiple threads to maximize CPU and GPU performance. This test is memory-intensive, so RAM capacity and throughput are important and can be a limiting factor.

The Core i9-1080XE takes the overall lead in this test, followed closely by the Core i9-10900K. Notably, the auto-overclocked Ryzen 9 3900X effectively ties the stock 10900K. 

We encountered persistent application errors when testing the Ryzen 9 3950X in this benchmark (and this benchmark only), so it isn't listed in the results. We'll add the chip as soon as we can diagnose and fix the issue. 

Adobe Premier Pro CC Benchmark

This benchmark measures live playback and export performance with several codecs at 4K and 8K resolutions. It also incorporates 'Heavy GPU' and 'Heavy CPU' effects that stress the system beyond a typical workload. Storage throughput also heavily impacts the score. 

The PCIe 4.0 interface benefits the Ryzen processors, but sheer brute computational force appears to be the name of the game here: The overclocked Core i9-10980XE takes a comfortable lead despite its PCIe 3.0 interface. The Ryzen 9 3950X is very impressive in this series of tests, especially given its palatable price point and mainstream platform. Meanwhile, the 10900K matches the Ryzen 9 3900X at stock settings, but trails after overclocking.

Adobe Photoshop CC Benchmark

The Photoshop benchmark measures performance in a diverse range of tasks, measuring the amount of time taken to complete general tasks and apply filters. 

The stock Core i9-10900K is incredibly strong in this benchmark as it matches the overclocked Ryzen 9 3950X and it nearly matches the Core i9-10980XE after we dialed in the 5.1 GHz all-core overclock. GPU acceleration is here to stay in professional applications, and the 10900K's high clock rates help push it to the lead in the GPU score, contributing to its overall lead. 

SPECworkstation 3 Benchmarks

The SPECworkstation 3 benchmark suite is designed to measure workstation performance in professional applications. The full suite consists of more than 30 applications split among seven categories, but we've winnowed down the list to tests that largely focus specifically on CPU performance. We haven't submitted these benchmarks to the SPEC organization, so be aware these are not official benchmarks. We've upgraded to the new 3.0.4 revision that supports spanning the tests across processor groups and sockets.

Media and Entertainment

We run the new Blender Benchmark beta in our regular suite of tests on the following page, but different render jobs can stress processors in unique ways. Here we can see a breakout of several industry-standard benchmark renders that largely favor chips with hefty core counts. As such, it's no surprise to find the Ryzen 9 3900X,  Ryzen 9 3950X, and Core i9-10980XE lead in three of the renders, but the Core i9-10900K scraped out leads over the 3900X in the Island and 3BMWs workloads. 

The Ryzen 9 3900X also beats the 10900K in the HandBrake tests, but Intel's new chip turns the tables in the LuxRender benchmark. 

NAMD and Rodinia LifeSciences

NAMD is a parallel molecular dynamics code designed to scale well with additional compute resources, and here we can see performance scale well largely based on core and thread counts. That benefits the Ryzen 9 3900X, which unsurprisingly sweeps this round of tests. 

SPECworkstation 3's Rodinia LifeSciences benchmark steps through four tests that include medical imaging, particle movements in a 3D space, a thermal simulation, and image-enhancing programs. Again, the Ryzen 9 3900X's two extra cores and four extra threads help it sweep the 10900K. 

Product Development and Energy, Financial Workloads

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. Core counts reign supreme in this test, again granting the Ryzen 9 3900X a win over the 10900K. 

Calculix is based on the finite element method for three-dimensional structural computations. This benchmark performs well on the Intel processors, with the -10980XE typically taking the overall lead in our tests. The strength of the 10900K's high clock rates comes into play here, lending it the pole position among the stock processors and the overall lead after tuning. 

SRMP algorithms are used for discrete energy minimization. AMD processors have traditionally struggled with these tests, suggesting the benchmark might be latency-sensitive, but Ryzen 3000 reversed that trend in astonishing fashion. You'll notice the stock Ryzen 9 3950X is faster than its own overclocked configuration, but this is a repeatable phenomena. We also see the same trend with the Ryzen 9 3900X as it notches another win over the 10900K. 

MORE: Best CPUs

MORE: Intel and AMD Processor Hierarchy

MORE: All CPUs Content

Rendering on Intel Core i9-10900K

Ryzen processors are apex predators in the realm of threaded workloads, and despite the 10900K's additional two cores, the Ryzen 9 3900X continues that trend as it leads in all of the threaded tests. The 3900X also scrapes out a lead in the single-core Cinebench test, but the 10900K leads in the single-threaded POV-Ray benchmark. The stock 10900K is faster than the 5.1GHz overclocked configuration in both of these single-threaded tests, indicating the workload was correctly assigned to 5.3 GHz-capable core. 

That highlights that our all-core overclock, while faster in multi-threaded tasks, involves sacrificing single-threaded performance in other applications. We've seen this situation plenty of time with Ryzen processors in the past, particularly with manual overclocks, but it's certainly new to Intel's chips. 

Encoding on Intel Core i9-10900K

The single-threaded LAME and FLAC encoding tests lend themselves well to the strengths of the 10900K, which takes a pronounced lead in both tests. Again, the Core i9-10900K's stock configuration beats out its own overclocked configuration.

While most encoders are inherently single-threaded, the new Intel/Netflix-designed SVT-AV1 leverages multi-threading to great effect. As such, the Ryzen 9 3900X and 3950X take the lead over the 10900K.  

The threaded HandBrake x264 and x265 tests really speak to the AVX performance improvements AMD made to the Zen 2 architecture. Naturally, the chips with higher thread counts enjoy a lead here. 

Web Browsing on Intel Core i9-10900K

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

We sure hope you're not considering buying a fire-breathing Core i9-10900K just to browse the web, but this type of workload is universal. Single-threaded performance still reigns supreme in these tests, so the 10900K sweeps the field. Notably, the overclocked 10900K rig beats out the stock configuration, highlighting the fact that the CCPC2 drivers responsible for pinning threads into the fastest core aren't 100% accurate. 

Office and Productivity on Intel Core i9-10900K

Compilation, Compression, AVX on Intel Core i9-10900K

Once again leveraging its 12 cores and 24 threads to great effect, the Ryzen 9 3900X scores another win against the 10900K in the LLVM compilation benchmark. Meanwhile, the 7zip and Zlib compression/decompression benchmarks also rely heavily upon threading and work directly from system memory, thus avoiding the traditional storage bottleneck in these types of tasks. The Ryzen processors sweep the Intel competitors on the 7zip compression side of these tests, and nearly pull off the same feat in the 7zip decompression test.  

The heavily-threaded y-cruncher benchmark, which computes pi using the taxing AVX instruction set, reveals what we consider to be erroneous test results based upon our previous experience with Intel chips based on the never-ending Skylake architecture. We expect better scaling in this test given the 10900K's extra two cores, but the eight-core 9900K and 9700K beat Intel's new part at stock settings, and widen the gap after overclocking. We've scrutinized these test results heavily, especially to determine if the overclocked Core i9-10900K throttles during the workload, but we can't find signs of enough heat generation to trigger the throttling algorithms. Given this Intel launch follows the standard trend of late-arriving firmwares, it's possible that the problem resides with the motherboard. In either case, we're following up with Intel on the matter.   

MORE: Best CPUs

MORE: Intel and AMD Processor Hierarchy Comparisons

MORE: All CPUs Content

Spurred by AMD's unending Ryzen onslaught and its own inability to move forward to a smaller process node and more efficient architecture for the desktop, Intel went for broke with the Core i9-10900K to retain its gaming performance crown. 

In fact, the Core i9-10900K is exactly what we would expect from an overclocked 10-core 14nm Skylake derivative: Exceptional performance in gaming and lightly-threaded workloads, competitive performance in multi-threaded work, and downright ugly power consumption and thermal output. 

And that's pretty much what you get with the Core i9-10900K – an overclocked 14nm processor right out of the box. However, Intel did throw in what equates to a 20% price cut (provided that the processors land at retail anywhere near tray pricing). That makes the processor at least somewhat more attractive for extreme performance enthusiasts that are willing to put up with the less-desirable aspects of the chip in exchange for leading-edge gaming performance. 

In the chart below, we plot gaming performance using average frame rates and a geometric mean of the 99th percentile frame times (a good indicator of smoothness), which we then convert into a frames-per-second measurement. Bear in mind that we tested with an Nvidia GeForce RTX 2080 Ti at 1920x1080 to alleviate graphics-imposed bottlenecks. Differences between our test subjects shrink at higher resolutions.

The Core i9-10900K is the fastest gaming chip available, but not by much considering the cost and platform trade-offs. Intel's own $370 Core i7-9700K is plenty for most gamers. However, performance addicts with high refresh rate panels, particularly if they plan to update to next-gen GPUs like AMD's Big Navi and Nvidia's Ampere, would be able to expose larger performance deltas with the pending GPUs. Streamers and multi-taskers might also find the chip attractive, but the requisite supporting components will add a lot of additional cost.

Intel's previous-gen Core i9-9900K has enjoyed some uptake with semi-professionals due to its performance in single-threaded workloads, and the third chart in the album above shows that, based on a geomean of the listed single-threaded workloads in our test suite, the 10900K builds on that lead. Flipping over to the multi-threaded chart, we can see the 10900K's sizeable step forward over the 9900K in threaded workloads, but it still trails the Ryzen 9 3900X that retails for ~$50 less and comes with a stock cooler.

If you choose to go the extreme performance route with the 10900K, you'll need to invest in a capable watercooling solution and a premium motherboard to unlock the full performance. The intense power consumption, which topped 320W during stress tests at stock settings, will limit performance tremendously if not adequately cooled, not to mention potentially reduce processor longevity. That makes for a pricey build.

In contrast, the Ryzen 9 3900X comes with a bundled cooler and can be overclocked on less-expensive platforms, while Intel still restricts overclocking to pricey Z-series motherboards. Speaking of which, overclocking the Core i9-10900K is extremely limited with conventional cooling and can result in less performance in some lightly-threaded applications, making it a dubious pursuit.

If you are dead set on the best single-threaded and gaming performance available, regardless of power consumption, heat, and cost, the Core i9-10900K is your chip. However, despite the Core i9-10900K's excellent performance, the slim leads in several key areas won't be perceptible to most users.

Considering the excessive power consumption, heat generation, requirement for pricey supporting components, and lack of PCIe 4.0, most enthusiasts are better served with less exotic alternatives.

MORE: Best CPUs

MORE: Intel and AMD Processor Hierarchy Comparisons

MORE: All CPUs Content

The Ryzen 3 3300X and Ryzen 3 3100 show how much the AMD vs. Intel CPU landscape has changed since AMD's Zen first arrived three years ago. At the time, the mainstream desktop had been stuck at four cores for a decade, and Intel's market-leading $350 Core i7-7700K served as the halo part. Now AMD's new four-core, eight-thread Ryzen 3 3300X beats the 7700K in every benchmark across the board—but for only $120, making it a natural fit for our list of best CPUs and the best value gaming chip on the market. Our Ryzen 3 samples also reached the highest all-core frequencies we've seen with a Ryzen processor yet, underlining the untapped potential for enthusiasts. 

The company's new Ryzen 3 has Intel's Core i3 and i5 in the crosshairs. While the Zen 2 architecture and 7nm process have given AMD the core count lead in both the mainstream and HEDT segments, its previous low-end lineup isn't quite as impressive compared to Intel's existing Core i3 processors. That's a situation AMD hopes to rectify with its Ryzen 3 series as it braces for Intel's looming Comet Lake launch. 

