Gamers who find themselves “limited” by SteamOS capabilities can now more confidently install Windows on their Steam hardware. Valve just released the Steam Hardware - Windows Resources page on Steam, which contains Windows drivers for the Steam Deck LCD, Steam Deck OLED, and Steam Machine. This means that those who want to replace the stock SteamOS with Windows 11 on their devices can now do so more easily, and they don’t have to hunt around the internet for drivers just to ensure that all the parts and components on their console will play nicely with Windows.

The drivers on the page include those for the APU, graphics, SD Card reader, Wi-Fi, Bluetooth, and Audio, depending on which Steam hardware you’re driving. However, Valve said, “We are providing these resources as is and are unfortunately unable to offer 'Windows on Steam Hardware' support. If you get stuck and need a way back to the default SteamOS, please follow these recovery instructions.”

This is the main reason why Valve refuses to subsidize the Steam Machine — since owners can essentially do whatever they want to the living room PC console, selling one does not automatically guarantee that the company will make back any discounts it offers through game sales. This is why even though the Steam Machine is highly rated for Steam Deck users who want to upgrade to a more powerful device, its relatively high price puts it at a disadvantage compared to competitors like the PS5 or Xbox.

Many gamers would likely prefer SteamOS over Windows, especially as it offers far less bloat than Microsoft’s operating system. Valve is in fact working on SteamOS to make it more compatible with a general release and allow it to be installed on hardware that features Intel CPUs/GPUs and Nvidia GPUs. But because many anti-cheat apps still aren’t compatible with Linux (hence, SteamOS), installing Windows on Steam hardware is currently the only way that gamers can enjoy titles that require them.

It’s unfortunate that SteamOS doesn’t have any dual-boot capabilities at the moment, as that would give gamers the best of both worlds — use Windows and accept the performance hit if they want to play games that require specialized anti-cheat apps and then switch to SteamOS for all the other titles they enjoy. Valve said that it’s already working on this capability, but hasn’t given a fixed timeline for when it will arrive.

Earlier this week, we reported on the Green Team officially retiring the creaky Nvidia Control Panel (NVCP), with all its major settings adjustments claimed to have been ported to the Nvidia App. Throughout its tenure, this long-in-the-tooth piece of graphics settings software stuck resolutely to the classic non-themed Win32 controls style, but we know there will be holdouts and those who miss it for one reason or another. Thankfully, Nvidia has left an NVCP installer in the Microsoft Store, for now. Let’s look closer at whether it is worth a separate download in mid-2026.

To be clear, you will still need to download a modern Nvidia graphics driver to use the separately available NVCP. The familiar control panel software that is now available via the Microsoft Store is simply a controls access UI - one that has now been relegated to a secondary, optional choice.

For existing Nvidia graphics card users, you probably won’t have to go out of your way to grab the old NVCP from the Microsoft Store. It should normally persist from previous driver installs and updates, unless you opt for a ‘clean install’ from now on.

Possibly the primary reason you will want to keep a copy of the NVCP handy is the updated Nvidia Apps’ missing “professional features.” From my nosing at the information available, RTX Pro / Quadro features - things like offering adjustments to Mosaic, Sync, stereo, and a few pro‑workflow toggles - are yet to be migrated. So, modern GeForce gamers shouldn’t worry about hanging onto NVCP for functionality.

Even if you don’t need the handful of missing features in new vs old, some folks will want to keep using the NVCP due to familiarity with Nvidia’s older lightweight settings software.

We’d also be tempted to download or keep a shortcut to the NVCP handy in case upcoming software from Nvidia messes up the controls accessible in the Nvidia App. The Green Team’s software has come under fire for a string of buggy releases lately. In the likely scenario that Nvidia will ship an upcoming version of its Windows drivers with some feature-breaking wrinkles or crashing issues, the NVCP might be handy for fallback or troubleshooting.

Microsoft has outlined a new feature of Windows called Cloud-Initiated Driver Recovery (CIDR). This newly introduced capability lets Microsoft remotely roll back a bad driver to a previously known good version on affected PCs. Moreover, it can work without user action or OEM intervention. It sounds like a magic bullet for a long history of Windows Update woes, but we’ll have to see if it works when the rubber hits the road. CIDR will only work with drivers distributed via Windows Update.

Windows Update can cause plenty of problems when a bad driver gets through testing and gets pushed to users. Indeed, buggy drivers have caused many a lost hour, gray hair, wrinkle, high blood pressure, and so on, among Windows veterans. Microsoft also notes that a bad driver often means a user has to manually intervene and roll back to “a low-quality driver for an extended period.” So, the new CIDR is cautiously welcomed.

Microsoft spells out the CIDR process in its Tech Community blog, and there we learn that recovery starts by the Windows developer triggering "a recovery action directly from the Hardware Dev Center (HDC) Driver Shiproom." Once a problematic driver is flagged, the system recovers the previously known-good version of a driver via the Windows Update pipeline. “This is handled through coordinated updates to the PnP driver stack and the driver flighting and publishing services,” says Microsoft.

Importantly, Microsoft notes that “recovery is delivered through the existing Windows Update infrastructure — no new client agent or partner tooling is required.” That should help CIDR work well with what we’ve already got and for it to become an established standard. Likewise, partners don’t need to get involved in CIDR, Microsoft will manage it. However, Microsoft asks that these partners “continue monitoring their driver quality metrics in the Hardware Dev Center dashboard and to respond promptly to any shiproom feedback on rejected submissions.”

CIDR is rolling out now for validation and testing, and it is expected to automatically support the Hardware Dev Center publishing process from September onwards.

Releasing a new GPU architecture is a complex affair. Nvidia should know this as well as any company, considering it's been making graphics cards for 27 years now, every one of which needed drivers. But the Nvidia Blackwell RTX 50-series GPUs haven't had the smoothest of launches. The RTX 5090 at least offers new levels of performance, while the RTX 5080 delivers only slightly better performance than the prior generation RTX 4080 Super. But both GPUs seem to be suffering from a case of early drivers and the need for additional tuning.

It's not just about drivers, though — or perhaps it is, but for Blackwell, Nvidia recommends new benchmarking methodologies. It presented a case at its Editors' Day earlier this month for focusing more on image quality and analysis, which is always a time-consuming effort. But it also recommended switching from PresentMon's MsBetweenPresents to a new metric: MsBetweenDisplayChange (which I'll abbreviate to MSBP and MSBDC, respectively).

The idea is that MSBDC comes at the end of the rendering pipeline, right as the new frame actually gets sent to the display, rather than when the frame finishes rendering and is merely ready to be sent to the display. It's a bit nuanced, and in theory, you wouldn't expect there to be too much of a difference between MSBDC and MSBP. Intel has also stated that MSBDC is the better metric and recommended using it for the Arc B580 and Arc B570 launches.

Part of the issue with MSBP is that it doesn't necessarily capture information correctly when looking at frame generation technologies. And, in fact, if you try to use MSBP with Nvidia's new MFG (Multi Frame Generation), you get garbage results. This wasn't the case with DLSS 3 framegen, but MFG reports data in a different manner:

All of the fields shown are useful data, but with MFG 4X the values in the MSBP column now function differently. You can view it as the total time required for the GPU to render the frame in the traditional manner, so in this case, it's about 27.66 ms per frame, followed by nearly instant "rendering" times for the three generated frames. The new flip metering hardware in Blackwell GPUs then attempts to evenly pace the actual display of the generated frames on your monitor.

Looking at the MSBDC column, we find a far more consistent sequence of frametimes, as expected. Instead of the "fast-fast-fast-slow" frametimes in the MSBP column, we get relatively similar frametimes of around 7.35 ms per frame. The MsBetweenSimulationStart column ends up giving the timing of user input sampling. So, in this case, the user input sampling happens every 29.18 ms, or at 34.3 FPS, while the generated rate of unique frames sent to the monitor runs at 136.1 FPS — basically four times the input sampling rate, as expected.

The above results, incidentally, are taken from Cyberpunk 2077 benchmarks running at 4K with the RT-Overdrive preset (aka path tracing or full RT), DLSS Quality Transformers upscaling, and MFG4X frame generation. As I noted in the 5080 review, the smoothness of the framerate as seen on the monitor does make the game feel better than if it were running at 34 FPS, but it also doesn't feel like it's running at 136 FPS because input sampling is happening at the base framerate of 34.

But back to the discussion of MSBP and MSBDC, it's a relatively easy "fix" to switch between the two metrics. Also, without using framegen of any form, we'd expect the resulting performance metrics to look pretty similar. But "similar" isn't the same as "identical," and since I already had all of this data, I decided to take a closer look at what's going on and how the new metric affected my benchmark results. It's probably not going to be what you'd expect...

The above gallery shows the 14 graphics cards I've tested (so far...) using our new test suite and PC. There are 22 games, six of which have ray tracing enabled and 16 that are running in pure rasterization mode. Each image shows the same GPU, using the MSBDC metric compared to MSBP, and the third column is the import one as it shows the percentage change — how much better or worse MSBDC is compared to MSBP.

The average framerates are as expected. Whether you measure at the start or end of the rendering process, overall the total time averages out to the same value. There are differences, however, as the data in MSBDC sometimes doesn't appear useful. There are MSBDC values of 0.000 ms at times, and this throws off my formulas, so I just tossed those values.

"Instant" frametimes would pull the average FPS up if included, but I'm not sure what they're supposed to represent, as obviously it should take some amount of time to render and display the next frame. Are they dropped frames or something else? It's not clear, so I asked Nvidia for feedback, as I'm using its FrameView utility to capture these benchmarks. Note also that the lower spec cards (RTX 4060/3060, the Arc cards, and AMD's 7600/XT/6600) were all tested last month using an older version of FrameView that appears to have been more prone to getting 0.000 results for MSBDC.

But really, the changes in the 1% lows are interesting to look at. All of the Intel Arc GPUs show improved 1% lows, increasing by up to 8.3% on average. That's a pretty decent improvement for what would otherwise seem to be a relatively minor change in how performance gets measured. The RTX 3060 and 4060 also show modest improvements, and these were tested with older 566.36 drivers. AMD's GPUs also show modest improvements in minimum FPS.

And then we get to the newer Nvidia cards, which were tested with preview 571.86 drivers — and 572.02 on the 5080. The 4080 Super, 4090, and 5090 all show worse 1% lows at 1080p medium/ultra, and equal or slightly better minimums at 1440p and 4K. The 5090, on the other hand, sees worse typical 1% lows at 1080p and 1440p, and basically equal minimums at 4K. There are plenty of games that show a marked drop in 1% lows as well, while a few see a modest increase.

The RTX 5080, meanwhile, gets a small bump in minimum FPS on average. There are still a few games where it sees a relatively significant drop (Baldur's Gate 3, both Flight Simulators, and A Plague Tale: Requiem), and also some relatively substantial gains in other games (Assassin's Creed Odyssey, Cyberpunk 2077, F1 24, God of War Ragnarok, Hogwarts Legacy, and Star Wars Outlaws).

And it's data like this that makes the drivers and launch of the RTX 5090, in particular, feel rushed. Improving the frame pacing and reducing stutters should be the goal. The RTX 5090, with 32GB of VRAM, appears to have some difficulties here, particularly at 1080p, suggesting the drivers were still a bit undercooked. Most likely, things will improve, but it's also interesting that the 4080 and 4090 lost performance at 1080p as well using the new MSBDC metric with the latest drivers (at the time of testing).

There's also another side note I want to mention: Idle power draw on the RTX 5090 was abnormally high. Like, "AMD RX 7900 XTX at launch" levels of high. The 4080 Super, 4090, and 5080 all tended to idle in the 20~30 watts range. The 5090 seemed to bounce between a low of perhaps 50W, but was often at around 90W. Could that impact performance in some way? Yes. It's yet another item to keep an eye on for the future, as the 50-series GPUs are supposed to be better at idle power, not worse.

The bottom line, as we hopefully conveyed in the 5090 and 5080 reviews, is that Nvidia's driver and software teams still have work to do. That's always the case, but for a new architecture, it's especially true. We'll be retesting the 5090 and 5080 again in the coming weeks (and months) with updated drivers, and my expectation is that 1% lows could see double-digit percentage improvements. Average FPS might see some decent gains as well, though that will probably be more on an individual game basis.

I guess Nvidia was feeling a bit jealous of Intel's "fine wine" approach to drivers.

In the family of Windows stop codes, DRIVER_POWER_STATE_FAILURE is intriguing. It is associated with stop code 0X9F in bugcodes.h and means some drive has reported an invalid or impossible power state. This typically involves sleep: that is, transitioning into sleep while the PC is running, or experiencing an out-of-bounds power state while waking from sleep. As with all stop codes in Windows, 0X9F crashes the PC and emits a BSOD to avoid damage to the system or storage devices that might otherwise occur if drivers and their attached devices kept on working.

Diagnosing DRIVER_POWER_STATE_FAILURE Causes 

If you view the crash dump file from this BSOD, which you can do by installing and launching BlueScreenView, it will often identify a file by name as part of the crash data. By searching on that name, you can often associate it with some specific device built into or, as is most typical for this error, plugged into your PC.

USB drives (and other USB-attached devices) are among the most common involved in DRIVER_POWER_STATE_FAILURE BSODs. This includes flash drives and drive docks or enclosures, which may mouse NVMe or mSATA SSDs, conventional HDDs, and so forth.

When dealing with a DRIVER_POWER_STATE_FAILURE BSOD, disconnect all external USB storage devices (keyboards, mice and monitors seldom provoke this stop code). As the name of the error suggests, if a driver for any such device has been recently updated, it may also make sense to roll back (if possible) or to replace the current driver with an older, known working driver for your PC. If you don’t have a current backup of your PC, you should create one as the first step in the repair process. Then if anything goes south along the way, restore that backup to get back to where you started without much muss, fuss, or stress.

Boot into Safe Mode to Uninstall Newly-Added Drivers

If you continue to get DRIVER_POWER_STATE_FAILURE errors, boot your PC into Safe Mode. When you get to the desktop, run Device Manager, and use it to uninstall any newly-added device drivers. Simply right-click any newly-added device, then select Uninstall device from the resulting pop-up menu. If this doesn’t roll you back to a previous driver, Windows Update should furnish one during your next reboot.

Tip: to launch Device Manager, click the WinKey-X combination on your PC and select Device Manager from the resulting pop-up menu. Surprisingly, you’ll find USB storage devices under the Disk Drives heading in Device Manager. As an illustration, I’m selecting the Mushkin 16GB UFD currently plugged into the PC’s USB port in the screenshot below (see how to take screenshots in Windows 10 or 11 if you ever need to capture your own).

In most cases, this maneuver will fix the DRIVER_POWER_STATE_FAILURE problem. The next time you plug that device in (or if you leave it plugged in, reboot it) Windows will install the latest driver in the Update Catalog for that device. If the problem recurs after that, you’ll need to find a different driver instead.

Update Device Drivers

It’s not always necessary to update device drivers – except, that is, when the current driver causes trouble. And indeed, a BSOD is about as troublesome as a driver gets. Thus, if what Windows supplies by default doesn’t work (DRIVER_POWER_STATE_FAILURE BSODs continue) you’ll need to find a driver that works, or forgo use of the offending device until a working driver becomes available. Be sure to visit the vendor’s website to see if a newer driver is available for the problem device (the PC or laptop maker, and driver repositories  -- e.g. Station Drivers -- are also good places to look). If you find a good candidate, please give it a try: it may fix your problem. If so, count yourself lucky.

Round Up the “Usual Suspects”

If your driver maneuvers don’t restore things to normal operation and stop the DRIVER_POWER_STATE_FAILURE errors, it’s time to run a standard set of Windows repairs to see if they help. This means opening an administrative Command Prompt or PowerShell session and running the following commands:

Dism /online /cleanup-image /restorehealth

SFC /scannow

The first of these commands uses the Windows Deployment Image Servicing and Management (DISM) command to check the components of the current running Windows image, and attempt repairs on any components that come up short during its integrity and checksum tests. In most cases, this will fix issues related to corrupt or damaged files in the Windows component store (by default C:\Windows\WinSxS).

The second of these commands runs the Windows System File Checker (SFC) and will replace any such file it finds that are suspect or damaged. Microsoft recommends running DISM before using SFC. This is interesting, because SFC has been around longer than DISM (it goes back to Windows 98, whereas DISM appeared in 2008 for Windows 7). That said, DISM has additional smarts in conducting repairs, and can work from online or offline image sources to provide the basis for repairs if the simple version of the command shown above doesn’t work (you’ll get an error message that attempted repairs failed). Read the Microsoft Docs item Repair a Windows Image for all those gory details, if you must.

Driver In, Driver Out 

In most cases DRIVER_POWER_STATE_FAILURE is addressed by getting the offending device out of the picture, and fixing whatever driver issues might present. If worst comes to worst you can always replace the problem device with another (preferably different) one, and keep on computing. Here’s hoping you don’t have to go that far!