The 3300X serves as Ryzen 3's new flagship part with four cores, eight threads, a 3.8 GHz base and 4.3 GHz boost, plus unified core design for a mere $120. AMD says this chip tackles Intel's $157 six-core/thread Core i5-9400F and provides 20% more performance, which would be quite the feat. 

The $99 Ryzen 3 3100 is the next step down the price ladder with four cores and eight threads and a 3.8 / 3.9 GHz base/boost frequency, but it's a steep step down on performance due to a different distribution of cores across the chiplet. In either case, AMD says this chip is 20% faster than the $120 quad-core/thread Core i3-9100 in gaming and has 75% more performance in threaded productivity work, making the 3100 a price-to-performance hero in its own right.

AMD's Ryzen 3 silicon is just the start of the value prop, though. The chips come with AMD's generally lax segmentation, so you get hyper-threading on both chips (which forced Intel to match with its upcoming processors), a bundled Wraith Stealth cooler (which could be better, honestly), and unlocked multipliers that enable overclocking on almost all AM4 chipsets (except A320). You will need a beefier cooler for serious overclocking with the 3300X, though. You also get access to the PCIe 4.0 interface, which unlocks twice the connectivity bandwidth of comparable Intel chips, and AMD has its upcoming price-conscious B550 motherboards that support the interface. Still, you'll have to wait until June for those. 

We're used to the claims of the world's fastest gaming chips for the halo desktop PC processors, but the Ryzen 3 3300X is unequivocally the best value we've seen for gamers on a budget, particularly if you're looking to build a sub-$1000 system. It also serves up plenty of performance for typical desktop PC work, too.

Ryzen 3 3300X and Ryzen 3 3100 Specifications and Pricing

Before we jump into the speeds and feeds, it's important to know that there's a big architectural difference between the Ryzen 3 3300X and the 3100, giving the former a big boost in gaming and other latency-sensitive workloads that extend beyond the incremental clock improvements we see on the spec sheet. We'll dive into that a bit further down the page, but the Ryzen 3 3300X came up as a far superior chip in our testing, particularly in gaming. In short, it's well worth spending the extra $20 on.

The new 65W Ryzen chips lower the bar for entry to Zen 2 from $199 with the Ryzen 5 3600 to $99 for the Ryzen 3 3100. The Ryzen 3 chips serve up plenty of performance from their four cores and eight threads, but you'll need to step up a tier to the six-core 12-thread Ryzen 5 3600 (or higher) if you're looking for intense performance in threaded productivity apps, multitasking, or game streaming/recording. 

But if you're focused on gaming with a discrete GPU and mainstream types of applications, the Ryzen 3 processors fit the bill with surprisingly high base clocks that rival Intel's finest Comet Lake chips. The adaptive Precision Boost 2 clocking mechanism also serves up the best performance possible given your other components, like the cooler, power supply, and motherboard. Both chips come with 16MB of L3 cache (more info below) and support for up to DDR4-3200 memory at stock settings. The same Ryzen 3000 dual-channel DDR4 memory controllers carry over, so memory overclocking is simple and a known quantity. AMD's Eco Mode, which unlocks better power/thermal efficiency by downshifting to a lower TDP tier (35W in this case), also looks promising with the 65W Ryzen 3 models.

If you're looking for a processor with integrated graphics, AMD has its previous-gen Zen+ based architecture paired with the RX Vega graphics engine in its Ryzen 5 3400G and Ryzen 3 3200G APUs. Still, their processing cores aren't nearly as fast as the Zen 2 cores in the new Ryzen 3 models. Instead, they focus on super low-end gaming. AMD's pending (though not yet officially announced) Ryzen 4000 APUs for the desktop will come with the Zen 2 architecture in the future.

We're on the cusp of Intel's Comet Lake launch, and early indicators are that Intel will prioritize its high-end processors, so the lower-end chips that fall into the same bracket as Ryzen 3 won't be immediately available. That leaves us with a bit of time before we can see the full impact of Intel's addition of higher clock frequencies and hyper-threading to the Core i3 lineup, in effect creating its own 'new' Core i7-7700K for the low end with its $154 Core i3-10320.

But as we often see, Intel's pricing sticks out here. You'll have to dip down further to the $122 Core i3-10100 to find the 3300X price equivalent, and it comes with a lower base frequency and less L3 cache. You'll have to drop down to Intel's Pentium lineup to find a sub-$100 chip that matches the Ryzen 3 3100, though the Core i5-9100F can be found on sale at times. And you'll need new boards for those Intel chips, with the new LGA 1200 socket, while AMD's offerings will drop into any number of existing AM4 motherboards.

Intel's Comet Lake chips do look competitive from a performance standpoint, but the company inexplicably discarded its overclockable Core i3 processor for this generation. As you'll see today, the overclockable Coffee Lake Refresh i3-9350K is a powerful gaming chip once you dial up the voltage and clocks, but the Ryzen 3 3300X is beastly in its own right: We achieved a 4.5 GHz overclock. That makes Intel's decision to get rid of an overclockable Core i3 model a possible Achilles heel.

Ryzen 3 3300X and 3100 Zen 2 Architecture

AMD's typical Zen 2 desktop architecture consists of up to two chiplets (CCD), each housing eight 7nm cores, that are tied to a 12nm central I/O Die (IOD) via the Infinity Fabric. AMD simply disables or enables cores based on the SKU, so this basic design comprises chips from four to 16 cores. This scalable approach also extends out to eight chiplets (64 cores) for data center chips, but AMD only uses a single CCD for the Ryzen 3 processors.

AMD splits each CCD into two four-core clusters called Core Complexes (CCX), and the Ryzen 3 3100 has two active cores present on each CCX (2+2). That means it's 16MB of L3 cache is split across two separate slices (2x8MB), which introduces latency, and inter-core latency is higher, too. That slows down latency-sensitive applications, like games and applications with heavy thread dependencies. 

In contrast, Ryzen 3 3300X's four cores are resident on a single CCX (4+0), so it has a single 16MB slice of L3 cache and inter-core latency is lower. As you'll see throughout our game suite, that speeds up performance tremendously. We'll also dive into a few cache and mesh latency measurements on the following page. 

This approach helps AMD in the die-binning process, with die with four fully-functional cores in a single CCX heading to the premium bin, while the lesser-quality die with defects spread across both CCXs can ship as value silicon. Instead of paring back the silicon, AMD uses the same IOD for these chips as it does with the regular dual-CCD models, and also uses the same CCDs. That means the new Ryzen 3 chips have a total of ~199mm2 of silicon with ~5.99 billion transistors.

While the clustered core approach lends the 3300X the upper hand in performance, it also creates thermal density issues when overclocking. That's because the four cores aren't spread as evenly across the full integrated heat spreader (IHS), creating difficulties dissipating the heat that's generated in a comparatively small area. 

AMD's new chips use Solder TIM to improve overclocking potential. As expected, we found that a capable liquid cooler can open up the heights of overclocking, but the 3300X's thermal density could hamper the maximum achievable frequency with some coolers. In either case, we reached 4.5 GHz, a record for our lab with ambient cooling and a Ryzen 3000-series processor (without disabling cores). However, we weren't able to accomplish any meaningful overclocking with AMD's Wraith Stealth cooler that will ship with the Ryzen 3 CPUs.

AMD B550 Motherboards for Ryzen 3 3300X and 3100

AMD's long-overdue and hotly-anticipated B550 chipset, which offers a lower price of entry for PCIe 4.0-capable motherboards, lands on June 16 with up to 60 models from the likes of MSI, Asus, Gigabyte, and the usual roster of other vendors. Pricing isn't confirmed yet, but early signs point to much better pricing than the existing X570 platform that's currently the only pathway to PCIe 4.0 support with a Ryzen 3000 chip. 

That leaves Ryzen 3 3300X and 3100 customers with a few options for now. You could drop the chips into the less-expensive B450 or X470 platforms and sacrifice PCIe 4.0 support in exchange for lower pricing, or you can purchase a pricey PCIe 4.0-capable X570 motherboard that isn't a good pairing for this class of chip. Otherwise, you'll have to wait for June to get the best of both worlds with the B550 chipset.

The Ryzen 3 3300X has 16 PCIe 4.0 lanes for graphics and an additional four PCIe 4.0 lanes for storage (M.2 SSD) direct to the CPU, with four more PCIe lanes linking the CPU with the chipset. With B550, AMD also threw in support for dual-GPU configurations on its mid-range motherboards, a first, and upgraded the six general-purpose PCIe 2.0 lanes that hang off the chipset to four lanes of PCIe 3.0. The chipset also now supports USB 3.2 Gen2. 

The B550 chipset brings plenty of advantages, including support for CPU and memory overclocking, but there’s a big trade-off. Due to BIOS ROM capacity limitations, B550 motherboards will only support current- and future-gen Ryzen processors. That means pre-Zen 2 processors aren't compatible, which includes the current-gen Ryzen 5 3400G and Ryzen 3 3200G APUs. 

This isn't an entirely surprising move. AMD has stuck to its pledge to support the AM4 socket until 2020 but has had compatibility difficulties (particularly with lower-end boards) as it expands its ever-growing roster of motherboards. That's led to varying levels of chip support on different motherboards, which can be confusing due to the wild variation in support matrices from different motherboard makers. Instead of supporting previous-gen chips on B550 platforms and disabling the functionality later, AMD chose to make a hard cut at the Ryzen 3000 series.

That leaves the B550 chipset without support for entry-level APUs, but the company strongly hints that the Ryzen 4000 "Renoir" desktop APUs will come to market by early next year.

Ryzen 3 3300X and 3100 Overclocking and Thermals

The Ryzen 3 3300X and 3100 were among the best-clocking Ryzen chips we've tested in our labs, largely because all-core overclocks easily exceeded the single-core boost frequencies.

We've encountered overclocking limitations on chips with six or eight cores enabled per compute die (CCD), but the Ryzen 3 3300X and 3100 come with a single CCD and only four cores enabled, which reduces heat generation. After a significant amount of tuning and validating our overclocks, we reached an incredibly impressive (at least compared to other Ryzen chips) 4.5 GHz overclock on all four of the 3300X's cores. 

Caveats apply, though. We're fairly confident this is a cherry chip, so your mileage may vary. We also paired the chips with a Corsair H115i cooler for our overclocking tests, and even with the fans cranking away at full speed, we pushed the very edge of safe temperatures (95C) with the 3300X's 4.5 GHz overclock. We tuned the cores to 4.5 GHz with a 1.35V vCore and left Load Line Calibration on auto settings. (Under load, we logged a 1.413V effective VID.) Memory overclocking was easy peasy via the XMP profile, which dialed the memory to DDR4-3600 with 16-16-16-36 timings.

Keeping the chip under the safe 95C ceiling wasn't easy, though. We recorded ~85C with all cores loaded in extended heavy non-AVX workloads. We ran the multi-threaded y-cruncher test for hours to assure stability because it has a heavy AVX distribution and generates the most heat, by far, of our real-world workloads. This test pushed our chip to the very edge of the safe zone with a peak measurement of 94C for a few brief seconds. We also tried to boot the chip at 4.6 GHz with a beefier 720mm custom loop, but to no avail. That means thermal constraints from external cooling aren't necessarily the limiting factor, at least with ambient cooling. In other words, the 7nm process is likely the limitation when paired with this core configuration.  

We stuck with our Corsair H115i for our overclocking test results. Bear in mind, this is a beefy 280mm water cooler that most people would argue isn't rational for this class of chip—and they'd be right. Our MSI MEG X570 Godlike motherboard is also stupidly well-appointed for this task, but we couldn't secure a B550 motherboard in time for testing. 

We're excited to get a Ryzen chip stable at 4.5 GHz and want to share the results, but view them as a showcase rather than something you can achieve with a cooler that doesn't cost more than the chip. Even then, you'd be pushing your new CPU to the ragged edge on thermals. 