Related Tutorials

• How to Fix a Windows Blue Screen of Death (BSOD)
• How to Make Windows 11 Look and Feel Like Windows 10
• How to Make a Bootable Windows USB Install / Recovery Drive
• How to Get Windows 11 for Free (or Under $15)

Intel is hard at work building Linux-ready graphics drivers for its next-generation Battlemange GPUs, which will soon arrive with Intel's Lunar Lake CPUs. Phoronix reports that Intel will have a functional display driver for Battlemage once the upcoming Linux 6.11 kernel cycle rolls in. This new display driver is by no means complete, but it provides display functionality for Battlemage GPUs for the first time in Linux, which is a big deal regardless.

Right now, Intel is actively working on Direct Rendering Manager driver changes, HPD display enabling, and a number of power-efficiency-focused features for the upcoming Linux driver. These abilities include panel replay functionality, which enables a driver to just refresh portions of the display showing new information, rather than refreshing the entire frame buffer. It also has Auxless advanced link power management, link-off-between-frames support, and content match refresh rate, most of which are more power-saving techniques. Content match refresh rate is a variant of VRR, which purportedly prevents frame drops in certain types of content.

It makes a lot of sense for Intel to be prioritizing power-saving techniques in its Linux-based drivers first. There's a good chance Battlemage will be released as an integrated graphics solution first in Lunar Lake processors before discrete GPU variants arrive. This means Battlemage's primary competition will be against integrated graphics solutions in Qualcomm's Snapdragon X Plus and X Elite processors, as well as AMD's all-new Zen 5-based Ryzen AI 300 processors. Power efficiency is one of the biggest strengths of these chips, especially the Qualcomm Snapdragon counterparts, so it's reasonable for Intel to prioritize power efficiency techniques first in its driver as it tries to close the gap on Qualcomm with its next-gen Lunar Lake architecture.

Even if mobile Battlemage doesn't arrive first, the architecture will be used to a much greater extent in Intel's integrated graphics solutions. Intel produces significantly more CPUs than GPUs, so inevitably, Battlemage will have significantly larger adoption as an integrated graphics solution rather than a dedicated gaming GPU or enterprise GPU.

In integrated graphics form, Intel is touting a 1.5x performance upgrade over its previous-gen Arc Alchemist-based integrated graphics chips. The architecture has been optimized from the ground up, boasting larger XMX engines (for AI workloads), enhanced vector engines, larger RT units, and larger caches. The XVE vector engines are now capable of doing eight 512-bit multiplies per clock and eight 2048-bit vectors per clock via the XMX engines. Intel is also increasing the SIMD engine to 16 lanes.

The vertex fetch throughput and mesh shading performance have been increased by 3x, as well, which are important attributes that increase 3D graphics performance.

Intel's integrated graphics projections for Battlemage won't be the same as the discrete GPU versions, because Intel is using low-powered transistors for its integrated graphics versions. But there's a good chance Battlemage in discrete GPU form will be a big upgrade regardless, due to all the changes Intel has made to its upcoming Battlemage GPU architecture.

AMD has stated that it's on track to release its Micro Engine Scheduler (MES) documentation in late May, followed by source code. Then it will follow through with releases of additional parts of the Radeon stack as open-source. The statement was seemingly made in response to a Tweet/X by Tiny Corp, which has been publicly corresponding with (and often criticizing) AMD on social media for several months.

The discourse between Tiny Corp and AMD has made more than a few headlines in recent months. The former designed and pre-sold its TinyBox AI server, which garnered interest due to its use of relatively affordable AMD Radeon GPUs. However, the firm’s social media and its founder, George Hotz, became quite vocal as it discovered the consumer-grade cards weren’t behaving in a server-grade or enterprise-ready manner.

In brief, Tiny Corp wanted greater / deeper access to AMD hardware, firmware, and driver IP. The company thinks that with sufficient access to the firmware and driver stack, it can make the Tiny Box function as advertised. Despite Tiny Corp being a small company,  AMD engaged, and even Dr Lisa Su joined the conversation in early March. Su said the “team is on it,” back in March, but Tiny Corp remains grumpy about the situation it finds itself in to this day.

“We are working to release Micro-Engine Scheduler (MES) documentation towards end of May and will follow up with published source code for external review and feedback,” noted the official AMD Radeon Twitter/X account in early April. “We have also opened a GitHub tracker, which will have the latest status on fixes and release dates.”

Today, we saw a significant update in documentation and open-sourcing progress at AMD. In a reply to a further barb from Tiny Corp, the red team’s graphics division reiterated the MES documentation statement from earlier in the month. It added, “Additional parts of the Radeon stack to be open-sourced throughout the year,” before directing interested parties to keep an eye on a GitHub repository.

Tiny Corp has already replied to AMD’s statement, characterizing the MES news as “a red herring,” and asking for further parts of the architecture to be open-sourced, with documentation for the hardware scheduler, which it sees as a cause of Tiny Box system deadlocks.

As PC enthusiasts not particularly interested in using servers like the Tiny Box, we are still very interested in any knock-on effects the release of additional Radeon documentation and open-source software could have for us. If bugs are quashed and optimizations are made via entities like Tiny Corp, that should be good for other Radeon users. Moreover, this opening-up initiative could be beneficial to the Linux developers and community seeking to get more from their Radeon hardware.

AMD has enabled Fluid Motion Frames (AFMF) on Radeon 700M integrated GPUs with a new preview driver. The release notes for the 23.40.01.10 Preview Driver (via VideoCardz) only briefly mentions added support for the 700M iGPUs, which feature in mobile 7040 and 8040 APUs, desktop 8000G APUs, and the Z1 and Z1 Extreme SoCs for handheld PCs.

AFMF is essentially a broader version of FSR 3 frame generation that should work in any DX11 and DX12 game. In exchange for near-universal game support though, AFMF can't use motion vectors to more accurately construct new frames, which means the quality of AFMF may be significantly reduced compared to FSR 3 and DLSS 3. AFMF, like other frame generation technologies, also increases latency, though AMD's Anti-Lag+ is supposed to partially offset this.

Its relatively low quality and increased latency (which is also present in other frame generation technologies) makes AFMF a double-edged sword, but for 700M iGPUs it could come in particularly handy. Even the 780M isn't all that fast compared to modern discrete GPUs, and can easily be brought to its knees in older AAA titles. The ROG Ally, which features the 780M as part of its Z1 Extreme APU, struggled to break 60 FPS even at low and medium graphics settings in many of the games we tested.

For those games that just barely cross the 60 FPS mark even with lower-quality visuals, AFMF could come in handy. It could be an attractive alternative to reducing the resolution significantly or disabling particularly nice graphics options. There's no getting around the increased latency, which may be less noticeable in certain games, however.

There is a catch to consider though: AMD recommends enabling AFMF only if your framerate is already 60 FPS. This is presumably due to the latency being relatively high with lower framerates, and that might make such a gaming experience hard to tolerate due to poor responsiveness. Since getting to 60 FPS in the first place is difficult for 700M series iGPUs, this might nullify the appeal of AFMF in many cases.

So, if AFMF won't be particularly usable on handhelds, then perhaps it could find a niche in the desktop Ryzen 8000G series. These APUs have higher TDPs and slightly higher clock speeds for their 700M integrated graphics, which may be just enough to cross the 60 FPS threshold to enable AFMF optimally. 

Intel is preparing to introduce its own frame generation technology similar to DLSS 3 and FSR 3, called ExtraSS (via WCCFTech). Detailed at Siggraph Asia in Sydney, ExtraSS is not just a clone of DLSS 3 or FSR 3: instead of using frame interpolation, it uses frame extrapolation. Although these two methods are very similar, extrapolation has some advantages that could set ExtraSS apart from its competitors.

On the whole, ExtraSS is pretty similar to DLSS's and FSR's respective frame generation technology. Intel has built on top of XeSS and makes use of motion vectors and spatial data to improve visual quality, but with extrapolation, the data used to make a new frame is very different. Instead of using two frames to create a new one to insert in between (that's the inter in interpolation), extrapolation takes just one frame to generate the new one.

The obvious disadvantage here is the lack of extra data to put into Intel's algorithm. Extrapolation requires a higher resolution input and could still result in lots of visual glitches and artifacts, as Intel admits in its white paper. However, the benefit is that there is a reduced latency penalty compared to interpolation, which has to delay frames so it can generate new ones (otherwise, they'd show up out of order).

The reason why Intel has decided to go with extrapolation when its rivals chose interpolation is because the company has "a new warping method with a lightweight flow model," which apparently makes extrapolation more feasible. However, we don't have footage of the demo (which was done in Unreal Engine), so we can't properly evaluate how good ExtraSS is, at this time.

The key drawback of DLSS 3 and FSR 3 has been the big latency penalty, which practically needs to be mitigated by Nvidia's Reflex and AMD's Anti-Lag technologies, respectively. ExtraSS, if its visual quality is good enough, could be a potent alternative since it doesn't depend on additional features to get its latency to a playable level. However, that's a pretty big if, considering DLSS 3 and FSR 3 already struggle in visuals, and they use a higher quality frame generation method.

AMD’s Linux developers have been spotted posting patches for new GFX12 Targets on GitHub (h/t Phoronix). Specifically, two new discrete GPUs from AMD, expected to be the first to feature the RDNA4 graphics architecture, have been added as targets to the Linux LLVM compiler. These could be the first two members of an upcoming Radeon RX 8000 family of desktop GPUs.

The idea that GFX12 represents RDNA4 architecture GPUs hasn’t been plucked from thin air — previous AMD Linux patches referred to RDNA1 as GFX10, RDNA2 as GFX10.2, and RDNA3 as GFX11. Meanwhile, in October, a mysterious set of GFX11.5 support patches popped up in Linux’s Mesa 23.3 graphics driver. GFX11.5 is expected to be an RDNA3 refresh, but that has yet to be proven as we haven’t seen any such hardware officially introduced as yet.

Looking more closely at the two new compiler targets, we see that GFX1200 and GFX1201 are marked as “dGPU” components in the patches – thus, discrete GPUs or desktop graphics cards. Development is still at a relatively early stage for GFX12, as the patches reveal no details about the upcoming graphics ISA or its features. Details should be fleshed out nearer to the release of these discrete AMD GPUs. For now, developers are told in the GitHub project that the code will treat the new GFX12 GPUs as if they “behave identically to GFX11.”

Phoronix notes that the development work on GFX12 support in LLVM is required for upcoming support of these GPUs in things like the AMD ROCm compute stack, the AMDVLK open-source Vulkan driver, the RadeonSI Gallium3D driver, and more.

AMD continues to tweak and refine its FSR 3 Fluid Motion Frames (AFMF) feature. The red team’s response to Nvidia’s DLSS 3 Frame Generation can now be activated in “any DirectX 11 and 12 title using HYPR-RX or the AMD Fluid Motion Toggle,” writes AMD. This change is noted in the release notes for Adrenalin Edition Technical Preview Driver Version 23.30.01.02, Oct 13 Update. Please note this is a preview driver, indicating it isn’t ready for general consumption but is targeted at technically confident enthusiasts.

It is good to see AMD open up this technology to a much wider list of games than its previously selected white-listed titles, indicating a new level of maturity. AMD notes that users can use per-app settings to enable / disable AFMF, as they see fit. This significant step for the technology provides users with far broader game support in one swift move. Moreover, AMD recently opened up AFMF to its Radeon RX 6000 user base. Remember, Nvidia restricts DLSS 3 Frame Generation tech to RTX 40 users.

Another sign of advancement with the latest AMD preview driver is that initial support for HDR visuals has been melded into AFMF.

Early adopters and reviewers of the first preview drivers supporting AFMF have previously commented upon frame pacing issues. If the release notes are taken at face value, the Oct 13 update addresses this undesirable side-effect of the new tech, “resulting in an overall smoother gameplay experience and improved image quality.”

There is a little more evidence of progress in AFMF with regard to some specific apps / situations. For example, AMD says that Starfield now has better smoothness and image quality with AFMF applied. However, it is aware of some sizable wrinkles remaining in driver installation and BSODs “on certain systems.”

AMD says it is going to continue monitoring user issues and feedback to steer its development tweaks and continue to work on fixes. A half dozen known issues are highlighted in the latest set of preview driver release notes. We still don’t have an idea when FSR 3’s AFMF will be ready for prime time, but hopefully, the time is getting much closer.

How To Choose a Graphics Card

There has seemingly always been a love-hate relationship between Windows operating systems and printers. From issues with third-party drivers to Windows Updates breaking printer functionality altogether, Windows users have seen it all. However, Microsoft has announced plans to simplify the installation process of printers for users and reduce the number of security and compatibility issues that can arise with third-party drivers.

Starting with Windows 10 21H2, Microsoft’s IPP Class Driver has supported Mopria-compliant printers. The IPP Class Drivers allow Microsoft to provide baseline functionality for printers without installing third-party drivers or utilities.

This method of delivering native support via Windows Update works well enough that Microsoft says it will no longer service what it calls “legacy v3 and v4” printer drivers for Windows. Rather than rip the Band-Aid off in one swift motion, Microsoft is implementing a gradual phase-out.

The first significant step in the process comes in 2025 when Microsoft says it will no longer allow new drivers to be published to Windows Update via the Windows Hardware Compatibility Program (WHQL). However, hardware vendors can update existing drivers already posted on Windows Update.

Starting in 2026, Microsoft says that when installing a new printer, Windows will adjust its ranking to always prefer the native UPP Class Driver over manufacturer-submitted drivers on Windows Update. Finally, in 2027, Microsoft will no longer allow third-party driver updates unless specifically designed to address security flaws.

As you might expect, there are quite a few caveats in place for Microsoft’s phase-out of third-party printer drivers. Windows users can still install third-party drivers with a vendor-supplied installation package. Drivers already uploaded to Windows Update before the phase-out stages (and even after the 2027 end of servicing) can still be installed on a Windows system.

Microsoft also says that security concerns relating to the legacy driver platform (non-Mopria) will still be addressed per the support lifecycle of the operating system. The company also says it won’t remove existing functionality from v3 and v4 legacy drivers once the phase-out is complete.

Since the IPP Class Driver natively supports printing, faxing and scanning over USB and network, Microsoft feels this will be a smooth transition for users. In addition, all the major printer manufacturers (Hewlett-Packard, Canon, Brother, Epson, Lexmark, Samsung, etc.) offer Mopria-compliant devices.

Now, if printer manufacturers could just do something about the price of ink…

Intel has launched a new graphics driver for users of its Arc A-Series and Iris Xe GPUs. This update delivers some highly desirable fixes for Starfield players. Among the improvements users should see with Intel Graphics Driver 31.0.101.4672 include: faster load times, better stability, and fewer visual artifacts.

Last week, Starfield launched in early access and Intel Arc graphics card users were among those gnashing their teeth over glitches, wrinkles and bugs. Sometimes the bugs were hard to ignore, like when the game simply wouldn’t launch, or when it crashed after launch. We were quite surprised that Intel hadn’t prepared a game-ready driver for this highly anticipated game, as it has earned a good reputation for updated drivers and optimizations over recent months.

A few days ago, Intel followed up with an ‘Emergency’ driver update. This beta patch at last allowed Arc GPU owners to start to dabble in the game, which was admittedly still in early access at the time. Intel confirmed that the beta driver would be put through further refinement and tuning stages. With driver version 31.0.101.4672 WHQL, Intel says (PDF) the following issues have been addressed:

• Starfield (DX12) Game load duration is significantly reduced.
• Starfield (DX12) may experience instability and application crash while launching and during gameplay.
• Starfield (DX12) may exhibit texture corruptions and scene flickers during gameplay.

Like with the interim driver release, Intel says that the high-preset or below should be selected for the best stability. The above fixes will be welcomed by early Starfield adopters but there remains a list of known issues in the game, released to the masses yesterday, which Intel will be working on:

• Starfield (DX12) may experience application instability in some areas of the game.
• Starfield (DX12) may exhibit corruption when using Dynamic Resolution Scaling. A workaround is to change the Render Resolution Scale slider value.
• Starfield (DX12) may exhibit texture flickering on light sources during gameplay.
• Starfield (DX12) may exhibit low texture details on certain objects in the game.

Intel is also working on issues in the following software title: Uncharted: Legacy of Thieves Collection, Halo Infinite, Dead by Daylight, Topaz Video AI, and Adobe After Effects. Last but not least unexplained frequent Arc GPU fan speed increases are being looked at.

For further analysis of how Starfield runs on a wide range of graphics cards, covering samples launched between 2017 to the present day, check out our Starfield PC Performance guide.

We're accustomed to seeing driver updates that increase GPU performance through game optimizations made at the driver level. This has been especially true with Intel's Arc discrete GPUs, where the company has gained gobs of performance from its latest DX9 and DX11 driver enhancements — allowing GPUs like the Arc A750 to sit in our list of Best GPUs currently.

Update: Intel released the following statement to us and others: "In a recent driver update, we changed the reported graphics clock of the A380. Actual performance and frequency were not affected and we are working on an update to revert the change in a future driver update." The following text has been updated to account for the new information.

Intel's latest driver release (version 31.0.101.4644) changed the reported base clock for the Arc A380. According to a Neowin forum post by Eternal Tempest, the new driver update is bundled with a hidden firmware update for the Arc A380 that boosts GPU clock speeds by 150MHz, going from a flat 2000MHz clock speed by default to 2150MHz with the firmware update. However, it doesn't appear that there was actually any firmware change from what Intel is now saying.