You won't reach overclocks as high with lesser coolers, so take that into consideration. We did considerable overclocking with the stock cooler and couldn't sustain an all-core overclock that exceeds the 4.3 single-core boost without shooting above 100C almost immediately. If anything, you're best sticking with Precision Boost Overdrive with the relatively flimsy Wraith Stealth cooler, but there isn't much headroom. We'll look at how the bundled cooler impacts performance and thermals below, but a Hyper Evo 212-class cooler is probably a better entry point for cooling if you want to do any serious tuning. Just don't expect to hit a 4.5 GHz overclock with that class of cooler.   

The 3300X's clustered core arrangement made it much harder to cool, but as you'll see below, the 3100 tolerates the voltage much more readily. That's because the 3100's cores are spread out widely underneath the IHS and solder TIM, which reduces temps due to lower thermal density. However, the Zen 2 architecture is obviously much faster with the 3300X's clustered core design, so don't take the 3100's easier cooling as an indicator of more performance. With the Ryzen 3 3100 and a 1.4V vCore, we reached ~80C under heavy AVX loads at 4.4 GHz with the H115i. We'll do a bit more experimenting with overclocking with the stock cooler as time permits. 

Ryzen 3 3300X and 3100 Boost and Stock Cooling Performance

AMD didn't send a retail boxed chip for our testing, a first, so we had to dust off an old Wraith Stealth cooler (which luckily still had the original TIM) for our testing. We ran a series of tests to log the boost performance of the Ryzen 3 3300X and 3100, only plotting the minimum and maximum frequencies recorded every 100ms. That means these readings could come from any core. We also plot the temperature with the thick dark red line at the bottom (right axis).

With the Wraith Stealth cooler, the 3300X often exceeds its maximum 3.9 GHz single-core boost (which is only guaranteed on one physical core) by 50 MHz during our series of lightly-threaded tests (methodology here). That's a solid result for lightly-threaded work, but flipping over to the next slide shows the results with the Corsair H115i. The more robust cooling solution unlocks longer and more frequent boosts, so better cooling pays off. However, the performance delta in these kinds of workloads (LAME encoding, single-core POV Ray and Cinebench, PCMark10, and VRMark) only amounts to a few percentage points.

Keep in mind that we're testing with two polar opposites in terms of cooling: AMD's lightest stock cooler and a 280mm AIO. You won't need an H115i to hit those boosts more frequently—we expect most mid-range (maybe even low-end?) coolers will improve boosts considerably for the 3300X. Flipping over to the 3100, we see that chip easily maintains its 3.9 GHz boost, too. We also threw in a slide with the first-gen Ryzen 3 1300X as a comparison point. 

We hit the chip with the multi-core Cinebench R20 benchmark five times in a row to test the impact of better cooling on threaded workloads. The first chart, recorded with the stock cooler, logs a flat 4.1 GHz all-core frequency and peak temperatures of 95C on the stock cooler. Remember, AMD says 95C is the safe temperature limit for these chips, so the plateau at 95C represents the chip effectively hitting its thermal limit and throttling back performance to prevent damage.

Dropping the Corsair H115i on the test bench removes that limitation from the equation as temps plateau at 72C. That enables an extra 125 MHz for the all-core clock, or 4.225 GHz. That means the Ryzen 3 3300X's stock cooler is mostly sufficient, and it likely wouldn't take much more cooling capacity to boost it further. We recorded a 4% performance delta between the multi-threaded Cinebench tests with the stock cooler and AIO. AMD's beefier Wraith Spire cooler would be a better fit for the 3300X, but it isn't surprising to see the Stealth bundled with these budget chips. 

Ryzen 3 3300X and 3100 Power Consumption and Efficiency

The Ryzen 3 3100 proved to have exceptionally low power consumption during the AVX-laced x264 and x265 workloads, and that translates to excellent power efficiency  in the renders-per-day efficiency metric that takes performance into account. That's the advantage of having a denser node, but AMD has also tuned this chip a bit lower on the voltage/frequency curve than the Ryzen 3 3300X, which equates to better efficiency. 

The Core i3-9100 and Core i5-9400F also post relatively low power consumption figures during those tests, partially due to their lack of threading. After converting to an efficiency figure, the Core i5-9400F comes up with a solid score, but the 9100's low performance results in a low efficiency metric. The Ryzen 3 3300X is obviously tuned more for performance, so it lands closer to the 65W threshold. It should go without saying, but higher power consumption isn't a bad thing if you get more performance in exchange. The 3300X also offers great performance in these workloads, which is reflected in the efficiency scores. 

You'll notice the overclocked Ryzen 3 3300X draws more power than the tuned Ryzen 3 3300X, but that's because it required more voltage to achieve a sustainable overclock. As such, you should view the power figures for the overclocked processors as more of a general indicator – additional fine tuning could yield lower power consumption for either chip. 

Ryzen 3 3300X and 3100 Infinity Fabric and Cache Latency

As mentioned on the first page, the Ryzen 3300X and 3100 represent AMD's first 4-core Zen 2 processors, but they get there via different paths. The 3300X specifically is the first Zen 2 chip to only enable one CCX per CCD, while the 3100 is the first Zen 2 chip to drop down to two enabled cores per CCX. As we can see in the illustrations AMD shared during its architecture breakdown, the company has long planned to use its scalable architecture to step down to single-CCD processors that enable lower pricing. 

The pared-back design also offers performance advantages as it eliminates another layer of Infinity Fabric latency. Communication between one four-core CCX to another requires a hop back to the I/O Die, which creates another layer of latency. By enabling all four cores on one CCX only, that layer of latency is removed. We boiled that down to a few simple numbers in the table above. The 3300X's latency between cores (inter-core) weighs in at 24.8ns, but that's split into two distinct layers that measure ~11.4ns for near cores and ~27.8ns for far cores, respectively.

In contrast, the Ryzen 3 3100 has a 55.4ns inter-core latency with two distinct tiers that measure ~12.7ns for near cores and ~82.1ns for far cores, with the latter leading to slower performance in latency-sensitive apps, like gaming.    

AMD removed the memory controllers from the separate dies when it moved to the Zen 2 architecture, instead placing them on the central I/O Die (IOD). That results in similar memory latency and bandwidth measurements for both the Ryzen 3 3300X and 3100.  

Test Setup

MORE: Best CPUs

MORE: Intel & AMD Processor Hierarchy Comparisons

MORE: All CPUs Content

There are several interesting show-downs to keep an eye on as you flip through our performance charts. The Ryzen 5 1600AF represents a new value king that you can find for ~$85 (at times). AMD gave this first-gen chip a 12nm and Zen+ makeover long after it had come to market, and it's plenty impressive given its pricing—at least until now. We also have the Asia-exclusive Ryzen 5 3500X, which is AMD's only six-core six-thread Zen 2 processor. It doesn't support simultaneous-multithreading (SMT), which gives it a leg up in some games. 

We literally dusted off Intel's Core i7-7700K so it could make an appearance. AMD says the 3300X beats the 7700K in every test they've run—but we're overclocking it to 4.9 GHz (yeah, that's low, but this sample has always been less-than-great) to see how the two compare after overclocking. We also have the first-gen Ryzen 3 1300X to see how far AMD's graphics-less Ryzen 3 lineup has come in the last three years (AMD skipped an equivalent retail chip for its second-gen lineup).

And, of course, we have the Core i3-9100 squaring up with the Ryzen 3 3100, while the Ryzen 3 3300X does battle with the Core i5-9400F. Intel also chose to remove an overclockable Core i3 from its looming Comet Lake lineup, so we included the Core i3-9350KF to show what we'll be missing out on (aside from a $159 price tag for a Core i3). 

VRMark and 3DMark on Ryzen 3 3300X and 3100

Synthetic gaming benchmarks often aren't generally indicative of real-world performance, but the 3DMark DX11 and DX12 tests are interesting because they measure the amount of raw computational horsepower exposed to the game engine. For now, most of today's game engines don't scale as linearly with additional compute resources, but these tests help us gauge how games could exploit processing resources as the engines become more sophisticated.

In the synthetic world of the Fire Strike benchmark, the Ryzen 5 3600 and 1600AF's extra cores and threads come into play as the workload exploits their enhanced parallelism. The Ryzen 3 3300X sets the tone early with a substantial lead over its main competitor, the Core i5-9400F. Similarly, the Ryzen 3 3100 easily outpaces the Core i3-9100. The DX12 Timespy physics test finds the Core i5-9400F taking the lead over the Ryzen 3 3300X, but the Core i3-9100 can't compete with the Ryzen 3 3100. 

VRMark test values per-core performance (a mixture of frequency and IPC), and it obviously prefers physical cores and lots of L3 cache. Here the Ryzen 3 3300X leads all of the Intel processors in their stock configurations, but the six-core six-thread Ryzen 5 3500X takes a small lead due to its lack of threading.

Civilization VI AI and Stockfish on Ryzen 3 3300X and 3100

Civilization VI also values per-core performance, so it isn't surprising to see the overclocked 9350KF take the lead in this test of AI engine performance in a turn-based strategy game. The Ryzen 3 3300X and 3100 outpace their competitors, the Core i5-9400F and Core i3-9100, respectively. 

Stockfish, an open-source chess engine, is designed to extract the utmost performance from many-core chips by scaling well up to 512 cores, with latency being less of a factor. That plays to the Ryzen 5 3600's generous helping of cores and threads, so it takes a huge lead over the rest of the chips (which also happen to be less expensive). The overclocked Ryzen 3 3300X and 3100 both beat the Core i7-7700K. The new Ryzen 3 chips also serve up ultra-competitive performance at stock settings. Yeah, that's a $99 Ryzen 3 3100 beating the Core i7-7700K in this benchmark, if only by a minor amount.  

Ashes of the Singularity: Escalation on Ryzen 3 3300X and 3100

Ashes of the Singularity: Escalation loves cores and threads, but clock rates also play a role. The Ryzen 3 3300X leads the test pool even at stock settings, with only its 6C/12T Ryzen 5 3600 predecessor unseating it for the lead. The Core i7-7700K holds its own against all but the Ryzen 3 3300X and 3600, but it's noteworthy that its 4.9 GHz overclock can't beat the 3300X. That's nothing short of impressive and likely stems from the 3300X's 16MB of L3 cache. 

The Ryzen 3 3100 falls to the $85 (sometimes) Ryzen 5 1600AF, which isn't surprising given the latter's six cores and 12 threads. Put in that perspective, the fact that the stock 3100 can keep up with a mere four cores and eight threads is impressive, and turning up the clocks to 4.4 GHz propels it above the performant Core i3-9350KF. It's noteworthy that, even after overclocking, the Ryzen 3 3100 can't keep up with the stock Ryzen 3 3300X. That's a trend that will repeat itself through the rest of the tests.  

Civilization VI Graphics Test on Ryzen 3 3300X and 3100

Most gaming software is still optimized for low thread counts and high clock rates, and some game engines even prefer single-threaded cores. You can say all those things about Civilization VI, and the four-core four-thread Core i3-9350K fits the bill nicely. Regardless, it can't unseat the Ryzen 3 3300X in spite of its 600 MHz advantage after we kicked up the clocks. That's plenty impressive performance from the 3300X. It's noteworthy that the lone Ryzen chip without threading, the Asia-exclusive 3500X that you basically can't buy at a reasonable price, nearly matches the 3300X. 

Ryzen 3 3100 just edges past the Core i3-9100 at stock clocks, but Intel's strict overclocking segmentation keeps the 9100 from climbing up the charts like the 3100 does when we tuned it to 4.4 GHz. The Ryzen 5 1600AF also puts up a decent enough fight as it comes within striking range of the Ryzen 3 3100 and Core i3-9100, and tuning might unlock a bit more juice.

Dawn of War III on Ryzen 3 3300X and 3100

Five long years of stagnation on the Skylake microarchitecture and the 14nm process has punished Intel and enthusiasts alike as we've been subject to incremental clock speed, and thus performance, increases every year. The Ryzen 3 1300X, which debuted at $130 back in 2017, reminds us that innovation is still a thing. AMD skipped product-in-box (PIB) graphics-less Ryzen 3 models with its second-gen Zen+ chips, instead relegating them to the OEM market, but what a difference three years makes. Now you can get nearly double the performance of the 1300X (at least in this title) for a mere $120 with the Ryzen 3 3300X. 

The Warhammer 40,000 benchmark responds well to threading, but it's clear that clock speed and IPC also matter. All these things play well with the Ryzen 3 3300X and it's beefy L3 cache. At stock settings the 3300X leads the pool of other stock processors, and with our admittedly right-at-the-edge overclock, it also beats the overclocked 7700K at 4.9 GHz and the 9350KF at 5.1 GHz. Both of those chips don't come with bundled coolers capable of sustaining those overclocks, either.

Far Cry 5 on Ryzen 3 3300X and 3100

The Core i5-9400F pulls off a rare win over the stock Ryzen 3 3300X and the two overclocked Intel processors to take the overall lead in Far Cry 5. Meanwhile, the Core i3-9100 leads over the Ryzen 3 3100, indicating this title plays particularly well with Intel architectures. 

Final Fantasy XV on Ryzen 3 3300X and 3100

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 with the higher-resolution setting. The overclocked Core i3-9350KF takes the lead once again, but it falls to the Ryzen 3 3300X in out-of-the-box configurations. The Core i5-9400F and i7-7700K are closely matched in this series of gaming tests, indicating that Intel's response to competitive pressure, which comes in the form of higher clock rates and added cores, has steadily added more value in its mid-range processors. We expect competition to be even more heated for the mid-range when Comet Lake launches. 

Hitman 2 on Ryzen 3 3300X and 3100

The Core i7-7700K's four cores and eight threads assert themselves in this title after overclocking, but once again the Ryzen 3 3300X takes the lead over it and the Core i5-9400F in the stock configurations. Ryzen 5 3600 and the 3300X have traded a few shots throughout our tests, with the latter often serving up the lion's share of the 3600's performance, but at a lower price point. Tuning the Ryzen 5 3600 would open up a bit more of a lead, but the Ryzen 3 3300X is undoubtedly a great value. 

The Ryzen 3 3100 takes an impressive lead over both the Core i3-9100 and the Ryzen 5 1600AF in this title. The 1600AF has been a favorite for bargain-basement chip hunters, leading to inflated prices recently, but the Ryzen 3 3100 is a much better buy for gaming and modern connectivity options. 

Project Cars 2 on Ryzen 3 3300X and 3100

Although Project CARS 2 is purportedly optimized for threading, clock rates obviously affect this title's frame rates. The Ryzen 3300X is impressive here, beating out the rest of the field in the battle of stock configurations. Intel's Core i3-9350K takes the unequivocal lead after overclocking. Meanwhile the Ryzen 3 3100 trails the Core i3-9100 at stock settings.  

World of Tanks enCore on Ryzen 3 3300X and 3100

The Core i3-9350K can be devastating in some titles when it's pushed to its max clocks, and that trend plays out in the World of Tanks Encore benchmark. Meanwhile, the Ryzen 3 3300X trades blows with the Ryzen 5 3600, which is one of AMD's most popular chips. If you're primarily interested in gaming, the Ryzen 3 3300X provides enough performance to take the shine off the 3600, especially because it leaves more room in the budget for a better graphics card. 

MORE: Best CPUs

MORE: Intel & AMD Processor Benchmark Hierarchy

MORE: All CPUs Content

Rendering on Ryzen 3 3300X and 3100

It isn't surprising to find the Ryzen 3 3300X near the top of the threaded Cinebench rendering charts; Ryzen processors are apex predators in the realm of threaded workloads. It is surprising, however, to see the big delta between the stock Core i7-7700K and Ryzen processors in the single-threaded Cinebench tests. The 9350K also sports impressive per-core performance, but it also trails the stock 3300X. Surprisingly, that also carries over to the single-threaded POV-Ray test. 

The Ryzen 3 3300X also impresses in the Blender benchmark, but we also see another trend arise in the heavily-threaded workloads. In our gaming suite, the overclocked Ryzen 3 3100 trailed the stock 3300X in nearly all of the real-world benchmarks, but it provides similar performance to the stock 3300X after tuning in these workloads. Basically, inter-core latencies aren't a massive hindrance in 3D rendering performance, unlike in games.

Encoding on Ryzen 3 3300X and 3100

The single-threaded LAME and FLAC encoding test results highlight the Ryzen 3 3300X's agility in lightly-threaded apps, which is obviously helped along by the single cohesive L3 cache. Intel's chips assert their frequency advantage to take the lead, but only after overclocking. Meanwhile, the Core i3-9100 scores a win over the Ryzen 3 3100 in the FLAC encoding test. 

The threaded HandBrake x264 and x265 tests really speak to the AVX performance improvements AMD made to the Zen 2 architecture. Naturally, the chips with higher thread counts enjoy a lead here, but the new Ryzen 3 processors offer a surprising amount of performance from four cores and eight threads, especially in relation to comparatively priced Intel models. 

Web Browsing on Ryzen 3 3300X and 3100

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

Single-threaded performance still reigns supreme in these tests, though, and the Ryzen 3 3300X is plenty snappy. In its stock configuration, it takes the lead across the board over the other stock processors. The Ryzen 3 3100 isn't as impressive in this round of tests as the Core i3-9100 largely takes the lead, but it's hampered by its locked multiplier, granting the overclocked 3100 the lead.

Office and Productivity on Ryzen 3 3300X and 3100

The Ryzen 3 3300X impresses throughout this group of benchmarks, proving equally adept in both multi- and single-threaded workloads. It often vies with the Ryzen 5 3600 for the lead in threaded applications, showing that it can handle office and productivity applications exceedingly well for its price point. 

The stock Ryzen 3 3100 has to lean on its thread count advantage to scrape out its wins against the Core i3-9100, with the latter often leading in lightly-threaded applications.

Compression, Encryption, AVX on Ryzen 3 3300X and 3100

The 7zip and Zlib compression/decompression benchmarks rely heavily upon threading and work directly from system memory, thus avoiding the traditional storage bottleneck in these types of tasks. A copious helping of cores and threads is a real advantage to crunch through the massive influx of incoming data, so the Ryzen 5 3600 and 1600AF take the lead in the compression/decompression benchmarks. The Ryzen 3 3300X proves its mettle once again as it leads the other chips, even beating the overclocked 7700K in the 7zip compression and decompression workloads.

The heavily-threaded y-cruncher benchmark, which computes pi using the taxing AVX instruction set, finds the Ryzen 3 3300X notching another win over the Core i7-7700K in both stock and overclocked configurations. The Ryzen 3 3100 also grapples with the i5-9400F as it takes a big lead over the Core i3-9100. 

MORE: Best CPUs

MORE: Intel & AMD Processor Hierarchy Comparisons

MORE: All CPUs Content

If you're looking for potent gaming processors at an incredibly low price point, the Ryzen 3 3300X and 3100 are great options at their respective price points. But as you can see in the charts below, the Ryzen 3 3300X stands out as a gaming champion for budget buyers. 

In the chart below, we plot gaming performance using average frame rates and a geometric mean of the 99th percentile frame times (a good indicator of smoothness), which we then convert into a frames-per-second measurement. Bear in mind that we tested with an Nvidia GeForce RTX 2080 Ti at 1920x1080 to alleviate graphics-imposed bottlenecks. Differences between our test subjects shrink at higher resolutions, and most buyers at this price point will pair the chip with lesser graphics cards that push the bottleneck back to the GPU. 

At stock settings, the Ryzen 3 3300X offered nearly identical gaming performance (based on a geometric mean of our benchmark suite) to the Ryzen 5 3600, and actually squeaked past it in our 99th percentile measurements. If you're interested solely in gaming, the Ryzen 3 3300X takes the shine off of purchasing the Ryzen 5 3600, even though the 3600's twelve threads would come in handy for multi-tasking, streaming, and/or recording. We do have to keep in mind that the performance deltas are slim as we reach the top of the chart, so any of these chips could push a mid- or low-end graphics card along quite nicely. As usual, any extra performance matters if you plan on upgrading to the next-gen GPUs like AMD's Big Navi and Nvidia's Ampere. 

The Ryzen 3 3300X also exhibited a great balance of performance in both lightly- and heavily-threaded applications throughout our benchmark suite. If you don't frequently use heavy productivity applications, you shouldn't be left wanting. The Ryzen 3 3100 isn't nearly as agile because its single-thread performance holds back some applications that require snappy responsiveness, like web browsers and office applications.   

AMD didn't send our samples with a stock cooler, so we pushed a spare Wraith Stealth into service to measure the impact on performance at stock settings. As we demonstrated on the second page, cooling matters, so you'll need a better cooler to unlock the best of the Ryzen 3 3300X. 

The Wraith Stealth cooler is the lowest-end budget cooler for the Ryzen 3000 series, and it shows. We experienced high temperatures in extended workloads, and the fan was generally noisy as it ran at full speed to keep temperatures in check. But, to keep things in perspective, Intel doesn't provide a stock cooler at all with its overclockable i3 model, and bundled coolers at this price point are generally lackluster. It's a shame AMD didn't bundle the Ryzen 3 3300X with the Wraith Spire, as that would allow enthusiasts to crack open more of the overclocking potential. 

Speaking of which, our Ryzen 3 3300X and 3100 samples turned out to be quite the overclockers, but most enthusiasts won't pair these chips with coolers that cost more than the silicon, so you should keep your expectations in line with your cooling. The Ryzen 3 3300X offers most of its performance right out of the box, though; our comparatively extreme cooling solution only unlocked an extra 3.8 average fps and 2.5 99th-percentile fps. And that's with an RTX 2080 Ti, so that gain will shrink as you move to one of the more affordable best graphics cards. When it comes to the Ryzen 3 3300X, most gamers should just set it and forget it. Overclocking makes more sense if you use threaded applications frequently, though. 

The Ryzen 3 3100 offered a bit more uplift from overclocking as it gained 9.3 average fps and 6.3 99th-percentile fps, but it took a powerful cooler to get there. Considering the advantages to be had with the Ryzen 3 3300X, we would strongly recommend paying the extra $20 for the far superior chip. 

The industry has come a long way in a few short years. Now you can get what used to be flagship-class $350 performance for a mere $120 with the Ryzen 3 3300X. That's a value that's hard to beat, and we don't think Intel's looming Comet Lake will change the calculus much, especially in light of Core i3's slightly higher pricing. But we'll find out soon enough, and since most people looking at the Ryzen 3 3300X will want a B550 motherboard to go with it, waiting for our Comet Lake benchmarks makes sense.

The Ryzen 3 3300X certainly sets a new bar for low-end gaming processors. Suddenly, they don't seem so low-end after all. High-end chips like the Ryzen 9 3900X and Ryzen 9 3950X might make a bigger splash, delivering new tiers of performance, but budget shoppers will find a lot to like in AMD's new Ryzen 3 processors.

MORE: Best CPUs

MORE: Intel & AMD Processor Hierarchy Comparisons

MORE: All CPUs Content

German manufacturer Schenker has revealed the company's latest XMG Apex 15, as per a ComputerBase report. The system aims to be one of the best gaming laptops by offering desktop-type performance in a mobile format.