A 150MHz clock speed upgrade is no joke and would be a substantial jump from a mere firmware update. In the world of modern GPU overclocking, getting a stable 150MHz core offset would be a decent result on any of Nvidia's recent GPUs. Most Nvidia GPUs usually top out anywhere between 100MHz to 200MHz on the overclock front, depending on GPU temperatures and silicon quality. Getting a similar increase while remaining within spec would be excellent. But there's plenty of room for questioning what's going on.

We're not sure what prompted Intel to make the 150MHz clock speed change, possibly just an inadvertent modification of the code. Intel has already said that it plans to revert the change in a future driver. But the Intel "Game Clocks" aren't even that meaningful to begin with. The base clock is what Intel guarantees the GPU to run at as a minimum, while the "Game Clock" is a typical (conservative) estimate of where the GPU will land.

The Arc A380 officially launched with a 2000 MHz base clock, with a 2050 MHz Game clock. The boost clock (maximum clock) meanwhile is 2400 MHz. The Gunnir Arc A380 card we used for testing bumps all of those by 50 MHz higher, thanks to a factory overclock — 2050 MHz base, 2100 MHz game clock, 2450 MHz boost clock. Here our the results of our updated testing, showing average clocks under 15 games at three different test settings:

Across the fifteen tested games, virtually every game at every setting ended up pegging the GPU at the 2450 MHz maximum boost clock. The lowest average GPU clock for any test run was 2445 MHz, meaning there were periodic dips from the maximum clock, but not anything you'd really notice in actual gaming.

So, it would be nice to get a free performance upgrade on an existing GPU, no matter where it comes from. But while Intel may have inadvertently changed the reported base clock and game clock for the A380, the actual real-world clocks remain the same. The cards will generally run at their rated maximum boost clock in our experience — and that applies to all Arc cards right now.

Besides the A380 clock speed update, the new driver update also adds game highlights for Madden NFL 24 and Wayfinder. The update also fixes three bugs surrounding a crashing problem in Uncharted: Legacy of Thieves Collection (DX12), a system hang when waking up from sleep mode, and another application crash in Blender 3.6.

Intel Arc Alchemist GPUs landed about a year ago, and now compete among the best graphics cards. Drivers have been a recurring theme with Arc, both good and bad, but Intel has made strides to improve things and close the performance gap. To that end, Intel has released new drivers specifically aiming to improve its DirectX 11 performance.

We spoke with Intel earlier this week, and the gains for now are limited to specific games. To provide some background detail, after the initial Arc launch, Intel found that many of the base driver elements for their GPUs — integrated and discrete — were suboptimal, particularly for Arc. With integrated graphics, you can basically always assume that the GPU is the bottleneck. Anything that could be done to move work back to the CPU to alleviate that bottleneck was a "good thing." Obviously, that same logic doesn't necessarily apply any longer when the GPU becomes ten times faster, as in the case of the Arc A770.

So, since the initial launch, Intel spent a lot of time and effort reworking the drivers. We saw this first in early 2023 when Intel touted gains in DirectX 9 performance. The initial drivers for Arc leveraged Microsoft's DX9 on DX12 tools, which provided compatibility and got things up and running. There was a lot of performance left on the table, however. We're told Intel's driver team worked with the open source DXVK project (DirectX on Vulkan) to optimize the various elements better — something it couldn't do with the closed-source Microsoft tool. Eventually, things reached the state that all the DX9 support (or most of it?) was transitioned to using the new DXVK-optimized path.

We're now getting a similar treatment for DirectX 11 games — not in the DXVK sense, but Intel has worked to build a from the ground-up different code path for DX11 support into its drivers. Since the existing DX11 support already works, even if performance isn't ideal, the focus has been on testing specific games and then "whitelisting" them in the drivers to use the new code path. At present, Intel has whitelisted ten popular DX11 games: Apex Legends, Counter-Strike 2, Destiny 2, DOTA 2, Genshin Impact, GTA Online / GTA V, League of Legends, Middle-Earth: Shadow of War, Overwatch 2, and Valorant.

In short, the latest drivers optimize how the graphics hardware interacts with the DX11 API. By fine-tuning various parameters and implementing more efficient rendering techniques (it didn't go into any low-level details), Intel improved overall performance on DirectX 11 games. Here are the results of its internal performance testing, comparing its latest drivers (internal versions 4571 and 4642, though version 4644 drivers are now out) with the original launch drivers (3490).

Across the ten games (eleven if you want to count GTA Online and GTA V as separate entries), the new DX11 code path boosted performance by anywhere from 5% to 33%. But what's particularly interesting is that Intel specifically targeted the driver optimizations at "mainstream" CPUs — it tested with a Core i5-13400F and an Arc A750.

Later in the slides, Intel notes that the gains with a top-end Core i9-13900K weren't as pronounced. The performance increased by only 0% to 28% across the selected games, with an overall average improvement of 12%, compared to the 19% seen with the Core i5 CPU.

We applaud Intel's efforts and testing, as this makes a lot of sense for potential buyers of the Arc A750. With Arc A750 cards now starting at $199, they're a great bargain, providing performance roughly equivalent to the RTX 3060 and RX 6700 10GB. Both of those still tend to sell in the $270 range (while supplies last on the 3060), meaning the A750 provides great bang for the buck. At the same time, while we might standardize all of our testing on a Core i9-13900K CPU to eliminate other bottlenecks as much as possible, people building a PC with an A750 aren't likely to go much above $200 for the CPU, and that's where the i5-13400F sits.

Intel had some other interesting news and bullet points to discuss. The major one is that PresentMon — the foundation of Nvidia FrameView, AMD OCAT, and CapFrameX — is getting an update. PresentMon was created by Intel as an open-source performance monitoring tool to capture frametimes and other metrics related to graphics performance. It has seen many updates over the years, and now Intel is joining AMD and Nvidia in offering a more robust solution.

The new PresentMon beta adds a GUI, making it far more user-friendly than the existing version. Intel has also added a bunch of new features, including a robust overlay, histograms, and more. There's also a new "GPU Busy" metric, which shows how much time was spent waiting on the GPU versus waiting on the rest of the system. Minimizing this difference has been a key focus of the new drivers.

Intel continues to work on validating additional games with the new DX11 code path, and they'll roll those into future drivers. Perhaps at some point, the new code will be deemed robust enough that Intel will simply flip a switch, and all games will default to it — and maybe instead have a blacklist where games known to have issues can stick to the older code.

The new drivers and the PresentMon Beta are available for anyone to try. In fact, the DX11 performance improvements discussed here have been present in the past three Intel WHQL driver releases. The full slide deck for Intel's Q3'23 quarterly driver update is below for your reference.

AMD's Instinct MI series products are thought to be among the most potent HPC and AI accelerators when they arrive in Q4, and AMD has officially announced two variants. However, it appears the MI300A and MI300X aren't the only variants AMD has planned for the market — the first signs of a new mysterious MI300C variant have emerged in Linux patches. It's possible that this new variant either marks the arrival of a CPU-only MI300 that will be AMD's first HBM-equipped processor or a specialized cut-down variant that adheres to the U.S. sanctions for exports into the Chinese market. In fact, AMD recently stated that the company is working on optimizing its product line to comply with U.S. export restrictions to export to China, making this a very realistic possibility.

The MI300, touted as the world's first data center APU, features a multi-chiplet design that combines AMD's Zen 4 and CDNA 3 microarchitecture. The chiplet-based design allows AMD to swap out tiles to make different combinations. The basic MI300 design has 13 chiplets, most 3D-stacked, and comprises nine 5nm compute logic chiplets over four 6nm I/Ochiplets.

Japanese outlet Coelacanth Dream has discovered a third and unannounced variant of the MI300 in Linux's EDAC (Error Detection And Correction) driver. The Linux code suggests that the MI300C could have at least one CPU chiplet onboard. However, the type of memory and number of memory channels remain a mystery.

First, it's important to understand the configuration of the two existing models: The MI300A is the vanilla CPU+GPU combo with three CCDs and six XCDs. This configuration amounts to twenty-four Zen 4 cores and 228 CDNA 3 compute units. There are also eight HBM3 stacks for 128GB of shared memory between the CPU and GPU. In contrast, the MI300X is the pure GPU part with zero CCDs and eight XCDs, equivalent to 304 CDNA 3 compute units. The eight-stacked HBM3 configuration remains intact, but the higher density stacks give it 192GB compared to the 128GB layout on the MI300A.

We have two theories as to what the MI300C could be. Our first and most likely theory is that the MI300C might be a CPU-only variant. AMD never clarified what the suffixes in the MI300 branding stood for. However, if we use our imagination, the "A" in MI300A could represent "APU." It makes sense since the MI300A is synonymous with today's AMD APU with graphics. Assuming that is some accuracy to the speculation, the "C" in the MI300C may symbolize "CPU," implying that only Zen 4 cores are present. If accurate, that would turn the MI300C into AMD's first CPU with HBM3 memory and turn into a potential rival for Intel's Sapphire Rapids HBM offerings. Rationally, AMD would need more than 24 cores to rival Sapphire Rapids HBM, which spans up to 56 cores. AMD could maximize the core count on the MI300C to hit 96 Zen 4 cores, essentially turning it into a 4th Generation EPYC Genoa chip with HBM3 memory.

Our second theory is that the MI300C could be a China-exclusive SKU. AMD recently stated that the company is working on optimizing its product line to comply with U.S. export restrictions to export to China. We all know how AI is currently flourishing, and the MI300, with some minor modifications, could thrive in the Chinese AI market that's continuously showing a high demand for AI GPUs and accelerators.

Finally, the MI300C could just represent AMD swapping in an additional CPU tile and removing a GPU tile, thus creating a different CPU-to-GPU ratio to tailor the chip for different operating environments. 

AMD confirmed in the company's recent earnings call that it has already started sampling the MI300A and MI300X to its clients, and the accelerators are on track to hit the market in Q4. The MI300C shouldn't be too far behind, so we'll likely get some insight into the mysterious variant sooner than later.

AMD's Radeon Pro W7600 and W7500 (via VideoCardz) are seemingly on our doorsteps. The duo of Navi 33-based workstation graphics cards won't vie for a spot on the list of best graphics cards for gaming. However, they'll offer professional users the power of RDNA 3 with an entry-level (for workstations, anyway) price tag.

According to the leak, the Radeon Pro W7600, like its mainstream Radeon RX 7600 counterpart, features the full Navi 33 die with 32 compute units and 8GB of GDDR6 memory across a 128-bit interface. The Radeon Pro W7600 delivers a peak single-precision (FP32) performance of 20 TFLOPS, slightly lower than the Radeon RX 7600's 21.75 TFLOPs. The 18 Gbps GDDR6 enables the Radeon Pro W7600 to offer an identical memory bandwidth of 288 GB/s as the consumer-focused Radeon RX 7600. However, AMD also restricted the Radeon Pro W7600 to a TBP limit of 130W, 35W beneath the Radeon RX 7600. As a result, the Pro variant makes do with a single 6-pin PCIe power connector, unlike the mainstream version that demands an 8-pin PCIe power connector.

The Radeon Pro W7500 meanwhile comes with trimmed down Navi 33 silicon with only 28 compute units out of the maximum 32. More importantly, AMD limits the GPU to just a 70W TBP. The reduced configuration has the Radeon Pro W7500 offering just 12 TFLOPs of FP32 performance. The bright side is that the graphics card doesn't rely on any external power connectors and still has a similar memory configuration to the Radeon Pro W7600, except with slower GDDR6 memory. AMD has slapped 11 Gbps GDDR6 on the Radeon Pro W7500, landing the memory bandwidth to 176 GB/s, 39% lower than the Radeon Pro W7600.

Radeon Pro W7600 and Radeon Pro W7500 Specifications

Compared to the existing Radeon Pro W6600, the Radeon Pro W7600 provides double the FP32 performance and 29% higher memory bandwidth at the same 130W TBP. In contrast, the Radeon Pro W7500 brings 20% more FP32 performance at a 46% lower TBP. However, the biggest drawback is the debilitated memory system on the Radeon Pro W7500. The graphics card's memory bandwidth is 21% lower than the Radeon Pro W6600.

Geared toward the professional and workstation markets, the Radeon Pro W7600 and W7500 provide four DisplayPort 2.1 outputs with UHBR 10 support. It's a massive upgrade over the Radeon Pro W6600's DisplayPort 1.4a output with only HBR 3 support. Furthermore, the Navi 33-powered graphics cards arrive with AV1 hardware encoding, a feature not present on the Radeon Pro W6600.

Pricing-wise, the Radeon Pro W7600 will hit the market at $599, $50 less than the Radeon Pro W6600 while offering higher performance. As per AMD's price comparison, the Radeon Pro W7600, which outpaces the RTX A2000, is $47 cheaper. Meanwhile, the Radeon Pro W7500 will retail for $429, representing only a $5 saving over the T1000.

AMD will reportedly announce the Radeon Pro W7600 and W7500 tomorrow. And while the leaked slides divulged the specifications and pricing, it's unknown whether the two Navi 33 professional graphics cards will be available tomorrow or won't arrive until sometime later this year.

A week after the game's launch, AMD has released a hotfix Adrenalin update just for Ratchet & Clank: Rift Apart that fixes the game's notorious application crashes and driver timeout issues. The driver update is known as Adrenalin Edition 23.10.23.03 for Ratchet & Clank™: Rift Apart, and can be downloaded at that link.

Ratchet and Clank: Rift Apart launched on PC with broken driver support from AMD. The game reportedly suffered from driver timeout failures on top of game crashes on the RX 7900 XTX and some other Radeon GPUs when AMD's dynamic resolution scaling or the game's ray tracing effects were enabled.

Thankfully, the game was apparently in a playable state without any of these features enabled, but it's not something you want to see on launch day when everyone wants to check out what the game can do and how good the game can look. At least now Radeon gamers no longer have to put up with these issues and can crank up the game's RT effects without stability issues.

On top of this, the team behind Ratchet and Clank: Rift Apart released another major update yesterday fixing more of the game's graphical issues — these were separate from the AMD driver problems, including low-resolution texture streaming issues, issues with water reflections, additional optimizations, and more. This should make the game even smoother and more playable than it was before, on top of the items fixed by AMD's new driver.

As previously mentioned, this was a hotfix (i.e. emergency) driver update for Ratchet and Clank: Rift Apart specifically, so the update does not come with any additional updates or bug fixes from 23.7.2. If you are not playing Ratchet and Clank: Rift Apart, this driver should behave exactly like 23.7.2. That means some of the other reported problems with the 23.7.2 drivers, including poor performance on certain mobile solutions, still remain.

The internet has often blamed AMD for bad graphics drivers, but according to @carygolomb on Twitter, that supposed reputation might be returning to haunt the company. The Twitter user reported that updating to Adrenalin driver 23.7.2 caused GPU performance to tank on his handheld gaming device and re-branded his Radeon 780M integrated GPU to the much slower 760M. On top of this, the user also reported issues with some internal and external devices no longer functioning with the new driver active, including his fingerprint reader, sound card, and Xbox 360 game controller.

To back up the user's claims, Digital Foundry replied to his Tweet confirming the bad performance problems with the new driver. According to the review outlet, one of its Aya Neo handheld gaming consoles sporting a Ryzen 7 7840U lost 33% of its gaming performance while running the new driver, going from 33 FPS to an unplayable 22 FPS in Metro Exodus: Enhanced Edition.

These issues are among the worst AMD-related graphics driver issues we've come across in a long time. Usually, AMD's GPU drivers are very reliable and don't have significant problems. But, with the dozens and dozens of driver updates that AMD pumps out year after year, a dud was bound to pop out at some point.

On top of these issues, the official list of known problems with 23.7.2 is also somewhat high for a fully-released WHQL Adrenalin driver. Some of these additional problems include high idle power consumption on RX 7000 series GPUs — which was supposed to be fixed by now, performance metrics overlay not fully working in some games, and the widely known Rachet and Clank: Rift Apart ray-tracing crashes that we talked about two days ago.

With how new the driver is, it is impossible to tell how widespread the performance issues are. But for now, it seems to be heavily affecting AMD's mobile Ryzen 7000 series processors, with nobody talking about these bugs with desktop Radeon graphics cards. Modern RX 7000 series GPUs seem unaffected, with Ancient Gameplays reporting slightly better gaming performance in some titles compared to the last driver.

AMD has released the first graphics driver software package offering support for its popular code-name Phoenix mobile APUs. This range of AMD Ryzen processors is particularly attractive for compact form factors as it offers some of the best iGPUs in the industry (the RDNA 3 architecture Radeon 700M series), plus the benefit of the newest Zen 4 CPU cores. Until today, with the release of AMD Software: Adrenalin Edition 23.7.2, folk buying PCs with these AMD Ryzen 7040 series processors had to rely on a system vendor-supplied driver.

Phoenix Support, and an Unannounced Processor

PC makers are often quite slow at delivering graphics driver updates when it is their responsibility, so it is good for consumers that Phoenix chips are now within the scope of the freely available Adrenalin driver. As of July 2023 there are a growing number of laptops, mini-PCs, and PC gaming handhelds featuring AMD Phoenix.

The new driver folds in support for the Ryzen 7045 Dragon Range family, as well as other Ryzen 7000 chips on mobile and desktop.

A couple of interesting support exceptions have arisen, though. The Asus ROG Ally exclusive Ryzen Z1 series isn't mentioned in the new Adrenalin driver. But curiously, there is mention of an unannounced chip that has gained support — the Ryzen 7 7840S.

Fixed and Known Issues

There is a typical array of fixed and known issues with Adrenalin Edition 23.7.2, with some games and specific graphics card models mentioned. The fix that will likely have most widespread appreciation is for Call of Duty: Modern Warfare II, where stuttering when Radeon Anti-Lag is enabled should now be cured. A close second may be the fix for audio sync issues when using AVC or HEVC video recording with your Radeon hardware.

As for known issues, the widely reported high idle power issues remain in some multi-display setups with mixed high-resolution and high refresh rate displays, when using a Radeon RX 7000 GPU.

Last but not least, Tom's Hardware regulars may have read our Ratchet and Clank: Rift Apart with DirectStorage article yesterday. Released today, this game is also going to be stuffed with Nvidia supported tech like Nvidia DLSS 3, Reflex, RTX IO, and DLAA. It supports ray tracing on Nvidia GPUs, with this graphical splendor applied to shadows, ambient occlusion, and reflections. Sadly, AMD's new driver released on the same day, lists Ratchet and Clank: Rift Apart as being somewhat unstable with RT on Radeon graphics.

Specifically, AMD says that "Application crash or driver timeout may be observed while playing Ratchet & Clank: Rift Apart with Ray-Tracing and Dynamic Resolution Scaling enabled on some AMD Graphics Products, such as the Radeon RX 7900 XTX." AMD seems to have its driver team focused upon this wrinkle, as it adds that it is working with the game's developers on solutions.

Nvidia has just released a new Game Ready driver, headlined as delivering support for a few new games. Specifically, there's Portal: Prelude RTX, which has RTX IO support (a variant of DirectStorage that we're told works on non-Nvidia GPUs as well), and Ratchet & Clank: Rift Apart, which also has RTX IO suport. Alongside the games, this is also the first driver with support for the GeForce RTX 4060 Ti 16GB custom cards from AIBs. However, you have to dig through the release notes (PDF) to find out that this is the long-awaited driver release which finally addresses DPC latency.

Last month we reported that Nvidia's software quality assurance spokesperson Manuel Guzman signaled a DPC latency fix had been developed and that it was just going through QA before it could be rolled out. Deferred Procedure Calls (DPC) in Windows were designed for high-priority tasks to defer required but lower-priority tasks for later execution. If a driver's DPC routines aren't written correctly, the user can see symptoms manifest as stutters, skips, and similar glitches in video and audio. Users who aren't sure if they are affected can run the Latencymon tool, to check their numbers; smaller values are better.

In Nvidia's 536.67 WHQL drivers, the release notes say that it fixed two bugs, one where applying Freestyle filters caused games to crash, and another one to address DPC latency. As you can see in the screenshot, we don't get a lot of information from Nvidia about the fix, only that using the new driver should eradicate the "Increase in DPC latency observed in Latencymon for Ampere-based GPUs [3952556]." That reference number points back to the raising of this issue among the 'Known Issues' last year.

Last month we pointed out that the value of Nvidia's fix can only be demonstrated by a mass of users testing and being happy with it. Signs so far (just a few hours after release) aren't 100% positive. We note that Manuel Guzman has been chatting with two users who don't feel the benefit of the new driver. At the time of writing, it looks like more independent testing needs to be done to verify the fix, and perhaps more work to truly fix the root cause may be needed.

Nvidia 536.67 Driver Notes

In this driver, Nvidia says that the following new games are now optimized for GeForce users:

• Portal: Prelude RTX, a free, fully ray traced Portal mod with path-traced graphics, DLSS 3, Reflex, and RTX IO
• Ratchet & Clank: Rift Apart with DLSS 3, advanced ray tracing, Reflex, and RTX IO
• Remnant II from Gunfire Games has DLSS 2

Additionally, a trio of new G-Sync Compatible displays has been validated, including the very tasty 48-inch Panasonic 48MZ1800 OLED 4K monitor. Seven new games have GeForce Experience Optimal Settings configurations, including BattleBit Remastered.

Last but not least, the new driver supports the new 16GB models of the GeForce RTX 4060 Ti from Nvidia AIBs. You can read more about this new driver via the official release notes or GeForce blog (linked top), and actual driver downloads can be grabbed via the GeForce site, or alternatively use GeForce Experience to auto-download and apply this new driver.

Raspberry Pi Pico handhelds might not be new anymore, but this one, developed by a maker known as Phlash Thunderstorm over at YouTube, definitely kicks things up a notch. Phlash Thunderstorm dubs the project the Pico Hero. It’s bigger than modules like the PicoSystem by Pimoroni and is open-source for anyone who wants to make one at home.

According to Phlash Thunderstorm, the goal was to create their own unique take on a Pico-powered gaming handheld. While you can find plenty online, the idea here was to make one with a larger screen, better speaker, and custom form factor with a little artistic flair. The result is this impressive little handheld that can run custom homebrew games and even emulate other systems.

In a demo video shared on YouTube, Phlash Thunderstorm shows off the Pico Hero successfully emulating a modified version of a Sega Mega Drive emulator. The original emulator was created by Bzhxx, who created it to work with ARM Cortex STM32H7 Cortex-M7 MCUs, which runs at 280 MHz. Phlash Thunderstorm got this emulator working on the Pico Hero handheld.

The unit is powered by an RP2040 Plus module which has the form factor of the regular Pico with a few upgrades. The screen is fairly large for a handheld this size, measuring 3.2-inches with a resolution of 320 x 240px, connecting using a parallel interface for improved performance. The Pico Hero has three buttons, a power switch, and an analog control stick. A 3-watt speaker is used for audio output, and a microSD card stores all the necessary data. Everything is held together with a beautiful, customized purple PCB made just for the project.

The software for the project was shared on the official project page over at GitHub. Phlash Thunderstorm explains that the library is Arduino-based, and developing with it is as easy as dragging and dropping it into the library's folder. If you want to get a closer look at this Raspberry Pi project, check it out in action on YouTube and follow Phlash Thunderstorm for future updates.

Phoronix reports that Intel engineers are already building SoundWire Linux driver support for its upcoming Lunar Lake CPU architecture. Specifically, this update brings with it initial SoundWire Intel ACE2.x support, which will be part of Lunar Lake's future audio capabilities. Intel is developing this new driver well ahead of schedule, with Lunar Lake expected to arrive in 2024.

According to Phoronix, Linux 6.5 will support SoundWire's new ACE2.x generation IP block, which is probably why Intel's engineers are building the new audio drivers right now. Not to mention the fact that Lunar Lake is just around the corner, and is slated to arrive as early as 2024. We don't know much about this new SoundWire revision, but as with all new audio technologies, expect better audio processing and audio quality compared to current SoundWire revisions.

SoundWire is an interface that was ratified in 2015 to transport audio data to integrated audio devices. Basically, it is an "audio operating system" that governs how data is transported to audio devices inside speakers, phones, laptops, automobiles, and other supported audio sources. According to Phoronix, Intel has supported SoundWire for a long time, which explains its early adoption of the latest SoundWire IP blocks.

This is also not the first Linux audio driver Intel has published for Lunar Lake either. Intel has also been busy adding two more audio drivers to Linux supporting Lunar Lake CPUs, including Sound Open Firmware 2.6 and MIDI 2.0 support.

Lunar Lake will be the successor to Meteor Lake, reportedly bringing back the performance per watt crown to Intel. Lunar Lake will share the same multi-tiled technology as Meteor Lake, sporting a CPU tile, graphics tile, SoC tile, and I/O tile, but will be upgraded to Intel's 18A manufacturing process that will offer 10% better efficiency over Meteor Lake's 20A node.

Originally, Lunar Lake was supposed to launch in 2025, but development has gone so well with the 18A process that Intel accelerated its timetable by a full year to 2024.

AMD has released a new graphics driver which fixes high idle power issues observed by Radeon RX 7000 users. AMD Software: Adrenalin Edition 23.7.1, which supports graphics cards from the Radeon RX 400 Series and newer, addresses an unusual high-power usage case in Windows, but at this stage we aren't sure if it has covered all incidences.

We first reported on the headlining high idle power on RDNA 3 cards issue last December. In our story about a fix for RX 7900 XT(X) power consumption - particularly addressing video playback efficiency on these powerful cards - we welcomed the driver update. However, that driver release came with AMD's first admission that “high idle power has situationally been observed when using select high resolution and high refresh rate displays.” It turns out his bug affects the whole RX 7000 family, which has only recently grown beyond the high-end with the release of the Radeon RX 7600.

The release notes features a bullet point which provides the scope of the idle power fix provided by the Adrenalin Edition 23.7.1 driver. We are sure this fix will be warmly welcomed by users of AMD's newest architecture, as it should reduce the amount of wasted electricity.

Other RDNA 3 architecture bug fixes in this latest Radeon driver include one which causes suboptimal performance or occasional stuttering in VR games and applications. An RX 7900 XTX specific fix has also been implemented to eliminate intermittent corruption when playing WWE 2K23.

The last two fixes are for a wider audience. A crashing and driver timeout issue which annoyed DaVinci Resolve Studio users has been quashed. Lastly, intermittent corruption observed when switching windows while playing Nioh 2 has been fixed. AMD mentions this Nioh 2 bug affected multiple Radeon products including the RX 6800 XT.

There is only one new feature mentioned by the official Adrenalin Edition 23.7.1 release notes. AMD has implemented support for additional Vulkan extensions. If any of your favorite games / apps use the Vulkan API you may appreciate these additions.

AMD's list of remaining known issues outguns the list of fixes, but at least we know the red team developers are working on these specific annoying wrinkles.

Lastly, we must note that AMD has temporarily disabled the 'Factory Reset' driver install feature. Back in March, we reported on the bricked Windows installs that this driver install option was precipitating. There is a workaround for now; AMD recommends users run its AMD Cleanup Utility if they want / need to start afresh.

Nvidia is on the verge of releasing a fix for its ongoing driver DPC latency issues. In a Twitter post, Nvidia's software quality assurance spokesperson Manuel Guzman said a fix is ready. However, the update release still had to go through the standard QA testing procedure before distribution.

While Nvidia has gone to great lengths to minimize input latency and improve responsiveness with projects like Nvidia Reflex, users of its drivers have been suffering from DPC latency issues for several GPU generations.

If you are blissfully unaware of DPC latency, it is system latency introduced by a Deferred Procedure Call (DPC) in Windows. DPCs were designed for high-priority tasks to defer required but lower-priority tasks for later execution. However, DPC buffers can get filled, resulting in stutters, skips, and similar glitches in video and audio. These jarring latency symptoms are highly undesirable in gaming, viewing videos, and creating or listening to music.

People who suspect they have DPC latency issues can download various tools to check and help minimize the issues. However, for a proper fix, you will want a device driver that has been coded to minimize excessive latencies in DPCs. If you have a laptop or pre-built PC, NotebookCheck maintains a list of systems and their tested DPC latencies - lower numbers are better.

According to Mr. Guzman, Nvidia has now reached a stage where it has fixed its driver DPC latency issues. Of course, the proof of the pudding is in the eating, and we shall see shortly when the fixed driver(s) arrive. As the Nvidia software QA spokesperson says, please "stay tuned."

GPU Monitoring Apps and Updating Nvidia Drivers

From Guzman's Twitter thread, where the DPC latency news emanated, there comes a solid recommendation to exit all your GPU monitoring apps before updating Nvidia drivers. The crux of this potential issue is that if you forget or neglect to follow this advice, "the GPU monitor utility may inadvertently set a lower power target resulting in lower performance."

Nvidia Gamers Using Windows 11

Mr. Guzman also recommends that Windows 11 users with Nvidia GPUs update to the latest version. Apparently, Microsoft's latest update addresses Timeout Detection and Recovery (TDR) errors which could otherwise interrupt gaming sessions.

Phoronix reports that Meteor Lake-S driver code is still being developed and implemented into the Linux kernel, with a new driver update going live this morning. This comes as part of what appears to be a dual-pronged desktop CPU lineup, as Raptor Lake Refresh has been confirmed as part of Intel's next generation of desktop CPUs. The code suggests that desktop-designed Meteor Lake processors could still be coming soon, with the possibility of it co-existing with the new Raptor Lake refresh.

The driver code Phornix mentions specifically upgrades the Intel pin control driver (pinctrl) with Meteor-Lake S support. These changes were made this morning and will be shipped with Linux 6.5 once it is ready. According to Phoronix, this new driver update compliments other Meteor Lake-S code already in the kernel, confirming that Intel developers are actively working on Meteor Lake S support.

Intel has not said anything about desktop Meteor Lake-S processors since Raptor Lake Refresh was confirmed by themselves in China. But these Linux updates seem to nullify any rumors pointing to a Meteor Lake S cancellation. If Meteor Lake S ends up launching soon (as the evidence shows), desktop Meteor Lake processors could co-exist with Raptor Lake CPUs in the same market, filling gaps in Intel's 14th Gen lineup that are better suited to Intel's more advanced Meteor Lake architecture.

Raptor Lake refresh and Meteor Lake couldn't be any more different from each other; the former is based on Intel's current 13th Gen Raptor Lake CPU architecture, while Meteor Lake is an all-new CPU architecture featuring several new technologies, including new cores, tiles, AI-acceleration and more. Raptor Lake Refresh, on the other hand, is a second-generation refresh of Intel's Alder Lake architecture that will probably feature higher clock speeds, different core configurations, and an upgraded memory controller compared to Raptor Lake if rumors are true.

Of course, it's impossible to say when these chips will come to market, but Intel's roadmap demands that a Meteor Lake-S lineup will have to launch during the Raptor Lake Refresh life-cycle and before Arrow Lake arrives next year.

Intel graphics card users who prefer the Linux OS have a significant performance boost to look forward to. On Friday, the latest release of the Mesa (v23.2) code included a performance optimization that has precipitated some very worthwhile games frame rate boosts. For example, using the latest code on an Intel Arc A770 graphics card in Linux saw Counter-Strike: Global Offensive running 11% faster.

Intel open-source Linux graphics driver engineer Francisco Jerez discussed the intel/gfx12.5 code change on GitLab. According to Jerez, some performance features which were expected to be enabled by default, had been accidentally disabled by the kernel. 

Confounding code detectives, the "Compressible Partial Write Merge Enable", "Coherent Partial Write Merge Enable" and"Cross-Tile Partial Write Merge Enable" bits all appeared to be enabled when a query was run on an idle system. Nevertheless, these L3 partial write merging features were "getting clobbered during 3D context initialization by an i915 workaround," and causing a "serious performance bottleneck."

This is an important performance sapping code error to catch, as it affects all DG2 (Arc Alchemist) graphics cards in Linux. The benefits will also be felt by upcoming users of Intel Meteor Lake integrated graphics.

As well as the 11% performance benefit mentioned above (in classic FPS CS:GO), Jerez also noted that Shadow Of the Tomb Raider improves by ~5.5%, and AztecRuins-VK improves by ~6.5% on a DG2-512 sample (Arc A770). No, performance penalties from the new code have been observed thus far.

Lastly, Jerez commented on the new Mesa code helping narrow the Windows / Linux performance gap. We note that Intel's Arc cards will probably see some more big gains like this on both Linux and Windows, due to relatively less mature driver software and the underlying raw performance figures (TFLOPs etc) indicating there remains untapped potential.

It is expected that the Mesa 23.2 code improvements will get to stable Linux OS distros sometime around late August or September.

New Zealand-based cyber security experts at CyberCX have detailed and demonstrated an alarmingly simple way to consistently access older BIOS-locked laptops. In the linked blog post, and video demo, an exec at the firm detailed how to short some EEPROM chip pins with a simple screwdriver to access a fully-unlocked BIOS. Then all it took was a quick poke around the BIOS settings screen to disable any BIOS password altogether.

Before we go further, it is worth pointing out that CyberCX's BIOS password bypass demonstration was done on several Lenovo laptops that it had retired from service. The blog shows that the easily reproducible bypass is viable on the Lenovo ThinkPad L440 (launched Q4 2013) and the Lenovo ThinkPad X230 (launched Q3 2012). Other laptop and desktop models and brands that have a separate EEPROM chip where passwords are stored may be similarly vulnerable.

Some of you will know that there are quite a lot of good used laptops sold for spares as they are practically disabled from re-use due to a BIOS lock in place. Individual owners or organizations may have neglected to document these passwords, forgotten them, or whatever - but the systems, which are also usually stripped of their best hard drive and/or SSDs, aren't as valuable on the used market as they could be. CyberCX pondered over its old IT equipment that had been password locked and wondered about trying to gain full access to the hardware again to use it as spares or testing machines.

From reading various documentation and research articles, CyberCX knew that it needed to follow the following process on its BIOS-locked Lenovo laptops:

• Locate the correct EEPROM chip.
• Locate the SCL and SDA pins.
• Short the SCL and SDA pins at the right time.

Checking likely looking chips on the mainboard and looking up series numbers eventually lead to being able to target the correct EEPROM. In the case of the ThinkPad L440, the chip is marked L08-1 X (this may not always be the case).