The XMG Apex 15 measures 14.2 x 10.2 x 1.3 inches (361 x 258 x 32.5mm) and weighs up to 5.7 pounds (2.6kg). The device flaunts a robust exterior, including a metal lid and a display frame, top and bottom case made of resin. 

The XMG Apex 15 utilizes AMD's B450 chipset to house the flagship AMD Ryzen 9 3950X desktop CPU. The processor brings nothing less than 16 CPU cores, 32 threads and a whopping 64MB of L3 cache. However, a few sacrifices had to be made to make the Ryzen 9 3950X more mobile-friendly. The XMG Apex 15 operates the Ryzen 9 3950X in its Eco-Mode, which forces to chip to stick to a 2.6 GHz base clock and 4.2 GHz boost clock to run within the 65W envelope, as opposed to its native 3.5 / 4.7 GHz. 

Schenker's gaming laptop features a 15.6-inch IPS panel at 1080p resolution and a 144 Hz refresh rate. To push the high-refresh rate display, the company allows you to choose between the an RTX 2060 or RTX 2070. These aren't the mobile or Max-Q variants either, so you'll get full desktop-level performance from either of the graphics cards.

The laptop comes with two SODIMM DDR4 RAM slots and can hold up to 64GB of memory total. 

There are three storage options. The machine has two M.2 2280 ports. One utilizes the PCIe 3.0 x4 interface, and the other uses PCIe 2.0 x4. There's also space for a 2.5-inch hard drive or SSD with a maximum height of 0.3 inch (7mm).

The XMG Apex 15 has one USB 2.0 Type-A port, two USB 3.2 Gen 2 Type-A ports, one USB 3.2 Gen 2 Type-C port, one headphone out and one microphone-in jack. Display outputs include one Mini DisplayPort 1.4 output with support for Nvidia G-Sync and one HDMI 2.0 port with HDCP 2.2. There's a Realtek Gigabit Ethernet port for internet connectivity, as well as Wi-Fi 6 and Bluetooth 5 functionality.

The laptop's other features consist of a backlit keyboard with a numeric keypad that offers 15 different color options, stereo speakers, Microsoft Precision Touchpad with two buttons, integrated microSD card reader, HD webcam that's compatible with Windows Hello and a fingerprint reader.

The XMG Apex 15 is equipped with a 62 Wh Li-ion battery. Being the powerhouse that it is, the gaming laptop draws power through a 230W power brick.

The base configuration, which features the RTX 2060, 8GB of DDR4-2666 Corsair Vengeance RAM and a 250GB  Samsung 860 Evo M.2 SSD starts at €1,337 (~$1,470). 

Whether you’re working from home, spending your socially-distanced downtime gaming (or like us, both) it’s a good time to have a high-performance desktop (HEDT) at your disposal. And Maingear is ready to deliver with its latest Rush tower.

Packing your choice of current high-end components (up to an AMD Threadripper 3990X or Intel Core i9-10998XE and dual-RTX Titan graphics cards), alongside an overhauled version of the company’s impressive Apex custom liquid cooling, the Rush is housed in a Lian Li 011 Dynamic XL case. It’s also ROG-certified, meaning Maingear worked with Asus to assure the system’s lighting works with Asus Aura Sync software for a unified light show.

Available starting today direct from Maingear, the Rush will start at $1,899, though expect to pay much more if you want high-end parts and that pretty, high-performing, solid-machined-acrylic Apex cooling solution.

Expect only Asus motherboard options with this ROG-certified system. As for other components, the company says a fully-outfitted Rush tower can fit up to four graphics cards (though with Nvidia SLI/AMD Crossfire effectively dead, gamers should stick to one high-end card), 128GB of RAM, multiple 2TB SSDs and your choice of high-end CPU/motherboard platforms, including large E-ATX options. 

If you’re wondering what’s changed with the Apex cooling setup since we looked at it in the Maingear F131 back in 2018, Maingear said it’s integrated a custom quiet pump, a flow-rate sensor and a high-capacity reservoir, along with parallel bridges to the GPU and radiator for greater cooling/overclocking potential. 

While it’s tough to validate the company’s claims that Apex is the “best performing” cooling solution on the market (especially from a press release), the clear acrylic design is  certainly the most visually striking custom cooling setup we’ve seen. If you’re looking for a desktop that delivers both high-end performance and the looks to match its muscle, the Rush might be tough to beat.

A maingear rep told us that the Rush won’t replace the company’s existing R1, Vybe, or F131 systems but will live alongside them as the go-to option for the “ultimate gaming PC.” With support for the top HEDT platforms and up to four GPUs alongside impressive custom cooling, the Rush is certainly a strong contender for that title. We look forward to putting it through its paces in a full review as soon as one lands on our test bench.

Nvidia today released the GeForce Game Ready 442.50 WHQL driver with improved support for The Division 2: Warlords of New York, ARK: Genesis and Apex Legends. The driver also brings support for three more G-Sync Compatible monitors.

Warlords of New York is a paid expansion for The Division 2 that introduces new areas, a higher level cap and multiple gameplay changes. Nvidia's latest Game Ready Driver is supposed to offer improved performance when the expansion debuts on March 3.

• The best gaming monitors: 144Hz, 4K and more
• Here are our picks for the best gaming laptops out now
• How to stop automatic driver updates AMD, Nvidia can now push through Windows

ARK: Genesis introduces "all-new mission-based gameplay" with "beautiful yet lethal new worlds filled with exotic creatures, craftable items and weapons unlike anything seen before on the ARK," Nvidia claimed. It's now available on all platforms.

Unlike the other games, the improvements for Apex Legends introduced with the latest Game Ready Driver weren't prompted by new content. Instead, Nvidia merely said the driver "includes a fix for the crash that has affected some players."

The driver also brought support for three new G-Sync Compatible gaming monitors: the AOC AG271FZ2, AOC AG271F1G2 and Asus PG43U. A full list of G-Sync Compatible monitors is available on Nvidia's website. 

You can download the GeForce Game Ready 442.50 WHQL driver via Nvidia's website or the GeForce Experience companion app. Additional details about the changes made with this driver are available in the full release notes (PDF).

Nvidia today released a new GeForce Game Ready 442.19 WHQL driver with support for the fourth season of Apex Legends, the Sam's Story expansion to Metro Exodus, Zombie Army: Dead War 4, as well as three additional G-Sync Compatible displays.

Those displays are the Dell AW2521HF, Asus VG259QM and LG 34GN850. Nvidia already said that the gaming monitors would be joining the G-Sync Compatible program; now it's made good on that promise by officially supporting them with G-Sync.

Nvidia said that it also improved the Max Frame Rate and Variable Rate Super Sampling (VRSS) features that the CES 2020 Game Ready Driver introduced in January. Max Frame Rate now supports values between 20 and 1,000 frames per second. VRSS, meanwhile, can now be easily activated in every game that supports it.

 Apex Legends Season 4 starts on February 4, which is the same day Zombie Army: Dead War 4 is set to debut. Metro Exodus: Sam's Story arrives on February 11. Now people with Nvidia graphics cards should be ready for those games at launch.

The GeForce Game Ready 442.19 WHQL driver is available for download via the GeForce Experience software and Nvidia's website. More information about the driver is in the full release notes (PDF), and Nvidia accepts feedback via its official forums.

Keyboard manufacturers seem convinced that people want more information at their fingertips -- literally. Das Keyboard made notifications via RGB lighting a central aspect of its Q devices (like the Das Keyboard Q5), and SteelSeries decided to do Das one better with the Apex 7, which features an OLED Smart Display in the top-right corner. This monochromatic panel is supposed to represent the future of keyboards. The problem is that the feature isn’t compelling enough to make up for the Apex 7’s shortcomings.

SteelSeries Apex 7 Specs

OLED Smart Display Disappoints

SteelSeries released the Apex 7 alongside the Apex Pro (as well as tenkeyless versions of both keyboards) earlier in 2019. Much of the company’s attention went to the Apex Pro, which features new OmniPoint switches that offer customizable actuation points, with the Apex 7 offering a similar design with traditional switches for people willing to sacrifice customizability for affordability.  But without those OmniPoint switches, the Apex 7 relies on the OLED Smart Display to stand out.

The display is a monochrome rectangle that displays a SteelSeries logo by default. It can be customized to display a 120 x 48 pixel image, too, for anyone who’d like to add a little bit of personal flair to the device. Otherwise, it’s primarily used to navigate menus involving the board’s RGB lighting, macro keys, and other settings, which makes it easier to customize the Apex 7 without having to install the companion software offered by SteelSeries.

SteelSeries offers several apps for the OLED Smart Display. Most use it to display additional information: One previews messages received via Discord, another displays in-game details about ongoing Counter-Strike: Global Offensive matches, and SteelSeries said one showing the currently playing Spotify song is “coming soon.” Those details could prove useful, but might not be enough to convince someone to Alt-Tab out of a game so they can view them on their monitor.

But the OLED Smart Display simply hasn’t made a difference in my day-to-day usage of the Apex 7 in the weeks I’ve spent with the keyboard. Its positioning means it’s unobtrusive, but it also makes the display hard to notice while the keyboard’s in a comfortable position. The number pad’s keys block at least half of the display even when the built-in legs underneath the Apex 7 are set to their lowest position. Steeper angles fare slightly -- but not significantly -- better.

I’ve had similar problems with other keyboards that were supposed to offer at-a-glance information. It’s an interesting concept, but it’s never managed to become core to my computer usage, despite repeated attempts to understand what the hype is about. I learned to touch-type specifically so I wouldn’t have to stare at my keyboard while I’m using my PC. Why break that habit--and ruin my posture--to look at the keyboard when there’s a monitor right in front of me?

Design

At 403 x 139 x 17mm (15.9 x 5.5 x 0.7 inches), the Apex 7 isn't a small keyboard. It's not a behemoth, either, but it takes up considerably more space on my desk than I would like. That's especially true when the wrist rest is attached--suddenly the number pad, thick USB cable, and wrist rest have become more than the sum of their parts. This led to some uncomfortable moments, which I'll discuss later, but it generally meant that once I put the Apex 7 down I was determined not to move it again. But the sizing issue is really a matter of preference--and price, since the Apex 7 TKL (sans the number pad) costs $30 less than its full-sized counterpart--in the end.

SteelSeries constructed the Apex 7’s case with an “aircraft grade aluminum alloy” that gives it a satisfying heft. I like the brushed metal--it’s a nice change of pace from the plastic cases I’ve encountered on most keyboards. Combined with the keycaps, which don’t rely on typographic monstrosities to convey the Apex 7’s status as a gaming keyboard, the frame makes it clear that the Apex 7 is supposed to be a premium device that wouldn’t look out of place in the office.

Like I mentioned earlier, the Apex 7 features two built-in legs that allow the back of the keyboard to be angled for more comfortable typing. There are only two settings--the legs are either tucked in or pulled out--which might disappoint anyone looking for more precise angling options, but plenty of keyboards employ a similarly limited setup. The included wrist rest can help make up for the limited adjustability, though, and its magnetic attachment makes it easy to type on the Apex 7 without worrying about it sliding around.

The Apex 7 connects to your PC through a USB cable that occupies two USB slots if you use the USB passthrough. I’m not particularly fond of this setup, largely because it can be hard to integrate into the hodgepodge of peripherals I already have connected to my system, but to each their own. A cable management tray helps keep things relatively neat, though, so long as you can ignore the thick cable protruding from the back of the keyboard.

Gaming Experience

SteelSeries offers three switch types for the Apex 7: linear Red switches, tactile Brown switches, and clicky Blue switches. All three feature 2mm actuation points and have been rated to withstand 50 million keypresses. My review unit featured the Red switches, which are often marketed to gamers because they’re supposed to be a bit easier to press than their counterparts, and quieter, too, so as not to annoy everyone in voice chat with a bunch of clacking sounds.