An embedded video in the CyberCX blog post shows just how easy this 'hack' is to do. Shorting the L08-1 X chip pins requires something as simple as a screwdriver tip being held between two of the chip legs. Then, once you enter the BIOS, you should find that all configuration options are open to be changed. There is said to be some timing needed, but the timing isn't so tight, so there is some latitude. You can watch the video for a bit of 'technique.'

CyberCX includes some quite in-depth analysis of how its BIOS hack works and explains that you can't just short the EEPROM chips straight away as you turn the machine on (hence the need for timing).

Some readers may be wondering about their own laptops or BIOS-locked machines they have seen on eBay and so on. CyberCX says that some modern machines with the BIOS and EEPROM packages in one Surface Mount Device (SMD) would be more difficult to hack in this way, requiring an "off-chip attack." The cyber security firm also says that some motherboard and system makers do indeed already use an integrated SMD. Those particularly worried about their data, rather than their system, should implement "full disk encryption [to] prevent an attacker from obtaining data from the laptop’s drive," says the security outfit.

CyberCX hints it will continue the above research. Perhaps it will look into being able to read the BIOS password in plaintext from the EEPROM, or check the viability of its screwdriver hack across more machines.

We showed benchmarks of the PNY GeForce RTX 4070 Verto Dual in our initial Nvidia RTX 4070 Founders Edition launch review. Then a bunch of other stuff came along, like the RTX 4060 Ti and RX 7600 launches, but now we're getting around to various third-party reviews. These are some of the best graphics cards, for gaming and other purposes, and the RTX 4070 looks to hit the sweet spot in terms of specs and pricing.

At its core, the RTX 4070 performs a lot like the previous generation RTX 3080, despite having 2GB more VRAM but a much narrower 192-bit memory interface. That's thanks to the Nvidia Ada Lovelace architecture, and specifically the inclusion of up to 8MB of L2 cache per 32-bit chunk of memory interface. The additional memory can help in a few instances, but higher resolutions also tend to prefer the increased raw bandwidth of the 3080's 320-bit interface.

Let's start with the usual specs table before getting into the specifics of the PNY card.

The above specs table shows the ten GPUs we'll be using for this review. The PNY RTX 4070 Verto Dual and Nvidia RTX 4070 Founders Edition have identical specs, since both use the reference clocks. There's not much more we can add.

The RTX 4070 lands at a $599.99 MSRP, and there are even cards periodically available for less than that — like right now, you can pick up the RTX 4070 from Asus, Gigabyte, MSI, or Zotac with a $50 instant rebate using code "ZIPDADS" (for Fathers' Day, if that wasn't obvious). PNY's card unfortunately isn't part of that deal. It's also "while funds last," so you may or may not be able to take advantage of that offer.

That sort of deal does speak volumes about graphics card sales in general. The past several Black Fridays and Prime Days, for example, had very few if any noteworthy GPU deals, but that's a straight $50 discount from Newegg on a graphics card that released less than two months ago. It's a safe bet that Nvidia could also drop the prices on all of its GPUs $50 or so without incurring too much pain.

Short-term sales aside, we can't help but feel the RTX 4070 should be selling for a lot less than its MSRP. It's faster than the previous generation RTX 3070, but it also comes with a substantially smaller die size. 294.5 mm^2 on TSMC 4N compared to 392.5 mm^2 on Samsung 8N likely means the AD104 chips do cost more, but Nvidia should still get about 190 chips per wafer. Even at $19,000 per 4N wafer, that's only $100 per chip (and Nvidia probably pays less than that due to the number of wafers it orders).

PNY GeForce RTX 4070 Verto Dual

PNY uses a nondescript box for its base model RTX 4070. That's fine, as most people are far more interested in the performance and graphics card than in the packaging. Inside the box sits the card in an anti-static bag... and that's pretty much it.

There's no support stand here, which, considering the card weighs 687g, shouldn't be a problem. The card's dimensions are 247x120x40 mm. Nvidia's RTX 4070 Founders Edition by way of comparison measures 244x111x40, so it's just a touch shorter in both length and height, though it's quite a bit heavier at 1021g.

PNY uses a dual-fan configuration, as implied by the model name. Each fan is 85mm in diameter, so smaller than some cards. More importantly, there's no integrated rim to improve static pressure and airflow. That gives Nvidia's reference card a double advantage: Its 91mm fans have more area, and they include a rim around the outside diameter.

The radiator also carries this same approach of not going hog-wild on cooling capacity. There are plenty of heatsink fins, but only three heatpipes running through them. The RTX 4070 only has a 200W TGP (Total Graphics Power), so the cooling should be sufficient, but as we'll see later, this modest setup ultimately results in higher fan speeds and noise levels.

As with most recent Nvidia GPUs, you get three DisplayPort 1.4a outputs and a single HDMI 2.1 port. With DSC (Display Stream Compression), you can get "visually lossless" 4K at 240Hz. Some might quibble about how good the image truly looks, but having tested it, we don't have any complaints — and most of the time, at 4K you won't be pushing framerates anywhere near the 240 Hz maximum refresh rate.

PNY RTX 4070 Test Setup

Our graphics card test system was updated earlier this year to use Intel's fastest Core i9-13900K, with all the bells and whistles. We tested the PNY RTX 4070 card using Nvidia's 531.42 launch drivers, though as far as we can tell nothing substantial has changed with the latest 535.98 that came out on May 30. All of the cards in this review have been retested in the past month or two, without Windows 11's Virtualization Based Security enabled.

Our initial review of the RTX 4070 looked at performance across all resolutions and settings that we normally test (1080p medium, 1080p ultra, 1440p ultra, and 4K ultra). It also had professional and AI benchmarks, and performance results with upscaling including DLSS 3 (see how to enable DLSS 3). For third-party cards like this PNY RTX 4070, we'll limit our testing to 1080p/1440p/4K ultra, without looking at the other workloads — everything should be relatively similar margins as the gaming results.

Our test PC includes an MSI Z790 DDR5 motherboard, G.Skill DDR5 memory, and a Sabrent 4TB SSD — which we promptly filled to about half its total capacity. be quiet! also supplied us with its new 1600W Dark Power Pro 13 80 Plus Titanium rated power supply. That would have been overkill, back before cards like the RTX 4090 were a thing.

We're including a collection of current and previous generation GPUs, with the Nvidia results all from the RTX 40-series. If you want to see how the RTX 4070 stacks up against other GPUs, check out our GPU benchmarks hierarchy.

We have Nvidia PCAT v2 (Power Capture and Analysis Tool) hardware as well, which means we can grab real power use, GPU clocks, and more during all of our gaming benchmarks. We'll have most of the details for power testing in a few pages.

PNY RTX 4070 Overclocking

Our standard approach to overclocking graphics cards is to use MSI Afterburner and then max out the power slider, but the PNY card is locked to 100%. So we left that alone and played with the GPU and GDDR6X clocks. We ultimately achieved a stable (for our benchmark suite) +200 MHz GPU and +1750 MHz GDDR6X overclock. The memory overclock suggests that this is 24 Gbps Micron memory, even though it's marked as standard 21 Gbps memory.

The thing about the overclock is that, without an increased power limit, it's not as effective as with other GPUs. We've got overclocked results from the 4070 Founders Edition in the charts as well (+225 MHz GPU, +1350 MHz GDDR6X), and the increased power limit seems to be the deciding factor.

Regardless, there's a pretty large gap between the RTX 4070 and the RTX 4070 Ti. With 5888 CUDA cores versus 7680 cores, you'd need to run the RTX 4070 at about 3.4 GHz just to match the reference 4070 Ti. That's not going to happen, possibly not even with liquid nitrogen. All told, our overclock on the PNY RTX 4070 yielded a 6% improvement in performance, while the RTX 4070 Ti ends up being around 25% faster than the stock 4070.

• MORE: Best Graphics Cards
• MORE: GPU Benchmarks and Hierarchy
• MORE: All Graphics Content

We have 1080p, 1440p, and 4K ultra benchmarks for this PNY review. Because it has a $599 MSRP, we feel that starting with 1440p ultra is most appropriate. We'll cover the 1080p results next, and finish with 4K gaming results (which are mostly academic, as you'd need DLSS to get good performance in many games at 4K ultra).

Here's the overview, first with the 15-game full suite, and then with separate rasterization-only and DXR-only (i.e. ray tracing) charts. We'll discuss the individual results further down the page.

There's very little to add to the RTX 4070 story when looking at the PNY card. It's effectively tied with the reference RTX 4070 Founders Edition, with all the results being within the expected margin of error.

Against other GPUs, the RTX 4070 and RX 6950 XT trade blows, with the 4070 leading in ray tracing games and mostly trailing in rasterization performance (Total War: Warhammer III being the sole exception on the latter). DLSS upscaling will generally favor Nvidia's new card, because it's both more widely available and it looks better, with Frame Generation being the potential smooth (as in, frame smoothing) cherry on top.

Across our 15-game test suite, the PNY and Founders Edition 4070 cards are within 2% of each other in every test. We'll need to look to other aspects to find some differentiating factors.

One interesting point is just how much the gap has grown between the various Nvidia GPUs this generation. The 4070 ends up being 33% faster than the 4060 Ti, while the 4070 Ti in turn delivers 24% higher performance. Compared to the previous generation, the RTX 3070 was only about 12% faster than the RTX 3060 Ti, while the RTX 3070 Ti tacked on just 7% more performance than the 3070.

That mostly applies to other 30-series versus 40-series comparisons as well, though the 4080 and 4070 Ti at least are about the same 20~25 percent difference as the previous generation. We'll have to see where the upcoming RTX 4060 and, eventually, RTX 4050 land. The 3060 Ti was 34% faster than the vanilla 3060, despite having less VRAM. On paper at least, the 4060 Ti has 46% more compute than the 4060.

• MORE: Best Graphics Cards
• MORE: GPU Benchmarks and Hierarchy
• MORE: All Graphics Content

The PNY RTX 4070, and RTX 4070 in general, does a bit better at 1080p ultra than at 1440p ultra. That's partly because the large L2 cache helps more at lower resolutions, and partly because CPU bottlenecks also come into play. Overall, the 4070 comes in just a bit ahead of the RX 6950 XT, where it was just a bit behind at 1440p.

We could point out that it's fair to expect any $600 graphics card to easily handle 1080p gaming. After all, even the $350 RX 6750 XT averages over 60 fps, though as usual, AMD's GPUs struggle a bit more in some of the ray tracing games. Likewise, AMD's GPUs do better overall in rasterization games. Anyway, let's move on.

The spread in the individual games, when comparing the PNY and Founders Edition 4070 cards, is meaningless. That's as you'd expect, given they have the same reference clock speeds.

The RX 6950 XT remains the closest competition, with a slightly higher price right now. PNY's card just edges past it, by 1% overall, with a 20% lead in the DXR test suite and a 9% loss in the rasterization games.

What the above performance charts don't show (and we'll get to this in a couple of pages) is that the RTX 4070 basically matches the RX 6950 XT in performance while using 178W on average at 1080p, while the RX 6950 XT uses 307W. That means you need a better PSU if you want to run AMD's previous gen halo card, and that's still not accounting for DLSS upscaling or frame generation.

• MORE: Best Graphics Cards
• MORE: GPU Benchmarks and Hierarchy
• MORE: All Graphics Content

Wrapping up with the 4K ultra results, as you'd expect, the PNY RTX 4070 starts to struggle. Several of the rasterized games still break 60 fps, but only Spider-Man: Miles Morales breaks 30 fps in our DXR tests.

4K definitely needs help, and if you enable DLSS upscaling, performance improves a lot. We could debate about whether or not AI-based upscaling to 4K counts, but unless you're using a substantially more potent GPU or running lighter games, it's really the only way 4K is viable on GPUs like the 4070.

Overall, the RX 6950 XT now leads the PNY 4070 by 3%, which again doesn't mean a lot. It's still the same split between DXR games favoring Nvidia and rasterization games favoring AMD. And while we haven't said much about some of the other GPUs, skipping 4K and getting a card like the RX 6750 XT certainly makes sense if you're on a budget.

Native 4K in the rasterized games ranges from 35 fps in A Plague Tale: Requiem to 81 fps in Forza Horizon 5. So all nine of those games are at least playable, even without upscaling. Of the games that don't hit 60 fps, it's also worth mentioning that four of the six games support DLSS, and two others also support Frame Generation.

The DXR games are another matter. Even with DLSS support in all six games, most aren't going to get to 60 fps or more without Performance mode upscaling, and possibly Frame Generation as well. But you could drop to high settings instead of ultra and probably make up the difference.

• MORE: Best Graphics Cards
• MORE: GPU Benchmarks and Hierarchy
• MORE: All Graphics Content

Our power, clocks, and temperature testing now utilizes the same test suite as our gaming benchmarks, as the PCAT v2 hardware and FrameView software lets us collect this data alongside frametimes. We're also using our updated Core i9-13900K platform, so we're less likely to have CPU or platform limitations playing a role.

We have 1080p, 1440p, and 4K charts for power, GPU clocks, and temperatures below. Then we do a test using Metro Exodus Enhanced at whatever "demanding for the GPU being tested" means, and after letting the game run for 15 minutes or more, we check noise levels. We'll present additional tables and information about efficiency (FPS/W) and value (FPS/$) at the bottom of the page.

Nvidia's Ada Lovelace architecture proves very efficient overall, and PNY locks down power use more than Nvidia's reference card, so even when overclocked power use doesn't break 200W. The maximum average power use we saw was actually just 192W when overclocked, versus 198W running stock (though we didn't test the overclocked card at 4K ultra or 1080p medium).

The PNY RTX 4070 uses less power than AMD's RX 6800 while delivering performance closer to that of the RX 6950 XT. Even if we drop the ray tracing results, it would still be equal to the RX 6800 XT in rasterized performance while using over 100W less power. Like we said, it's an efficient GPU.

GPU clock speeds on their own don't mean too much, but it's interesting how much clocks have shot up with the latest Nvidia RTX 40-series. Taking a page from AMD's RDNA 2 book, Nvidia focused on improving clocks and is now routinely hitting 2.6–2.8 GHz, where the previous generation RTX 30-series usually topped out at around 1.9 GHz (without serious manual overclocking).

Here's where we also start to see some of the low-level limitations of the PNY RTX 4070. It's nearly as fast as the Founders Edition, but clocks are typically slightly lower, and the gap widens at more demanding settings. Overclocking also hits power limits on the PNY card, so that it runs about 100 MHz slower than the FE card at 1440p.

Temperatures are the other part of the equation, and it's possible to tune cards for higher clocks by cranking up fan speeds and noise levels. The PNY RTX 4070 doesn't run hot, but it does run warmer than the reference card by about 4C–5C. Let's look at noise levels to see what's going on...

We check noise levels using an SPL (sound pressure level) meter placed 10cm from the card, with the mic aimed right at the center of one of the fans (the right fan on the PNY card). This helps minimize the impact of other noise sources like the fans on the CPU cooler. The noise floor of our test environment and equipment is less than 32 dB(A).

After running Metro Exodus for over 15 minutes, the PNY RTX 4070 settled in at a fan speed of 57% and a noise level of 42.7 dB(A). That's okay in terms of noise, but it's still 3 dB(A) higher than the Founders Edition. It's also interesting that both the FE and PNY cards are at relatively high fan speed percentages, meaning there's not as much room for additional cooling.

We also tested with a static fan speed of 75%, which caused the PNY RTX 4070 to generate 51.4 dB(A) of noise. So the stock noise levels are pretty average, with larger cards typically making less noise. Conversely, at 75% fan speed this is one of the quieter results... though the RTX 4070 FE still does better.

GPU Value and Efficiency

This final gallery of images shows the full performance test suite, along with the above power, clocks, and temperature information. Latency is also provided, at least in some of the games (depending on the GPU and drivers used).

We've calculated efficiency in FPS/W for the various games, plus value in FPS/$ using the best current online prices we could find (usually at Newegg or Amazon, though B&H and Best Buy were also checked). We've summarized those results in the above table (based on 1440p performance and power), sorted by overall value.

It shouldn't be too surprising that the RX 6750 XT is the best overall value out of the GPUs we've tested, followed by the RTX 4060 Ti and RX 6800 XT. Lower prices will go a long way toward making a GPU a better value, and those are the least expensive cards we've included in this review. The Nvidia and PNY RTX 4070 cards are tied for fourth, followed closely by the RX 6950 XT. Note also that the spread in FPS/$ for a lot of the cards is very narrow, with the RX 7900 XT to RTX 4070 all in the 0.111–0.122 range.

Meanwhile, the RTX 4080 ranks as the most efficient GPU we've tested to date, at least for 1440p ultra, with the PNY RTX 4070 coming in second — just edging out the RTX 4070 FE. All of the Nvidia RTX 40-series GPUs have better efficiency than any of the AMD GPUs, with a pretty sizeable drop going from the RTX 4070 Ti (0.367 FPS/W) to the RX 7900 XT (0.281 FPS/W).

• MORE: Best Graphics Cards
• MORE: GPU Benchmarks and Hierarchy
• MORE: All Graphics Content

The PNY RTX 4070 doesn't try to do anything wild. There's no RGB lighting, no factory overclocks, and not even an option to crank up the power limit for manual overclocking. It's not a card targeted at enthusiasts, in other words, but instead something that OEMs and system integrators might find useful. It will happily chug along at less than 200W, providing good performance.