I didn’t mind the switches themselves, even though I usually prefer tactile switches. The switches were a bit louder than the Romer-G switches found on my Logitech G Pro, but not as loud as other keyboards I’ve used. Unfortunately there did appear to be some problems with key presses not registering during particularly hectic moments in games like Overwatch when I had to quickly switch between keys.

The Apex 7 offers N-key rollover, so those moments should not have posed any problems. But I noticed that inputs were regularly ignored during frantic moments--which is exactly when gamers expect their keyboards to perform flawlessly. A quick example of a common scenario: grappling into the air as Wrecking Ball, using the WASD keys to reposition, pressing Q to drop a bunch of mines, pressing Ctrl to slam into the opposing team, and then pressing E to activate shields.

That paragraph might read like a bunch of gobbledygook to people who don’t play Overwatch, but the gist is that I’m doing a bunch of things at once to make sure my mech-piloting hamster doesn’t turn into so much roadkill. Yet I would regularly find that my bombs didn’t drop (so the Q didn’t register) or shields didn’t activate (which means the E didn’t) while using the Apex 7. I didn’t have any problems performing similar maneuvers on the Logitech G Pro after multiple attempts.

Some errors are to be expected when playing an online game. But with everything else about my setup being stable--the Apex 7 is literally the only thing I changed--the problem is most likely the keyboard’s fault. If the OLED Smart Display’s relative uselessness makes it hard to justify the Apex 7’s price, missing inputs makes it hard to justify using the keyboard while playing competitive games, period. That makes it hard to recommend to most gamers these days.

Typing Experience

That said, these problems didn’t affect the Apex 7 during everyday use. I wrote most of this review on the Apex 7 and didn’t have any more typos than I usually do. That might have resulted from me being particularly cautious, though, or actively gauging how the keys felt while I was writing just to make sure my initial impressions weren’t clouding my final judgment. Sometimes it takes a little while to get used to a keyboard. So I turned to the Aesop’s Fable test at TypingTest.com.

The results confirmed what I'd already suspected about the Apex 7, which was that I was typing faster but also making more mistakes. TypingTest scored me at 101 words per minute (wpm) after taking 10 typing errors into account. I usually score around 103 wpm on the Logitech G Pro and Microsoft Sculpt Keyboard, with fewer errors on each, so I'm clearly hitting keys faster on the Apex 7. I just also happen to be hitting more keys by accident in the process.

Having the included wrist rest made typing on the Apex 7 more comfortable as well--at least until I had to reach for my mouse. Having to accommodate the number pad after getting used to using tenkeyless keyboards meant I was putting more stress on my shoulders by spreading them further apart. I mostly noticed this while I was gaming, so on the whole the typing experience was fine, but anyone who doesn't need the number pad should consider the TKL model of the Apex 7 rather than the full-sized keyboard.

RGB Lighting and Controls

The Apex 7’s RGB lighting is controlled with the same SteelSeries Engine 3 software as the Apex Pro. I’m not as keen on RGB lighting as some of our other writers and editors--I’m content to let all the pretty lights cycle through colors of the rainbow in the default “Coast” setting. I could use SteelSeries Engine 3 to keep the RGB lighting in sync for my headset and mouse, both of which are also from SteelSeries, but I can’t see the lights on my ears or under my palm so I left ‘em alone. (Yes, I’m aware that this makes me the outlier in an age when RGB lighting is finding its way into almost everything, but that’s just fine by me.)

For anyone who’s more interested in customizing their setup than I: SteelSeries Engine 3 makes it easy to create individual profiles that automatically start in certain programs and games. There isn’t as much customizability with the Apex 7 as there is the Apex Pro, thanks to the missing OmniPoint switches, but different lighting profiles can help set the mood based on what’s on-screen. The software also simplifies the assignment and function of macro keys. Built-in color profiles also make it easy to choose a new lighting scheme with as little fiddling as possible.

My relative disinterest in advanced RGB lighting controls somewhat ironically helps me justify the OLED Smart Display. The built-in display can be used to cycle through five built-in lighting schemes (Vapors, Coast, Haze, Prism and Orange) or navigate a Custom setting. Is this as convenient as using SteelSeries Engine 3? No. The options are limited and the menus aren’t easy to navigate. For anyone who only has a passing interest in RGB lighting, though, not having to install additional software could be the best justification for the OLED Smart Display. Gamers who frequently move their keyboards between multiple PC will also appreciate this.

Bottom Line

It’s not hard to see why SteelSeries made the Apex 7. We liked the Apex Pro, but most people aren’t willing to spend $200 on a keyboard, regardless of what kind of switch it uses. Relying on its existing switch options gives SteelSeries an easy way to take advantage of all the design work that went into the Apex Pro--including the OLED Smart Display--but with a lower price meant to help those design elements reach more people. Other companies have done similar things.

But we concluded our Apex Pro review by saying that the OmniPoint switches and their per-key customization options helped make up for the keyboard costing more than competitive products that were actually more comfortable to use. Without the advantage of those OmniPoint switches, the only thing the Apex 7 really had going for it was the OLED Smart Display, and that simply isn’t a compelling enough gimmick to make up for the product’s other shortcomings.

There’s no denying that the Apex 7 is an attractive keyboard--it’s certainly easier on the eyes than most “gaming keyboards” on the market--but I don’t think that’s worth the $160 asking price. And if the problem I had with missed inputs affects more than just our review unit, then this keyboard is even more problematic.

In short, the Apex 7 is a middle-of-the-line keyboard with a relatively high price that only seems low because its higher-end sibling costs even more. Most people would be better served by competing products that either have more substantive features, or ditch them in pursuit of a lower price.

Image Credits: Tom's Hardware

MORE: How We Test Mechanical Keyboards

MORE: Mechanical Keyboard Switch Testing Explained

MORE: All Keyboard Content

If you want a mechanical gaming keyboard you can finetune to be truly “you,” the SteelSeries Apex Pro ($199 as tested, $179.99 for the tenkeyless version) is it.Anyone can appreciate its striking design and attractive RGB lighting. But hardcore gamers and finicky typers will appreciate unprecedented customization brought by unique switches that allow you to individually set the actuation point (how far you have to press for a key to register) for nearly every key. A screen and other bonuses solidify and rationalize the keyboard’s high price.

But the Apex Pro’s linear keys still feel mushy, regardless of the actuation point. If you’re more of a typer than a gamer, you’ll want to consider other (and likely cheaper) gaming keyboards.

SteelSeries Apex Pro Specs

The Apex Pro can store 5 custom profiles and uses a 32-bit Arm processor. The keys are guaranteed for 100 million presses, which is twice the average of typical mechanical gaming keyboards, thanks to SteelSeries’ OmniPoint switch (more on that in the section below).

The keyboard also supports 104-key rollover and 100% anti-ghosting, meaning you can press 104 keys at once and they’ll all register. While I hardly have any reason (or the physical ability) to strike all 104 keys at a time, simply smashing of as many keys as possible on an online n-key rollover test, proved that it can register at least 32 keys simultaneously. 

Design

Before getting into the customizable actuation, I can say that the Apex Pro is an impressive keyboard from looks alone. It’s built on a Series 5000 aluminum alloy frame which SteelSeries says is aircraft-grade. The matte finish almost feels like plastic, and the dark build makes a great foreground for the RGB lighting.

As expected for a mechanical keyboard, the Apex Pro connects to your PC via the attached USB Type-A cable. More surprising is the cable management system. The back of the keyboard has a long channel for snapping in the keyboard’s cable and extending it outwards from one of three locations. Whether your PC’s port is to the right, left or center of the keyboard, you can still keep things tidy. Although, with the cable so long (about 7 feet), I wish there was even more space. The bottom also has 3 rubber pads that keep the keyboard in place atop your desk.

The flip-out feet add a small amount of height at the back, though some might wish for a greater boost. But what it lacks in height, it makes up for with a soft-touch magnetic wrist rest, which snaps on and off in a second. The deliciously smooth wrist rest feels amazing, especially when sliding across the keyboard, and helps deliver the type of experience you should demand from a $200 keyboard. It’s not as thick and plush as a fabric one, like the HyperX wrist rest, but it should be durable and long lasting without having fabric to wear down, and its 8 rubber feet add even more stability.

The keyboard’s screen is OLED, but with its tiny size and black-and-white display, OLED is all about readability. What makes it special is the customization it enables. You activate its menu by double-tapping the button to its right that brings up a menu (Illumination / Macros / Actuation / Profiles and Settings) which you can navigate with the audio wheel and press in. It’s nice that you can navigate these things without downloading the free Steelseries Engine 3 companion software -- especially if you’re going to use the keyboard on multiple computers. Of course, the slick audio wheel also controls your PC’s volume.

Up top, next to a subtle SteelSeries logo, is a USB 2.0 Type-A passthrough port, like many mechanical keyboards have. To use it, plug one of the keyboard cable’s two-pronged USB connectors into your PC. This virtually takes one USB port from your PC, which may be pushed against a wall or trapped by cables and moves it to a more accessible spot--perfect for plugging in your mouse or a Bluetooth dongle. The USB passthrough port shines bright white when it’s ready for use.

At 2.14 pounds and 17.2 x 1.9 x 4.4 inches, the Apex Pro is smaller than other 104-key keyboards, such as the Hexgears Impulse (2 pounds, 17.25 x 6 inches) or Patriot Memory Viper V765 (2.5 pounds, 18.4 x 6.4 inches). If desk space is extremely tight, this may be worth considering, but of course a tenkeyless keyboard will take up much less space.

OmniPoint Switches: Adjustable Actuation

All of the Apex Pro’s keys have a linear feel, with the keys traveling straight down and lacking any tactile sensation or bump until you hit the bottom, much like Cherry MX Red switches. The majority of the keys use OmniPoint switches for adjustable actuation points, but the numpad, macro and arrow keys use traditional MX Red switches.

To let users set varying actuation points, SteelSeries invented OmniPoint switches that feel like familiar Red mechanical switches, but work in a totally different way. Your press actuates normal mechanical switches by making two pieces of metal touch. However, when you press an OmniPoint switch, you’re pushing a magnet down its stem and closer to a sensor. Using the “Hall Effect,” or magnetic fields, the keyboard can tell the distance between the sensor at the base and the magnet’s stem, so you can tell it which distance equals a key press. Without any physical contact causing degradation, these switches promise longer durability.

With the SteelSeries software or the menu in the OLED screen, you determine at which distance a press is registered --10 steps from 0.4-3.6mm -- for each of the 61 main keys. At 0.4mm, you barely have to press a button for it to register. For comparison, note that Blue and Green-style switches actuate at 2mm, while Kailh Box White actuates at 1.8mm. At 0.4mm, there’s opportunity for special uses, such as speed-driven first-person shooters, while lower actuation points should help with accuracy.

Unfortunately, both the software and OLED screen let you set actuation on a 1-10 scale instead of using actual measurements. But through the screen, you can test it out in real time, as a bar fills up as you press to show you far the key has traveled and when it actuates.

As mentioned, not every key has an OmniPoint switch. When SteelSeries first teased the keyboard in January, they told us that this is because the OmniPoint switches are expensive to make. Considering how expensive the peripheral already is, omitting the keys which are least-likely to be used in games is a smart move.

The Omnipoint switches are a bit different from those used in the Cooler Master MasterKeys MK850 and Wooting One analog keyboards, which work like an analog stick on a console controller and read how much pressure you’re applying. The Cooler Master’s Q, W, E, R, A, S D and F keys can sense a press across 4mm of travel but need calibration every time the pressure sensitive tech is reactivated. Like the Wooting One, its analog keys leverage optical sensors, which use a sensor and light to detect the distance between each press, but only the Wooting One lets you set your actuation point. Similarly, Razer’s Opto-Mechnical switches, used in the Razer Huntsman Optical, tap optical sensor tech, but you can’t customize the actuation point and they’re not pressure-sensitive.

Typing Experience

With the Omnipoint switches’ linear feel, pressing a key is smooth, with the keys going straight down without any bump or click and rising up with your finger as it lifts. You’ll hear a clacking noise as you type, and if you’re extra aggressive, you may notice the echoey ding of the aluminum frame. It’s much louder than the Razer Cynosa Chroma Pro I’m currently using with a plastic build and “mecha-membrane” switch (3.5mm of travel that feels a bit short). If you prefer tactile switches, this keyboard will feel spongy to you. Without a tactile sensation or much resistance, my fingers felt more tired, sort of like they were typing through water, particularly during long sessions.

Because of the adjustable switches, you get newfound control over the typing experience. Do you want extra sensitive keys? Or do you want the keyboard to demand firm, confident typing to help you avoid accidental presses and typos?

I quickly noticed the difference when changing the overall actuation. While a light touch--I’d call it more of a tap than a press--a press registered if actuation was set to 1. At 10, I needed a truer press and could see letters taking longer to appear on my word processor than with actuation set to 1.

To see how this affected my typing accuracy I tried the 10fastfingers with all OmniPoint switches set to the lowest, highest and middle actuation points.

I hit quicker speeds with actuation set to 1 (116 wpm)  versus 10 (110 wpm) but the same speed with it set to 5. Ironically, I got my best accuracy with actuation set to 1 (99.48%) and lowest at 10 (96.33%), but the difference is small. In general, my accuracy rate is usually around 98%, and I don’t usually accidentally hit keys, which the high actuation point would theoretically help me prevent. For me, accuracy was hurt by adjusting to the amount of space between each of the keys and perhaps, based on the numbers, assuming a key was actuated earlier than it truly was.

Gaming Experience

Compatible with PC, Mac, PlayStation 4 and Xbox One, the Apex Pro’s adjustable actuation shines when gaming, where you can strategically program actuations for different action. The general experience is responsive and comfortable, especially with the wrist rest.

When I switched from a middle actuation setting to the highest for Middle-earth: Shadow of War no longer was I accidentally hitting a button that made me wander to the left and wonder what happened. With less sensitive keys,I had to deliberately press a key to make it happen. Playing Battlefield V and Grand Theft Auto V at the most sensitive setting, the keyboard was reliable and quick. However, I didn’t notice automatically become a better gamer: Hardcore gamers who need a lot of repetiive presses, like in a MOBA game, may disagree.

RGB Lighting and Controls

The RGB lighting against the Apex Pro’s midnight gray frame not only pops, but brilliantly illuminates the chassis. The lights burst from beneath the raised keycaps, emanating a can’t-miss glow across the keyboard. Even from a few feet away, the RGB makes a beckoning call, although if you don’t like that much attention, you can shut the RGB off or turn down the brightness.

But if you’re an RGB glutton like me, you’ll love the configurability. You get the most customization options by downloading SteelSeries Engine 3 software for free. But if you don’t want to add more software to your PC, you can toggle through pre-set profiles with the OLED screen. The display will show huge letters as you try five different designs: Vapors, Coast, Haze, Prism and Orange.

Unlike other RGB keyboards, these aren’t just a flashing rainbow or breathing between two colors. These profiles are advanced. For example, Vapors has the rows transition between a rave-ready theme of indigo, purple, fuchsia, pink and teal that sends a wave of white across the row of each button pressed. Coast is similar, with colors going from blue to yellow, orange and then red as you look from left to the right, and with each press, the buttons turn blue before fading back to their original hue. Even without the software, you can enjoy mesmerizing RGB lights.

Also in the OLED screen’s Illumination menu is a Custom section. The menu options here let you make the lights all one color with a white ripple effect with each keypress, breathe a color with a white ripple effect, breathe in 1 of 2 multi-color schemes, or have the lights off. The custom menu also has a Reactive section, for picking from 4 different reactions to keypresses (including none). The Custom menu’s AFK (away from keyboard) section controls what happens when you’re away from the keyboard (such as turning the lights off). Meanwhile, Timeout ranges from 5 seconds to 2 hours / never).

Navigating all this without software can be challenging. Besides having to get used to the clicking in the wheel and pressing the OLED button to enter and exit menus and submenus, without software it seems turning off the reactive mode or selecting an AFK mode replaces any of the preset options you already have. But you can still get away with not using the software--so long as you don’t mind having a reactive affect with your RGB.

Those who do download the software will be greeted with a very robust system.

You can customize the 5 on-board profiles, including the effects and its speed and wave, the reactive layer and when AFK effects start taking place. As with setting actuation, the software has tools to select each key individually, select all groups with the same setting, select all keys and undo. You’re also free to create a whole new config from scratch but won’t be able to activate it with the keyboard’s screen (but you could just replace the preset profiles with your own).

You can have a profile autolaunch with a certain app or game. Perhaps you want a toned down, non-reactive RGB with Microsoft Word and rainbows and white ripples everywhere with Counter-Strike: Global Offensive (CS:GO). Remember, your profiles also store actuation, so that Microsoft Word-ready profile might also have higher actuation for help with accuracy, while your gaming one has low actuation points for speedy movements.

Programming macro keys is a similar situation, with a low level of control possible via the OLED screen and much more available in the software, The OLED screen lets you record and delete macro keys (the screens show you the key combination to enter to stop recording). With the software, all keys are available for setting keybindings, and there’s Macro Editor that pops up in its own window.

OLED Screen

In addition to letting you quickly change you RGB lighting, macro keys, overall actuation point, the OLED screen also holds a Profiles and Settings menu (timeout, brightness and lock indicator, which’ll display icon(s) on the screen if in num/caps lock, and about section.

The screen can also share info with you if you play CS:GO (see things like your current KDA and money), Discord and Spotify. But don’t think you’re getting a second display like you do with a dual-screen laptop. For example, the screen won’t tell you what song/ artist you’re playing on Spotify but can pause, play and change the song.

When not showing off a menu, the Apex Pro’s screen can depict your favorite 120 x 48 pixel image. If you don’t have one, SteelSeries graciously offers you 2 of its logos (black on white or white on black) or lets you draw one. As if the per-key RGB and actuation weren’t enough customization, I opted to make this keyboard even more me by inscribing my name.

I opted to embellish the keyboard with my own signature. Although it's OLED, perhaps that tech’s biggest gain is its high contrast. But rest assured all text is legible and sharp from normal viewing distance, although you can spot individual pixels if you peer closely.

Bottom Line

With the SteelSeries Apex Pro, you can have a keyboard that’s truly you, from the screen to the S key. With innovative adjustable actuation points, a screen sufficient for mainstream users to enjoy the keyboard without software and a welcoming wrist rest, it starts earning its $199 MSRP. Configurable profiles, including 5 on-board ones, let advanced users design profiles with per-key lighting and switch settings and easily pick one for their current app.

Although SteelSeries’ linear switches are groundbreaking, others are comfier during long typing sessions, including the Wooting One, which also has customizable actuation for keys with more clatter and a tactile bump. It’s also cheaper ($140), but actuation is not per-key and switching profiles is far less seamless. You also get eight pressure-sensitive keys with the Cooler Master MasterKeys MK850 ($175, also linear Red switches) but can’t customize actuation.

Unless you want high-level customization for individual keys, you can find a better typer for less money. But the Apex Pro’s per-ley adjustable switches make a noticeable difference, especially for competitive gaming, and are integrated more seamlessly than competitors. If you’re okay with linear switches and want a keyboard that you can really make your own, count on the Apex Pro.

Image Credits: Tom's Hardware

MORE: Best Gaming Keyboards

MORE: All Keyboard Content

From Beer, to the Iron Throne and Pork Ramen

As we visited the many tech booths at Computex in Taipei this week, we didn’t stop to smell the roses; instead we paused to gawk at the coolest case mods. With nostalgic references, moments of zen, nods to athletes and enough RGB to satisfy the most gluttonous color addicts, here are the mods we loved the most.

It’s 5 O’Clock Somewhere

A beer is always welcome, but a beer on a busy showfloor like Computex’s is like a gift from above. This mod from Taiwanese modder Karens XU, aka K Mod Studio, features an Intel Core i7-8700K, booth host G.Skill’s Trident Z Royal RGB RAM, MSI Meg Z390 Godlike motherboard, Zotac Gaming RTX 2080 AMP Extreme graphics, a Plextor M9PeGN 512GB M.2 PCIe SSD, Cooler Master’s MasterFan MF121R RGB and Bitspower water cooling. But more importantly, it was dispensing real cups of Heineken.

“Scorpio”

As a menacing scorpio myself, I couldn’t resist the poisonous lure of Suchao Modding Design’s Scorpio mod. Lurking around Colorful’s iGame booth, it had an i3-8350K, iGame GeForce RTX 2080 Vulcan X OC, Z390 Vulcan X and 8GB DDR4-3200 RAM, plus a 1TB Colorful SL500 SSD.

“Quasar”

We found this out-of-this-world md over at PSU vendor Seasonic’s modding showcase. The modder was inspired by popular science fiction, and boy does it show. Naturally, it uses a Seasonic power supply.

Steampunk Typewriter

The ASRock booth took us back to a simpler, clickier time with this retro-looking build. Coming from Mark’s Fabrication, components include the Trident Z Royal, ASRock’s Z390 Taichi motherboard and lots of vintage style.

Lotus Flower

Sometimes you need to get a little zen. And for tech geeks, that can require 3D printing your own lotus flower case and custom CPU cooler. Filipino modder Samca worked full force with T-Force, using its XCalibur RGB DDR4-3600 (8GB) memory, Delta R RGB SSD (250GB), plus EVGA’s H370 Stinger motherboard, RTX 2060 XC Gaming and Supernova 650 GM PSU and an i5-9500.

The Iron Throne

With HBO’s Game of Thrones officially over, who doesn’t want their own Iron Throne to cope? Bitspower had one at their booth from YJ Mod, who used an Intel CPU, Asus ROG Strix Z390-I Gaming board, OCPC X3treme Aura DDR4-3600 RAM and zero spoilers for this timely build.

Apex Legends Loot Tick

The Philippines' JPModified repped battle royale fans with this mod for G.Skill. Those who can find and destroy this Loot Tick will score an i7-8700K, ROG Strix Z370-I Gaming, RTX 2070 Founder’s Edition, Trident Z Royal RAM, G.Skill Ripjaws S3 SSD and Silverstone SX700-LPT PSU. But please don’t destroy JP’s mod; neither shipping it to Taipei (it took four different sections in one crate) nor 3D printing those feet were easy. 

Olympic Stadium

Tokyo will host the Olympics in 2020, but in 2019, it’s Bitspower’s Computex booth that held the Olympics, mod that is. Ice Freeze’s gold medal-worthy build uses an i7-8700K, ROG Strix Z390-I Gaming, Klevv Cras X RGB DDR4-3466 and WD Black SN750 NVMe SSD (250GB).

War Machine

This PC looks like something you shouldn’t pick a fight with. The monstrosity is a mod of an Acer case, which modder Jengki Wmp filled with an i7-8700K, Acer motherboard, Nvidia GeForce GTX 1080 Ti, Plextor 512GB SSD and FSP Hydro G 850 W PSU and kept cool with Bitspower water cooling.

RGB Monster Truck

This monster truck better not enter any rallies because we’d just cry if it got covered in dirt. Instead, we found it safely at the iGame booth, rocking the vendor’s RTX 2080 Ti Vulcan X OC, Z390 Vulcan X, Vulcan U 3000 DDR4 RAM and an i9-9900K.