If you want a flashier RTX 4070 card, or something quieter, there are plenty of options. Depending on the week or month, some of those cards might even cost less than the PNY RTX 4070 Verto Dual. That's fine, and there's room for all types of cards. Some will undoubtedly like the no-frills aesthetic that's on tap.

But if we're being frank ("Hi, I'm Frank..."), we want graphics cards to either beat the Nvidia reference card on price, features, or performance. PNY doesn't do any of those. It's close enough to call performance a tie, but the RTX 4070 Founders Edition will still be a better card for most people.

Unless you hate the superfluous 12VHPWR 16-pin connector and the adapter cable that comes with Nvidia's card. And there are definitely good reasons to dislike it! In a typical case, you end up with a stiff adapter cable that sticks out several inches above the top of the card. Then you have to make the side panel of your case fit on properly, without putting too much strain on the cable.

If you have a high-end ATX 3.0 power supply, maybe it won't matter. But how many people are actually buying high-end PSUs with native 12VHPWR support for a 200W graphics card? We still don't have a good explanation for why Nvidia has pushed so hard on use of the 16-pin connector (or the 12-pin precursor for it's RTX 30-series Founders Editions). It's not like they're provably better, and the melting cable issue still hasn't really ended.

Is that one item a big enough reason to opt for the PNY RTX 4070 over a different model? Perhaps! But if so, know that there are MSI, Asus, Zotac, and Gigabyte models that will also suffice. In fact, it's only the higher-end third-party 4070 cards that have 16-pin connectors, and not even all of those.

PNY has other RTX 4070 options as well, with some RGB lighting if that's what you're after. It's a good GPU, going toe to toe with the previous generation RTX 3080 while cutting the price, increasing features, and improving efficiency. That's not the most exciting description of a new GPU, but you could do far worse.

• MORE: Best Graphics Cards
• MORE: GPU Benchmarks and Hierarchy
• MORE: All Graphics Content

Inno3D has thought of a way to hide the power connectors on the brand’s upcoming GeForce RTX 4070 and GeForce RTX 4060 Ti, two of the best graphics cards on the market. Unlike Asus’ concept of ditching the power connectors, Inno3D ingeniously placed them at the back of the graphics card.

Inno3D is preparing to launch the GeForce RTX 4070 AX eSports and the GeForce RTX 4060 Ti iChill X2 White in the Chinese market. Consumers will surely like the idea of having the power cables routed to the back of the graphics card since it would facilitate cable management and improve the overall aesthetics of the build.

The power connectors reside on the back of the PCB, located in an área between the heatsink and backplate. Inno3D rotated the power connector in a 180-degree orientation so it faces the motherboard instead of the side panel of your case. The good news is that the graphics card only requires a single power cable, so consumers don’t have to worry about routing multiple cables.

In Inno3D’s case, the two upcoming graphics cards leverage a standard 8-pin PCIe power connector. However, we can see the utility of the design for models that use the 16-pin power connector. With all the cases of the 16-pin power connector meltdowns, consumers are paying extra attention when connecting and vending the cable. Inno3D’s design offers ample clearance space for bending the cable, and consumers can discard the power adapter, assuming they have an ATX 3.0-compliant power supply. 

Although both graphics cards have hidden power connectors, they differ slightly in implementation. The GeForce RTX 4070 AX eSports comes with a partially-removable backplate. Users can lift one-half of the backplate to reveal the 8-pin PCIe power connector. On the other hand, the GeForce RTX 4060 Ti iChill X2 White isn’t as long as the GeForce RTX 4070 AX eSports, so there was no way for Inno3D to use the same design. Nonetheless, consumers won’t have to remove the entire backplate to access the power connector. Instead, the GeForce RTX 4060 Ti iChill X2 White features a sizeable cutout where you can plug in the same 8-pin PCIe power cable.

According to MyDrivers, the GeForce RTX 4070 AX eSports and the GeForce RTX 4060 Ti iChill X2 White will have bundled white power cables to match the graphics cards’ white theme. Inno3D didn’t share the availability or pricing for the graphics cards yet.

Nvidia's latest GeForce 535.98 Game Ready driver ushered in support for Diablo IV and fixes for some general bugs. Among the improvements, the driver also brought a significant uplift in GPU compression, boosting performance on PCIe 5.0 SSDs, such as the Crucial T700, one of the best SSDs.

DirectStorage, which substantially decreases loading times in games, leverages GPU compression. Game assets come in a packaged format that requires decompression during playtime. Typically, the processor is in charge of decompressing the game assets. GPU compression takes the load off the processor by having the graphics card process the decompression, so driver-based optimization can significantly improve GPU compression.

Gamers don't necessarily need an SSD to take advantage of DirectStorage. The technology works fine even on old platter drives. However, SSDs get the most performance benefits from DirectStorage.

Compusemble discovered a GPU compression performance uplift when comparing the previous GeForce 535.03 driver with the 535.50 beta driver. The latest GeForce 535.98 driver already includes the optimizations. Compusembler employed the T700 and an undisclosed PCIe 4.0 SSD for its comparison. It uses the typical avocado benchmark to evaluate the performance between the two SSDs and Nvidia GeForce drivers. Admittedly, the list of PC games with DirectStorage support is extremely short, but it would have been nice to corroborate the avocado benchmark results with real-world results from Forspoken.

The GeForce 535.05 beta driver boosted the bandwidth on the T700 from 28.09 GB/s to 33.12 GB/s, amounting to an 18% improvement. Meanwhile, the same driver did little for the PCIe 4.0 SSD, only increasing the bandwidth by 2%. So, how much impact does more bandwidth have on loading times?

According to Compusemble's results, the T700's loading dropped to 0.20 seconds from 0.24 seconds. The first impression may not be exciting since PCIe 5.0 SSDs are already very fast. However, the math shows that the newer driver decreased the T700's loading time by 17%, aligning with the increase in bandwidth. On the contrary, the PCIe 4.0 SSD only enjoyed 3% lower loading times.

The evidence reveals that the PCIe 4.0 SSD has seemingly reached its maximum potential. Unfortunately, the drive is at a dead-end, receiving minimal enhancements from the later GeForce driver. On the other hand, there still seems to be untapped performance for PCIe 5.0 SSDs. The T700's performance is testimony to this. DirectStorage has a bright future for PC gaming. Microsoft has been hyping up the technology for threes year now. We just need more games to support it.

Nvidia today announced a new GeForce Game Ready Driver update that's bound to turn the head of anyone dabbling with local Stable Diffusion installations. The latest GeForce Game Ready Driver Release 532.03, which will be released later today, packs in learnings from Microsoft's Olive toolchain, a hardware-aware model optimization tool that aims to perfectly stitch your diffusion model's processing to your graphics cards' capabilities.

According to Microsoft, Olive has the ability to modify the base model according to the hardware capabilities available — whether in local installations, on the cloud, or on the edge. Olive is a way to abstract away all the different Machine Learning (ML) acceleration tool chains distributed by the different market players. Part of its mission, then, is to reduce market fragmentation around ML acceleration techniques.

Nvidia says the new Olive integration into GeForce will allow for performance improvements in AI operations that leverage the fixed-function RTX Tensor cores present in Nvidia GPUs. Using an Olive-optimized version of the Stable Diffusion text-to-image generator (paired with the popular Automatic1111 distribution), performance is improved over 2x.

“AI will be the single largest driver of innovation for Windows customers in the coming years,” said Pavan Davuluri, corporate vice president of Windows silicon and system integration at Microsoft. “By working in concert with Nvidia on hardware and software optimizations, we’re equipping developers with a transformative, high-performance, easy-to-deploy experience.”

There shouldn't be a reason for AMD to not be able to implement Microsoft's Olive toolset in its own graphics drivers, although it's currently unclear how much these optimizations actually depend on the Tensor performance (remember that AMD doesn't possess an equivalent hardware solution on-board its RX 7000 GPU family). For now, it seems that Nvidia users in particular are poised for even better performance than can be extracted from AMD's GPUs.

Interestingly, while Nvidia makes it clear that the RTX architecture's Tensor cores are mainly responsible for the cards' performance in ML acceleration, graphics cards listed for support with version 532.03 include cards starting from the GTX 700 family. Do any of these older architecture cards that don't sport any Tensor cores see a similar 2x improvement in ML acceleration? Okay, maybe not a GTX 700, but how about GTX 10-series and 16-series parts? If you have one of those and test this out, let us know in the comments.

Besides Microsoft's Olive integration, the latest GeForce Game Ready v 532.03 drivers also bring support for the NVIDIA RTX 4060 Ti, they're Game Ready for Lord of the Rings: Gollum, and they fix some known bugs in Age of Wonders 4 and Bus Simulator 31.

Solidigm's new Synergy 2.0 SSD driver and software are designed to offer up to an incredible 120% increase in some types of 4K random read workloads and up to a 170% increase in 4K sequential tasks, thus delivering leading-edge game loading and system boot performance by leveraging smart algorithms to prioritize the data you use most frequently. 

Solidigm, an SK hynix venture incorporating elements of Intel’s old SSD business, acquired its name from the combination of 'solid state storage' and the word paradigm. It’s only fitting, then, that they have taken efforts to separate from the pack by changing the SSD paradigm from the top down. This strategy has led to the release of Solidigm’s Synergy 2.0 software which works above the SSD hardware and firmware layers. 

Solidgm's overall approach is two-pronged: one side is the Synergy Driver to directly improve the user’s experience, and the other is the Synergy Toolkit as an SSD toolkit application. Together these software components help get more out of Solidigm SSDs via targeted real-world optimizations.

The first prong of Solidigm’s software strategy is the driver, known as the Solidigm Synergy Driver. This includes three prominent performance features including Smart Prefetch, Dynamic Queue Assignment, and Fast Lane.

The most-touted feature is Fast Lane, previously known as Host Managed Caching (HMC). This uses read SLC caching to improve boot and application load times by identifying the most frequently used (MFU) user data. This can improve reads by up to 120% under ideal circumstances, which is with 4KB random reads on a 50% full drive. SSDs perform worse once they are filled from the fresh-out-of-box (FOB) state and the dynamic SLC cache shrinks with drive utilization. Therefore, this feature is best used at between 25% and 75% drive usage, with 50% being the best target.

The Smart Prefetch feature identifies predictable read streams, typically sequential reads with a queue depth of one, to prepare data before it is needed. This is a typical gaming workload with the 4KB I/O size being the most common and having the most performance to gain, although this feature works with up to 128KB chunks in up to eight 512KB streams. Solidigm demonstrated an up to 170% speedup for QD1 4KB sequential reads, but in practice, this should only improve load times by single digits.

Dynamic Queue Assignment works by assigning I/O queues to less-utilized CPU cores, which usually isn’t an issue but can be a bottleneck with certain workloads. This is said to improve QD32 4K random write performance by up to 20%, but would also improve QD32 4k random reads. In general, this feature is designed for high queue depths and particularly with smaller I/O. This has potential use for some types of content creation workloads.

The second prong of Solidigm’s software strategy is an SSD toolkit, or the Solidigm Synergy Toolkit. This toolkit is compatible with all SSDs, including those of competitors. Universal features include real-time health monitoring including S.M.A.R.T., drive information, diagnostics, and secure erase. Drive information includes firmware and driver versions, and firmware may be updated for Solidigm drives through this application. Also shown are the host memory buffer (HMB) status and any partitioning. The write cache can also be evicted on the P41 Plus, which does impact the Fast Lane feature.

You will have to use the Solidigm P41 Plus SSD to explore the new 2.0 driver, as it is the only SSD that currently fully supports the complete functionality. The Solidigm P44 Pro, Intel 665p, and Intel 670p are also supported by the driver, but lack the Fast Lane feature. Solidigm intends to add these features to future drives. Solidigm claims this is a firmware limitation but it may be due to needing something like the P41 Plus’s unique SLC cache configuration. The Intel 660p is not officially supported at all, despite using the same controller as the 665p.

It’s true that software is often an afterthought with SSD design, although Microsoft’s DirectStorage API has encouraged some interest, and Solidigm’s driver fully supports it. Solidigm is also excited about the Synergy 2.0 software improvements, assuring us it has real-world benefits that may not always show up on synthetic benchmarks. The long-term intention is to improve this software over time while developing better hardware products. As such, additions to the Toolkit are forthcoming and the driver will see further optimization.

Getting more out of your device is always a good thing, so we are excited to see what Solidigm's new software brings to the table. You can download Solidigm’s 2.0 software on its website and begin using it today. Meanwhile, we're working on our own series of tests to characterize performance. Stay tuned. 

Turtle Beach is gunning for rivals Razer and SteelSeries with its new flagship gaming headset, the Stealth Pro. And, as an aside, companies need to start actually naming new products instead of just tacking “Pro” onto the name of an existing line, but I won’t fault Turtle Beach for doing this.

Announced last month and available starting April 23, the Stealth Pro is a wireless over-ear gaming headset packed with a laundry list of features and specs that match up almost exactly with SteelSeries’ Arctis Nova Pro — which currently sits at the top of our list of best gaming headsets.

The Stealth Pro has all the gaming features you wanted and didn’t know you wanted, including dual-mode wireless connectivity with simultaneous audio playback, active noise cancellation, a swappable battery system, and a detachable boom microphone and built-in beamforming mics. It sports 50mm dynamic drivers that are hand-matched for balance and has a remappable multi-function button and wheel so you can control everything on the hardware. The Stealth Pro comes in both Xbox and PlayStation versions, both of which can be pre-ordered now for $329.99.

Design and Comfort of the Stealth Pro

The Stealth Pro is a wireless over-ear (circumaural) headset with a detachable boom microphone. It’s part of Turtle Beach’s Stealth line, but it’s been redesigned to look less like a gaming headset and more like a pair of headphones you might wear on your commute. To that end, it comes in just one colorway, black, with silver accents, and the company has no current plans to release additional colors. While I understand this choice, it’s a little disappointing — the navy-and-copper version of the Stealth 700 Gen 2 Max might be my favorite headset colorway (ever).

The Stealth Pro is on the heavier side: my review unit weighed 14.74oz (418g) with the boom mic attached (14.11oz/400g without). For comparison, its predecessor the Stealth 700 Gen 2 Max weighs 13.4oz (380g), while the Corsair Virtuoso RGB Wireless XT weighs 13.47oz (382g). Both the SteelSeries Arctis Nova Pro Wireless and the Razer Barracuda Pro Wireless weigh around 12oz (340g). The Stealth Pro isn’t overwhelmingly heavy — it’s much lighter than the Audeze Maxwell, which weighs 17.28oz (490g) — but it is, for the most part, heavier than the headsets it’s competing against.

The Stealth Pro has a flexible metal headband covered in a slightly-rubbery, soft-touch material with “Turtle Beach” debossed across the top. Its earcups, which are connected via metal yokes, swivel both directions (and swivel flat for storage), tilt, and are individually height-adjustable. The earcups’ speaker plates also feature a soft-touch (though not rubbery) finish, which attracts a lot of fingerprints and, smudges and other unattractive things. I know these soft-touch finishes are kind of Turtle Beach’s thing, but covering up metal with plastic makes the headset look a lot cheaper than it is.

As far as comfort goes, the Stealth Pro is... okay. It’s not uncomfortable, but it’s not so comfortable that I forgot I was wearing it. Its clamping force is on the stronger side and the headband is a little too wide (and I don’t have a small head). Again, while it wasn’t uncomfortable, the fit felt a little off. Perhaps it would benefit from a suspension-style headband like what’s on the SteelSeries Arctis line or the Audeze Maxwell.

The headband and earpads are padded with memory foam covered in nice-feeling black leatherette. The leatherette is comfortable and does a reasonably good job of passive noise cancellation even without the headset’s active noise cancellation turned on, but it does hold in heat. The earpads are glasses-friendly — the foam is less dense in the area where your glasses sit, so you can wear this headset over glasses without getting a headache. I tried this headset on with my sunglasses (I no longer wear glasses), and it definitely works — the earpads put almost no pressure on the sides of my sunglasses.

Most of the headset’s controls are located on the right earcup. The face of the earcup features a remappable multifunction button and wheel. By default, the button toggles active noise cancellation and the wheel controls volume. The wheel has a secondary function, which is activated by holding down the multifunction button for three seconds. The wheel’s default secondary function is game / chat mix. The multifunction button and the wheel’s secondary function can be remapped using Turtle Beach’s companion software or mobile app.

The bottom edge of the right earcup also contains three buttons — Bluetooth, power, and “Superhuman Hearing.” The power button is just a power button, but the other two buttons are multi-function. The Bluetooth button lets you pair and control a Bluetooth device, while the “Superhuman Hearing” button toggles Turtle Beach’s “Superhuman Hearing” feature — a game audio setting that pumps up key sounds such as footsteps and gunfire. Pressing and holding the Superhuman Hearing button for two seconds toggles mute for the headset’s built-in microphones, so you can mute yourself even if you don’t have the boom mic attached. These buttons are well-spaced and easy to differentiate (the Superhuman Hearing button is textured), but they’re relatively stiff and can be difficult to press.

The right earcup also has a USB-C port for charging and updating the headset’s firmware.

The left earcup houses the boom microphone jack, which is covered by a small plastic cover that I immediately lost, as well as the headset’s hot-swappable battery. The hot-swappable battery works just like the hot-swappable battery on the SteelSeries Arctis Nova Pro: it sits under a magnetic speaker plate, and so long as you remove and replace it within a few seconds, your connections won’t drop. My only issue with the Stealth Pro’s system is that the speaker plate is trickier to remove when the boom mic is attached. It’s not impossible, but it’s not the seamless hot-swap you get from the Arctis Nova Pro.

The Stealth Pro comes with several accessories in the box: the detachable boom mic, the wireless transmitter/battery charging station, two USB-C to USB-A cables (one 6.5-foot/2m cable for charging, one 3-foot/1m cable for connecting the transmitter to your PC), two batteries, and a carrying pouch. The carrying pouch, while not as robust as the rigid travel case that comes with the Razer Barracuda Pro Wireless, is pretty nice — it’s made from thick, stiff velvet fabric, and it has an inner pocket that’s large enough to hold all of the headset’s accessories (including the wireless transmitter).

Spec

Audio Performance of the Stealth Pro

The Stealth Pro sports 50mm “Nanoclear” dynamic drivers with a frequency response range of 10 Hz - 22,000 Hz — the same as SteelSeries’ Arctis Nova Pro Wireless. This is a little wider than you’ll find on most gaming headsets (which typically stick to 20 Hz - 20,000 Hz), but it’s not the widest — Corsair’s Virtuoso RGB Wireless XT has a frequency range of 20 Hz - 40,000 Hz, while Audeze’s new Maxwell has a frequency range of 10 Hz - 50,000 Hz.

The Stealth Pro’s default sound profile (called “Signature Sound” in the EQ presets) is relatively balanced and fairly neutral — though it’s perhaps a bit too mid-range focused for me. Audio is not as detailed or layered as it is on, say, the Audeze LCD-GX or the Audeze Maxwell, but the Stealth Pro still managed to separate out most of the delicate sounds in Billie Eilish’s bad guy. Extremely heavy bass sounded a little muddy, but Kaskade’s POW POW POW didn’t turn into a swampy mess, so that’s something. My biggest issue (coincidentally, considering the Stealth Pro’s mid-focused default profile) was with mid-range heavy songs, such as K-391’s Lonely World — which sounded compressed in the mids and lower-mids on the Stealth Pro (and a little less dynamic overall).

When it came to gaming, however, the Stealth Pro was much more impressive. The Stealth Pro features Turtle Beach’s “Superhuman Hearing,” which is a game-focused setting that boosts certain sounds — such as footsteps, explosions, nearby artillery, etc — in games for a competitive advantage. It works, and it works well. I was especially impressed with how it managed to provide a significant boost to these higher-mid-range sounds without going overboard — I turned on the “footsteps” EQ preset for Audeze’s Maxwell headset last week, and I’m pretty sure my dog thought I was trying to communicate with her over high-range frequencies. But turning on Superhuman Hearing for the Stealth Pro nicely boosted the sound of footsteps and gunfire in Fortnite without aggravating my Pomeranian.

The Stealth Pro also sounded good in games with more ambient environmental noise. The jungle in Uncharted 4 sounded lush and layered, while cities in The Elder Scrolls Online had plenty of depth and busyness — even though the Stealth Pro has a much narrower soundstage than, say, the Audeze Maxwell. This headset might not be my go-to for listening, but it’s a great gaming headset.

The Stealth Pro has active noise cancellation (ANC), which is still a fairly uncommon feature on the best wireless gaming headsets — though we’ve been seeing it more often on premium headsets designed to cross over into the lifestyle sphere. Both the SteelSeries Arctis Nova Pro Wireless and Razer Barracuda Pro Wireless have it, as does the Epos H3Pro Hybrid. The Stealth Pro’s ANC is adjustable, from full transparency/ambient mode to full noise canceling. And it’s very good — even at 50% it was able to completely cancel out the hum of my 47 PC case fans (an exaggeration, but it’s still no small task). At 100% it was too good, and I found myself getting dizzy from the lack of... well, noise. At 0% (full transparency/ambient mode), it actually amplifies the sounds around you.

Microphone of the Stealth Pro

The Stealth Pro comes with a detachable cardioid boom microphone as well as built-in beam-forming microphones in the earcups. The boom mic plugs into the mic jack on the left earcup and sits at the end of a 4.5-inch (114.3mm) flexible arm. The microphone jack is located on the face of the earcup (rather than the edge), and the boom mic can be rotated up and out of the way.

Rotating the mic up mutes it, which is a fantastic feature — I’ve used a lot of headsets and it’s hard to tell if your mic is hardware-muted with buttons or switches (mostly because it’s hard to remember which button/switch direction is mute/unmute), but there’s no question with this mic — if you can’t see it, it’s muted. But if you need an additional visual cue, the wireless transmitter flashes red when your mic is muted. It’s a solid green when your mic is on, and this also works when the boom mic is unplugged and you’re using the headset’s built-in mics. 

All the hardware features of the mic aside, the first thing I heard from my coworkers when I logged on with the Stealth Pro was “You don’t sound as good as you normally do” and “I feel like there’s more... static?” Naturally, I immediately started mic testing when I logged off, and after playing around with some settings in the software I didn’t think it sounded that bad so I logged on the next day only for everyone to say, “You still sound worse than normal.” At this point I thought it might be the connection type — I was connected via Bluetooth because I was playing a full-screen game on my PC and didn’t want to switch windows, so the next day I logged in on my PC. I got the same comments — “You sound a little staticky? Or maybe that’s on my end?” — but ultimately people agreed that I sounded fine and the microphone clearly works well enough, even if it’s not perfect. 

I’m not sure if my coworkers are spoiled by my normally incredible broadcast-quality recording equipment, but the boom microphone on the Stealth Pro sounds fine — it’s not the best gaming microphone you can find, but no headset boom microphone is or ever will be. My voice sounds relatively full and is clear enough, and the microphone is definitely on par with boom mics on similarly priced headsets. 

But you can judge for yourself — here’s a test recording on the Stealth Pro boom mic:

The Stealth Pro’s built-in microphones are not quite as good as the boom microphone, but they’re not bad — they’re pretty sensitive, so they managed to pick up my voice fairly loudly and clearly (though they also picked up a lot more background noise). While I wouldn’t use the Stealth Pro’s built-in mics to do a guest spot on NPR, they seem like they’ll work fine for most casual situations — including communicating with your teammates while gaming. 

Features and Software of the Stealth Pro

The Stealth Pro can be configured using the Turtle Beach Audio Hub V2 PC software or the Turtle Beach Audio Hub mobile app. I initially had trouble getting the PC software to recognize the headset when it was connected to my PC, but this seemed to be a pre-production bug. I was eventually able to connect, but I had better luck in general with connecting to the mobile app — which has the same functionality (minus the PC-specific hotkey mapping).

You can use the Audio Hub V2 or the Audio Hub app to adjust the Stealth Pro’s various EQs, which include a number of “basic” adjustable presets for both audio and mics, as well as a 10-band “advanced” EQ for those who want more control. The Audio Hub also lets you toggle several settings that you can (for the most part) also access on the headset itself, including ANC, Superhuman Hearing, noise gate, and mic sidetone. 

But the main reason you’ll want to use the software is probably going to be so you can remap the multifunction button and the wheel’s secondary function — which you can also do on the mobile app. The button can be programmed to toggle ANC, noise gate, or cycle through game presets, or it can be set as a standard multi-function button. The wheel’s secondary function can be set to adjust mic sidetone, ANC level, Bluetooth volume, Superhuman Hearing volume, bass boost volume, treble boost level,  noise gate volume, or mic sensitivity.

The Stealth Pro’s software is a little rough around the edges — like most gaming peripheral software, unfortunately. But the good news is that any changes you make using either the software or the mobile app immediately save directly to the headset — so you shouldn’t need to use the app too often once you’ve got the Stealth Pro set up.

Wireless Performance of the Stealth Pro

The Stealth Pro offers two forms of connectivity — both of which are wireless: low-latency 2.4GHz wireless and Bluetooth 5.1. It does not have a 3.5mm jack for a wired analog connection, which is (I suppose) not too surprising. I would have liked an analog option, but I’m probably in the minority there.

It connects to 2.4GHz wireless via the wireless transmitter (in lieu of a dongle), which is not particularly convenient for those who don’t want to deal with the extra heft of the wireless transmitter, but also not terribly surprising — after all, the Arctis Nova Pro Wireless connects via its base station. On the back of the wireless transmitter you’ll see a switch for jumping between console/PC mode, a USB-A charging port, and a USB-C port for connecting to your console or PC. It’s decidedly less useful than the Arctis Nova Pro Wireless’ base station, which not only lets you control a variety of hardware settings but also has two USB-C ports so you can connect to multiple systems at once.

The Stealth Pro lets you connect to both 2.4GHz wireless and Bluetooth 5.1 and can play audio from both connections simultaneously. This might seem like a niche feature, but it’s one of my must-haves, because it means I can play a full-screen game on my PC while also taking a meeting on my phone (this is what I do almost every morning during the Tom’s Hardware morning meeting).

The Stealth Pro doesn’t give you an easy way to adjust the mix between the two audio sources, however. You can adjust the volume of each source separately by assigning “Bluetooth volume” as the secondary wheel function (the primary wheel function controls the main — 2.4GHz wireless — source audio), but this is pretty clumsy. There are headsets that handle this better, such as Logitech’s G735, which has a rocker button that you can use to mix the 2.4GHz wireless and Bluetooth sources.

Battery Life of the Stealth Pro

The Stealth Pro comes with two 12-hour batteries that can be hot-swapped for a total of 24 hours of battery life on a single charge. The wireless transmitter doubles as a battery charging dock, so — in theory — you’ll never run out of juice, so long as you’re gaming at home on your PC or console.

This is exactly the same system as SteelSeries’ Arctis Nova Pro Wireless, right down to the base station-slash-charging dock. And while it’s a great system for at-home use, it’s less convenient when you’re away from your PC (or console). Assuming you don’t want to lug a wireless transmitter around with you, the headset now only has 12 hours of battery life when you’re on the go, which is... not much. Most headsets easily get twice that — the Stealth 700 Gen 2 Max gets up to 40 hours on a single charge. Even the Arctis Nova Pro Wireless’ swappable batteries get 18 - 22 hours each. 

I like the hot-swappable battery system in general, because it means you essentially have unlimited battery life while you’re near the wireless transmitter. But when you’re not near the wireless transmitter, you’re probably going to want a little longer than 12 hours.

Bottom Line

I was excited for the Stealth Pro — I’ve been a big fan of Turtle Beach headsets for years, and the Stealth 700 Gen 2 Max is one of the few headsets I have that I know will just work (hey, it’s hard out here for a peripherals editor). But the Stealth Pro is sort of underwhelming. It’s clearly going for SteelSeries’ Arctis Nova Pro Wireless, but it just doesn’t go far enough. 

Don’t get me wrong, the Stealth Pro has its strengths — its audio performance didn’t blow me away when I was listening to music, but games sounded great thanks to Turtle Beach’s Superhuman Hearing audio presets. The microphone, despite being hated on by my coworkers, is solid for a headset mic, and the built-in mics are better than expected. The ANC is good — too good — and the headset also comes with a nice carrying bag that will at least keep everything together when you’re traveling with it. 

But if Turtle Beach is going to go up against SteelSeries, it’s going to have to try a little harder. It’s nice to see a swappable battery in a headset that’s not SteelSeries, but not one that gives half the battery life. And while I’m glad to see Turtle Beach is stepping outside of its comfort zone with a more lifestyle-oriented design, this headset definitely doesn’t look like it costs $330. 

MORE: Best PC Gaming Headsets

MORE: Best Wireless Gaming Headsets

MORE: Best Gaming Keyboards

MORE: Best Gaming Mouse

I love seeing gaming headsets from audio companies, because they approach it from the audio side first, not the gaming side. So I know they’re going to sound great. And Audeze’s newest wireless gaming headset, the Maxwell, does sound great. But sounding great is just one aspect of what makes the best gaming headsets, well, the best. 

The Maxwell is a large, wireless over-ear gaming headset with a steel and aluminum frame, a detachable boom microphone, and 80+ hours of rated battery life. It features newly-designed 90mm planar magnetic drivers (more on what that means later) and an out-of-the-box sound profile that’s tuned for gaming. And again, it sounds fantastic. Design-wise, it’s a departure from Audeze’s previous wireless gaming headsets, the Penrose and the Mobius, which were smaller, lighter, and mainly made of plastic. 

The Maxwell is available now and comes in two versions: a PlayStation version ($299) and an Xbox version ($329). Both versions work with PC, but the Xbox version is licensed by Microsoft and includes an embedded Dolby Atmos license (both versions of the headset also work with existing Dolby Atmos licenses). We reviewed the PlayStation version.

Design and Comfort of the Maxwell

The Maxwell is a large, sturdy headset with an aluminum and steel frame and glass-infused nylon earcups with plush, leatherette-covered foam earpads. The headset has moderate clamping force, which, combined with its weight of 1.06  pounds (490g), means you’ll definitely feel it on your head.

In the many hours I spent wearing this headset, I didn’t once feel discomfort from the strength of the clamping force or the overall heftiness of the unit. I did find the large size of the earcups to be a little inconvenient while traveling — specifically, it was difficult to lean against anything with such bulky earcups — but that shouldn’t be too much of an issue if you plan on using the Maxwell for its primary purpose of at-home gaming.

The Maxwell isn’t the sleekest headset — in fact, it looks a little clunky next to some of the best wireless gaming headsets, especially headsets designed to cross over into the lifestyle category, such as the Razer Barracuda Pro Wireless. While headsets like the Barracuda Pro Wireless have at least sort of attempted to sculpt down the size of their earcups, the Maxwell simply owns its bulkiness. I’m sure some of the Maxwell’s earcup size and chunkiness have to do with the dual-chamber design, which acoustically isolates the driver from the headset’s electronics. But, while I find this forgivable (in pursuit of audio quality!), there’s no denying that this headset is huge. 

The good news is that it feels very sturdy (the bulk probably helps with that), with a steel headband, aluminum yokes, and thick, weighty earcups. It’s a big step up from Audeze’s previous wireless gaming headsets, the Penrose and the Mobius, which were largely made of plastic.

The Maxwell’s understated, monochromatic design doesn’t scream “gamer,” but this isn’t too unusual. Many of the best gaming headsets, especially premium ones, are moving in that direction. The headset’s frame is a dark gunmetal gray, while the leatherette parts (earpads, headband strap, and headband padding) are black. ‘Audeze’ is printed across the top of the headband in light gray, and the company’s ‘A’ logo is printed in blue on both sides of the headband, just above each yoke. The headset has no lights, aside from a pinprick-sized LED indicator on the left earcup that flashes different colors to signal connection status and battery life.

The earcups swivel in both directions (and also swivel flat for storage) and tilt for comfort. The only major point of adjustment, however, is the leatherette tension headband, which can be shortened or lengthened via three sets of notches on each side. The headband is easy to adjust (you can just pull it off of the screws, you don’t have to break out a screwdriver, or anything), but this system does seem a bit limited. After all, it only allows for five lengths — six, if you count removing the strap completely. 

I was able to find an adjustment that worked for me, but I would have preferred more nuance. I suppose you could use a leather punch to add additional holes to the strap, but ideally, you wouldn’t have to go to those kinds of lengths just to make a $300+ headset fit. SteelSeries’ Arctis Nova headsets have a similar suspension headband system, but those headsets also have independently height-adjustable earcups, while the Maxwell’s earcups are fixed.

The Maxwell features soft, foam, leatherette-covered earpads that are contoured to fit the curvature of your skull. They’re similar to the earpads on the Audeze LCD-GX, but because the Maxwell has smaller earcups, the contour is less dramatic. The headset doesn’t have active noise cancellation, but the earcups do an excellent job of passive noise cancellation. The leatherette stays fairly cool for leatherette, but it still retains more heat than mesh fabric-covered earpads (such as those on the SteelSeries Arctis Nova 7). 

The earpads are replaceable and twist off easily. Audeze doesn’t currently have replacements for the Maxwell listed on its website, but it does have replacements for the Penrose/Mobius gaming headsets listed. So I assume the Maxwell replacements will make their way to the shop eventually.

The left earcups houses all of the Maxwell’s controls and ports. On the left earcup’s face, there’s a power button that also acts as a play/pause (single press) and Bluetooth pairing (double press) button, as well as a physical microphone mute switch. On the edge of the earcup, starting at the front, there’s a button for toggling the headset’s microphone AI noise reduction feature, a 3.5mm microphone jack, a USB-C port, a 3.5mm aux jack, a clickable wheel for chat mixing and sidetone, and a clickable wheel for volume. Most of the buttons and wheels on this headset have multiple functions, which makes them a little confusing at first — but no more confusing than most multi-mode wireless gaming headsets these days. 

The Maxwell notably does not have a way to manually switch between wireless connections (2.4GHz and Bluetooth), which is frustrating as it also does not allow for simultaneous listening on those two connections. Instead, the headset is designed to automatically switch between those two connections, prioritizing Bluetooth. I have a lot of thoughts on this, but we’ll dig deeper into it in the Battery Life and Wireless section.