“ElmorLabs Ice Cube”

ElmorLabs kept it cool with its mod for G.Skill with the chillest looking build on the Computex showfloor. It includes Trident Z RGB RAM and and ROG Maximus XI Apex motherboard.  Liquid nitrogen (LN2) cools it to -50 degrees Celsius, great news for its i9-9900K.

“Best Pork Ramen Store”

Adorable or scary? You decide. We don’t know why these pigs would want to hang out so close to a place that cooks pork, but we do know Vietnamese modder Amber Spider outfitted this surreal creation with an i7-9700K, MSI MPG Z390I Gaming Edge AC and RTX 2060 Ventus 6G OC, HyperX Predator DDR4 RGB RAM, Bitspower Paris cooling and an FSP Dagger Pro 650W PSU.

"I Choose You"

Kanto meets Reynolds in Stefan Ulrich’s Pokémon themed mod. This year, Thermaltake’s booth featured mods of its case Level 20-series cases. Uniting the original three Pokémon choices Professor Oak gives you (both in display and in coolant color) in the video game with nods to the new Detective Pikachu movie, there’s nothing for a Pokémon fan (new or old) not to enjoy. 

Ready, Aim, Fire

Despite using the same style case as the other mods at Thermaltake’s Computex booth, the one belonging to Mike Samsin, from the Philippines, looks like it could physically obliterate the others. Let’s just be grateful that thing isn’t loaded.

“Cyberbike”

If you’re looking for a less combative form of transportation, perhaps the Cyberbike is more your style. Displayed by Enermax, it uses the firm’s Revolution SFX 550W PSU, Liqfusion 240 CPU cooler and Trident Z Royal RAM.

“Mini TV”

AMD PC mod inception. Found at Alphacool’s booth, it’s an AMD Ryzen 5 2400G-based system on an ASRock A300M-STX board with a Plextor M9Pe SSD, chilled by an RGB Alphacool Eisblock XPX CPU block and accompanying pump and radiator, all housed in a modded version of AMD’s first-generation Ryzen Threadripper box. For authenticity, we hope you need to use the Torx screwdriver included in the Threadripper packaging to get to what’s inside.

“Mortar V2”

Speaking of things we hope aren’t loaded, this mortar-inspired mod, spotted at the AITC booth, was another one of the few AMD mods. It hosts an AMD Threadripper 1900x CPU, AITC’s Rapidez RGB RAM, an Asus Prime x3999 motherboard, Corsair 1600W PSU and Bykski cooling.

Team Blue

No question what brand CPU this Vikingdom’s mod uses. No matter your CPU team preference, you’ve got to admit this is a sweet rendition of Intel’s unique 9th Gen packaging.

Fittingly, it taps an i9-9900K and Intel’s 750P 512GB SSD, plus Trident Z Royal memy, ROG Strix Z390-I Gaming and GTX 1080 Ti O11G Gaming and ROG Thor 1200W PSU.

Traditional China

Ni hao! Inspired by traditional Chinese culture and respect for the environment (as per PSU vendor FSP’s philosophy), Zhengzhou 90IT computer studio’s mod depicts a delightful scene, especially with its bridge and tiny fisherman and koi pond.

In addition to an FSP Dagger Pro 650W PSU, traditional China is sitting atop an i9-9900K, ROG Strix Z390-F Gaming and RTX 2080 08G Gaming, Trident Z Royal RGB, Samsung 970 EVO Plus M.2 SSD (500GB) and Bitspower Paris water cooling all packed into Thermaltake’s Core P5 case.

Tall Glass of Water

G.Skill actually had to lower their platform for MP Customize’s mod, a meter-tall sight to be seen featuring an i9-7900K, Asus ROG Rampage Extreme motherboard, Zotac RTX 2080 Ti ArticStorm, Samsung NVMe SSD 960 PRO (512GB), Bitspower cooling and impressive cable management thanks to Cablenaut.

“Cap-Ten’s” Ship

German modder Cap-Ten made us want to watch sci-fi with his creation. He took off with an Intel CPU (unspecified), MSI MPG Z390I Gaming Edge AC motherboard, Trident Z Royal memory, MSI GeForce GTX 1070 Aero and Crucial SSD M.2 525GB with power supplied by the Enermax Revolution SFX 550W.

Skull Mountain

We found AI-Mask’s frightening mod at G.Skill’s booth, but it’s presumably from the depths of Hell. It houses an i9-9900K, ROG Maximum XI Gene and all the bones of our ancestors.

I’ve Come From the Future

Looking like something from a distant and way more shiny future, we found this over at Bitspower’s booth. Modder Nguyen Dinh Ban modded a case by fellow modder nhenhophach. It has an i9-9900K, Gigabyte Z390 Aorus Xtreme board, Corsair Dominator Platinum RGB DDR4-3200, Nvidia’s RTX 2080 Ti, Samsung 970 Pro NVMe M.2 SSD and FSP Hydro PTM+ 850W power supply.

Whether you're typing or gaming, your keyboard experience is all about the switches. You can get switches with a low-actuation point that triggers when you barely press or ones with a high actuation point that makes you push the key nearly all the way down to the bottom before it registers. But with a few special exceptions like the analog Wooting One keyboard, you can't adjust the actuation on a keyboard.

Enter SteelSeries' new OmniPoint switches, which let set their actuation point in 10 different positions, ranging from a light 0.4mm to a hard 3.6mm. So, if you want a light touch for blasting enemies in Counter-Strike: Global Offensive, you can set it to 0.4mm and rapid fire practically just by breathing on the key caps. Then, when you're writing a report, you can go for a more reasonable actuation point of 2mm so that you avoid accidental keypresses.

These switches are SteelSeries' own design and are exclusively available in its upcoming Apex Pro keyboards. I had a chance to go fingers-on with an Apex Pro here at Computex Taipei and I was impressed not only with the technology, but also with the keyboard's features and design.

Apex Pro's Other Features

Due out in a few weeks for $199 (or $169 for the tenkeyless version), the Apex Pro has some neat extras you don't find on every premium keyboard. A black and white OLED screen sits in the upper right corner of the keyboard with a dial and a button next to it. You can use these navigational controls to set everything, from the lighting configuration to the macros and the actuation level, without installing any software on your PC. You can also configure all these options through the SteelSeries app in Windows where you can upload an animated GIF to display on the screen if you want.

The Apex Pro comes with a comfy, detachable soft-touch wrist rest, which in my brief experience, felt much better than the cheap plastic units you get with most gaming keyboards. However, it's not padded or filled with gel so it's not going to be quite as comfy as our favorite keyboard accessory, the HyperX Wrist Rest.

The back surface of the Apex Pro has a USB Type-A pass-through so you can attach a mouse or another peripheral to it. Rather than acting as a hub, the keyboard has a built-in cable that splits into two USB Type-A ports, one for the Apex Pro and another for the pass-through port. The port itself has a white backlight to help you find it in the dark. There's also a cable routing tunnel in the bottom.

How OmniPoint Switches Work

The OmniPoint switches have a linear feel that's very similar to the Cherry MX Red switches you'll find on so many other keyboards, but they work in a way that's different from normal mechanical switches. Where a normal mechanical switch actuates when two pieces of metal make contact during the key press, these OmniPoint switches use magnets and sensors that don't need to touch.

As you press the key down, you're moving a small magnet in its stem closer to the sensor, which uses the "Hall Effect" to determine the exact distance between the base and the stem. When it hits the actuation point you configured in software, the system registers a key press. 

Aside from their configurability, "Hall Effect" switches have a couple of advantages over regular mechanical switches. Because there's no physical contact, there's less wear and tear on the mechanism so the switches last longer. More importantly, the debounce time -- the delay the system waits between presses to make sure it doesn't register the same stroke twice -- is just 0.7 ms as opposed to 6 ms on a regular switch. Combine that with an incredibly shallow actuation point and you can fire your in-game weapon more quickly.

Using a demo that's built into the keyboard, I was able to adjust the actuation point just by rotating the dial. Then, I watched the OLED screen which measured how far I depressed the key and lit up to show when it had actuated. With a 0.4mm actuation point, the system registered a key press, just from me resting my fingers on the home row, a potential horror show for typing. At 3.6mm of travel, I had to practically bottom out to register my key press.

The slider in the app lets you choose among 10 different actuation points. For typing, I'd probably go with the middle point or just one notch below whlie, for gaming, I think I'd choose something a little deeper than 0.4mm.

But the good news is that you don't have to pick just one actuation point for the whole keyboard. You can configure different actuation points for different keys. If you are jumping too often because you accidentally tap the space bar, you can give it a deeper actuation point so you have to really hit it hard to register. And you can configure the actuation points to change for every application on your computer. So, your favorite game could have one profile and Microsoft Word could have another.

Because the OmniPoint switches cost more to manufacture than regular switches, the keys in the function row, the number pad, the arrow keys and other keys outside the main letter / number block use standard red mechanical switches that SteelSeries makes in partnership with Gateron. 

The Apex Pro and the OmniPoint switches that power it have the potential to really change the way we think about gaming keyboards. We look forward to testing these more thoroughly when they get closer to launch.

Features and Specifications

Fractal Design's Define series has long been a favorite in the enthusiast community for its high-quality construction and class-leading performance. That said, the company had fallen a bit behind when it comes to lighting options--until now. The Define S2 Vision bundles classic Fractal Design styling with integrated addressable RGB (ARGB) lighting and enough tempered glass to show it all off. But at $280 (£215), there might be other options out there.

Specifications

Constructed of tempered glass, plastic, and steel that’s painted black inside and out, the Fractal Design Define S2 Vision RGB measures 21.4 x 9.2 x 18.3 -inches (543 x 233 x 465mm) and weighs almost 26.5 lbs (12kg). This $280 (£215.22) chassis comes with a two-year limited warranty.

Aside from its performance and build quality, the case’s biggest selling points are the added tempered-glass panels and ARGB lighting. The Fractal Design Define S2 Vision RGB incorporates the company's latest AR-4020 ARGB LED lighting strip, the Adjust R1 RGB controller, and four Fractal Design Prisma AL-14 PWM ARGB fans. A non-RGB 'blackout" version of this chassis is also available for about $90 less.

The entire front panel and the majority of the top panel are tempered glass, attached to a plastic exterior frame. The front features filtered vents running down both sides, and three of the case’s Prisma AL-14 PWM ARGB fans are mounted behind it.

Two USB 2.0 ports, two USB 3.0 ports and a single USB 3.1 Gen 2 Type-C port, headphone and microphone jacks, power and reset buttons and a hard drive activity light are embedded in the leading edge of the top panel. The top of the chassis is equipped with a strip of LED lights and can be configured in two ways: either with a glass panel that has an unrestricted view of the interior, or with an included cooling bracket and filter set built for fans, radiators and all-in-one coolers up to 360mm. 

Much like Cooler Master and NZXT, Fractal Design has also ditched the use of rubber-coated locating pins and thumbscrews, opting instead for a tempered-glass panel bonded to a metal frame. This assembly attaches to the chassis via push pins. The opposite side panel is also made of tempered glass. Both side panels are secured in the rear with metal thumbscrews.

The rear of the chassis has a fairly standard layout; here you'll find seven standard expansion-card slots (plus two vertical for showing off your graphics card), a motherboard I/O area, an opening for a bottom-mounted PSU, and a 120mm/140mm exhaust-fan mount with slotted screw holes that let you adjust the position of the fan to fine tune airflow or make room for system components.

A large removable plastic mesh filter cover is the entire length of the bottom of the case. The large, rubber-coated feet keep the case approximately one inch off the ground to facilitate air intake to the power supply.