The Maxwell comes with several cables and accessories: a detachable boom microphone, a 2.4GHz wireless USB-C dongle, a 1-foot USB extender, a 4-foot aux cable, and a 5-foot USB-C to USB-C cable. It does not come with a carrying case (or even a protective cloth bag), which is a little disappointing (the large, unblemished earcups make me wary of stuffing it into a backpack unprotected). While it’s true that most gaming headsets don’t come with carrying cases, I think higher-end headsets should. After all, spending $300+ on a headset is much more palatable if you can use it for more than just gaming.

Audio Performance of the Maxwell

Most gaming headsets (and most headphones and speakers in general) have dynamic drivers, which produce sound using a passive diaphragm attached to a moving voice coil — which is why dynamic drivers are also called “moving coil” drivers. Audeze specializes in a different type of driver: planar magnetic. Planar magnetic drivers produce sound using a diaphragm with the conductor embedded inside it (instead of attached to it, like a voice coil is), positioned between two magnetic arrays. This allows for more detailed control over the diaphragm, which means (in theory, at least) more accurate sound reproduction. 

The Maxwell sports newly designed 90mm planar magnetic drivers with a frequency range of 10 Hz - 50,000 Hz. This is much wider than the average frequency range of gaming headsets, which mostly stick to a range of 20 Hz - 20,000 Hz. The $350 SteelSeries Arctis Nova Pro Wireless has a slightly wider range of 10 Hz - 22,000 Hz (up to 40,000 Hz when wired) and the $270 Corsair Virtuoso RGB Wireless XT has a range of 20 Hz - 40,000 Hz. Both of those headsets have dynamic drivers, but the $330 HyperX Cloud Orbit S, which also uses planar magnetic drivers, has the same frequency range as the Maxwell. The Maxwell is an active headset with an internal DSP and a built-in DAC, so it will need to be turned on even if you’re using a wired analog connection. 

The Maxwell’s default sound profile is tuned for gaming — which makes sense, as it’s a gaming headset — with punchy lows, slightly pulled-back mids, and bright, slightly elevated highs. The headset’s wide soundstage did an excellent job of layering details in the jungles of both Uncharted 4 and Shadow of the Tomb Raider, two of my favorite games for immersive environmental audio. 

I made it through enough of The Callisto Protocol to confirm that the Maxwell’s crisp, detailed audio and passive noise isolation is much too impressive for me to be playing horror games with. Directional sound is excellent — even without spatial audio — in games like Fortnite and Overwatch 2, especially on the headset’s “competition” EQ preset (there’s a “footsteps” EQ preset as well, but the highs are a little too high for me). The Maxwell doesn’t include spatial audio in its PC app, but it does support Sony’s 3D audio on PS5.

But while the Maxwell’s default sound profile is good for gaming,  it’s less so for listening to music. Luckily it only takes a handful of EQ adjustments to rein in the headset’s aggressive highs and pump up the mid-range for a balanced, relatively neutral listening profile. The app has four customizable EQ presets that, once configured, save to the headset itself — so you’ll only have to do this once. 

Once I had the headset’s EQ sorted out, listening to music on the Maxwell was a pleasure. It did a great job of producing full-bodied, clean-sounding bass in Knife Party’s Boss Mode and maintaining the power of the low beat in Kaskade’s POW POW POW without getting too muddy. In K-391’s Lonely World, vocals and sustained mids sounded full and balanced (again, once I adjusted the EQ — the mids are a little recessed out of the box).

Microphone of the Maxwell

The Maxwell comes with a detachable boom microphone, which plugs into the 3.5mm mic jack on the left earcup. The mic sits at the end of a 4-inch gooseneck arm and comes with a removable foam pop filter. It appears to be the same boom mic that came with the Penrose and the Mobius, but the Maxwell features a built-in AI-powered noise suppression feature that works to filter out background noise while you’re talking.

The boom mic sounds pretty good — it adds a certain broadcast-quality fullness to your voice that you won’t get from most gaming headset mics — but it’s a little finicky. Depending on the connection, it picks up quite a bit of self noise and often sounds fuzzy, especially in the higher mid-ranges.

The AI noise suppression feature has two levels (high, low), and did a great job of filtering out consistent background noise, such as my 37 PC case fans, without thinning out or otherwise distorting my voice. It even managed to dull the clicky sounds of my keyboard to the point where the people I was talking to were still able to understand my words perfectly. (It didn’t, however, disguise the fact that I was typing on a very clicky keyboard during our Tom’s Hardware morning meeting.)

When you unplug the boom mic, the Maxwell switches over to its built-in mics. That’s right, this headset has an on-board beamforming microphone array for when you don’t need (or want) an external boom mic. It’s nice to see a gaming headset with both options, as attached boom mics do not look very cool when you’re not sitting in front of a PC or a TV. With that in mind, the Maxwell’s built-in mics are mostly for convenience — they’re nowhere near as good as the boom mic, and, frankly, they’re not even particularly impressive for built-in mics (the beamforming array in the Barracuda Pro Wireless is much better). But they work well enough if you need to take a phone call (indoors) and you don’t have the boom mic handy. 

Ultimately, while the Maxwell’s mics will work in most casual situations (how good do you really need to sound over Discord), if you’re streaming, podcasting, or otherwise creating content, it’s always worth investing in one of the best gaming microphones.

Specs

Software of the Maxwell

The Maxwell works with Audeze’s Audeze HQ software, which is designed to support the brand’s wireless gaming headsets. Unfortunately, at the time of this review, the Maxwell software situation is still... let’s say, incomplete. Specifically, the desktop version of Audeze HQ does not yet include an EQ for the Maxwell (Audeze assures me it’s coming in a future update). This is inconvenient, because — as I mentioned earlier — the headset does need some tweaking if you’re planning on using it in a non-gaming capacity.

Luckily, the mobile (iOS and Android) version of Audeze HQ does include an EQ, so you can make your adjustments in the mobile app and they will save to the headset’s onboard memory. However, you will need to disconnect the headset’s dongle from your PC to use the mobile app. 

EQ aside, Audeze HQ also has a handful of other useful features — such as the ability to update your device’s firmware and your dongle’s firmware, which you’ll need to do when you first take the headset out of the box. You can also adjust the headset’s auto-shutdown timer and voice prompt volume, and toggle the volume limiter on or off.

You can also use Audeze HQ to control the mic sidetone (how much of your own voice gets pumped into the headset), mix between game and chat audio, and select an EQ preset. You can perform all of these actions directly on the headset, using the multi-function chatmix and volume wheels. Clicking the chatmix wheel once lets you adjust sidetone, while clicking it twice toggles sidetone on and off; clicking the volume wheel once lets you scroll through tracks, while clicking it twice lets you scroll through the EQ presets. The Maxwell comes with six pre-tuned presets and four customizable presets, which, again, can only be customized using the mobile app for the time being. The presets include “Audeze” (the out-of-the-box preset), treble boost, bass boost, immersive, competition, and footsteps.

Wireless Performance of the Maxwell

The Maxwell has four connectivity options: 2.4GHz wireless, Bluetooth 5.3, wired via USB-C, and analog. While the headset can connect to both 2.4GHz wireless and Bluetooth 5.3 at the same time, it can’t play audio from both wireless sources simultaneously, like the Corsair Virtuoso RGB Wireless XT or the Logitech G735 can. It can, however, play dual-source audio simultaneously provided one of the sources is wired — e.g., it can play audio simultaneously over an analog connection and Bluetooth. It’s an active headset with an internal DSP and a built-in DAC, so it will need to be powered on even if you’re using a wired analog connection.

This isn’t necessarily an issue, as most gamers don’t need a headset that can play from two wireless sources at once — however, the Maxwell also doesn’t allow you to manually switch between the two sources. Instead, it switches between the two sources automatically, prioritizing Bluetooth. It will automatically switch to Bluetooth if you get a phone call, or if you open an app on your phone that is playing sound. 

The reasoning behind this makes some sense — it’s so you won’t miss phone calls. As a millennial, I am extremely averse to talking on the phone (for any reason, ever), so that probably colors my opinion. But, phone calls notwithstanding, it’s still incredibly frustrating to not be able to manually switch back to the 2.4GHz wireless connection without silencing your phone completely. I pick up my phone and fiddle with random apps all throughout the day, and any app that has sound will trigger the switch. Also, while the switch to Bluetooth is instantaneous, the switch back takes a second (which makes sense, logically, but is frustrating if you’re trying to get straight back into your game).

Connection switching aside, both of the Maxwell’s wireless connections have impressive strength and range. The Maxwell is the first headset I’ve used that’s been able to maintain a wireless connection when I walk from my home office to my kitchen — which isn’t very far away, but i separated by a four-foot-thick industrial factory wall made of steel and concrete (nothing has been able to penetrate this wall — until now). As for actual range: I wore the headset when I took my dogs on a walk, and it was still able to pick up my PC audio perfectly when I was outside, six stories below my apartment and about 50 yards from the building.

Battery Life of the Maxwell

The Maxwell also excels in battery life. Audeze rates the headset's battery life at 80+ hours over both 2.4GHz wireless and Bluetooth, which is... very, very good, especially for a planar magnetic headset. Shockingly good. (But not HyperX-Cloud-Alpha-Wireless-300-hours good.) I took the fully charged Maxwell on a flight to Tokyo (and back), and the battery was still at 75% after almost 30 hours of continuous listening (at higher than 50% volume level) — so Audeze’s estimation looks to be slightly conservative. The headset also supports fast charging, so a mere 20 minutes of charging time will get you 20 - 30 hours of playback.

Bottom Line

The Maxwell sounds fantastic, whether you’re gaming, chatting, or listening to music. You won’t find a better-sounding gaming headset at (or near) this price point. It’s also surprisingly comfortable, despite being one of the larger and heavier headsets we’ve tried, and its 80+ hour battery life far outlasts most of the competition (though, so does its weight). The Maxwell makes more sense for the gamer-leaning-audiophile than does the Audeze LCD-GX, which is basically just an audiophile headset with a mic, and it’s also much cheaper. 

But there’s more to gaming headsets than being really, really... really ridiculously great-sounding. (A lot more, actually.)

I didn’t have too many problems with the Maxwell’s wireless connectivity — not counting the purposeful lack of control over connection switching — but I did have some. Occasionally the dongle wouldn’t connect (this happened more often when it was plugged into a lower power USB port or a USB hub), and in order to change the EQ in the mobile app, I had to unplug the dongle. Also, the mobile app is the only place you can change EQ settings (for now). There have been firmware updates since the headset launched, and I’m sure there will be more, but this is something to keep in mind. 

For a little less ($250), Razer’s Barracuda Pro Wireless gaming headset is lightweight, has active noise cancellation and excellent beamforming mics, and comes with a nice travel case. It can’t step to the Maxwell in terms of audio quality, but everything else works much better. For a little more ($350), SteelSeries’ Arctis Nova Pro Wireless sounds very, very good and also features active noise cancellation and simultaneous audio over 2.4GHz wireless and Bluetooth. But if audio is your main priority by far, the Maxwell is hard to beat.

MORE: Best PC Gaming Headsets

MORE: Best Wireless Gaming Headsets

MORE: Best Gaming Keyboards

MORE: Best Gaming Mouse

Intel’s Skylake integrated graphics has reached end-of-life (EOL) status, according to official Intel documentation spotted today by TechPowerUp. Many recognize Skylake as the 6th generation of Core processors, but the family launched with 9th generation iGPUs that will no longer benefit from driver updates, even if a critical issue comes to light.

However, the EOL of Skylake graphics isn’t that surprising. For users of these systems, it will likely mean very little in practice unless a gaping security vulnerability is found in the old drivers. Users of discrete graphics cards will be less worried by this EOL news, as they can disable the iGPU in BIOS, avoiding the Intel graphics driver software.

Introduced in 2015, Intel’s Skylake CPUs started to feel decidedly vintage when Windows 11 arrived with its Intel 8th Gen Core or newer architecture requirements. Last summer, we reported on Intel ending day-0 GPU driver support for 10th gen and older CPUs, with Skylake at the bottom of the list of generations remaining on legacy support. Now that Skylake has dropped off the legacy list, Kaby Lake chips are on the bottom rung of the ladder, waiting to be allocated a plot in silicon heaven.

Intel’s 7th Gen Core Kaby Lake CPUs debuted with Intel Gen 9.5 integrated graphics. Thus, it might be harder to EOL Kaby Lake, as the same graphics architecture persisted through 8th, 9th, and 10th Gen Core processors.

The last regular iGPU driver release available for Skylake users is version 31.0.101.2115, from December 2022. This legacy driver is available here, supporting various editions of Windows 10 & 11, with the Skylake EOL notice in full.

As noted above, the EOL news probably has little consequence for desktop Skylake system users. However, this news and Intel’s legacy driver change last summer might devalue 10th Gen Core and older systems relying on iGPUs with few or no options for a GPU upgrade (e.g., laptops, AiOs or mini PCs). We must also remember that Windows 11 kicked many otherwise serviceable machines into a kind of upgrade limbo. We don’t know what Windows 12 requirements may be regarding CPU, GPU, security, connectivity, RAM, storage, etc.

Nvidia published a new YouTube video this morning showing Cyberpunk 2077's new RT Overdrive mode in action with a GeForce RTX 4090, along with a breakdown of the changes. The graphics are incredible, but so is the performance hit on the RTX 4090. The fully path-traced renderer brought the RTX 4090 to a snail's pace and ran the game at just 16 FPS natively with the mode active. To achieve favorable frame rates, DLSS 3's frame generation and upscaling were required. But let's take a moment to step back and dissect precisely what we're talking about here.

RT Overdrive mode is a brand-new rendering mode being added to Cyberpunk 2077 on April 11th. The new graphics mode replaces the game's current ray-tracing pipeline with a fully unified path-tracing solution that significantly expands the game's rendering engine. There's no hybrid rendering, only ray traced pixels. Nvidia has previously demonstrated this mode, and it notes that on average it needs 635 ray calculations per pixel — compared to just 39 rays per pixel with Battlefield V, one of the original launch games for RTX ray tracing.

The regular ray tracing modes in Cyberpunk 2077 were a hybrid rendering solution, consisting of rasterization in combination with ray-traced reflections, global illumination, and other effects. With the fully path-traced solution, every single light source in the game is simulated to generate the most realistic lighting, shadows, and reflections possible.

All light sources in the game are path-traced, in conjunction with Nvidia technologies including NVIDIA Real-Time Denoisers (NRD), NVIDIA RTX Direct Illumination (RTXDI), and Shader Execution Reordering (SER) to improve image quality and accelerate performance. It's basically the same overhaul that Nvidia did with Portal RTX, only now applied to a significantly more modern and demanding game engine.

Overdrive is the new Crysis

Nvidia's YouTube video demonstrates the absurd performance penalty of running a AAA title like Cyberpunk 2077 with full-blown path tracing enabled. Nvidia used a GeForce RTX 4090 to run the Overdrive tech demo at 4K resolution, and the GPU couldn't even hit 30FPS at native resolution.

To achieve playable frame rates, Nvidia used DLSS 3 frame generation in combination with image upscaling (DLSS 2 Performance mode) to get frame rates back into the triple digits. This is similar to Portal RTX — another fully path-traced game, where the RTX 4090 also needed DLSS 3 frame generation and/or upscaling to achieve playable frame rates.

This should put into perspective how much more demanding 3D light simulation is compared to just faking it. But thankfully, we're now at the point where good quality light simulation can be done in real-time, even if it still generally needs upscaling to get to decent frame rates.

We don't know the system requirements yet for this new Overdrive mode, but its safe to say you'll probably need a top-tier RTX 30-series GPU, or one of Nvidia's RTX 40-series GPUs, to run Overdrive mode at high frame rates. Considering Nvidia was running at 4K, you can probably get decent performance at 1080p with far less potent GPUs than the RTX 4090, and Cyberpunk 2077 does support both DLSS 2 and AMD FSR 2.1 upscaling.

We'll have to see what performance looks like on other GPUs when the free update arrives on April 11.

The matter of fake GeForce graphics cards floating around Chinese e-commerce platforms has been a widespread issue for some time now. However, a new report from MyDrivers suggests that Nvidia is finally doing something about it. The news publication claims that the chipmaker is working with the top Chinese e-commerce companies to eradicate counterfeit GeForce gaming graphics cards, which are common tenants on our list of best graphics cards.

Obscure graphics card brands, including 51RISC, Corn, or MLLSE, were previously trapped in China's domestic market. However, these brands have silently crept into the U.S. market over the years through platforms like eBay or, more recently, via third-party marketplaces at Amazon or Newegg. Most of the time, consumers will find GeForce graphics cards from these Chinese brands at lower prices or sometimes with ridiculous price tags. Common sense would tell you to avoid them and stick to Nvidia's official partners. However, every once in a while, a less-experienced consumer would fall for the lower prices.

The end of the Ethereum mining boom left remnants of overused mining graphics cards on the market. Some sellers are repainting the memory chips on old graphics cards to sell them on the second-hand market. Others are repurposing mobile graphics cards into desktop form and creating phony names like the GeForce RTX 3070 TiM. 

According to MyDrivers, Nvidia has reportedly denied its relationship with these bogus brands, highlighting that they're not official partners. Furthermore, the report claims that the chipmaker allegedly doesn't know how the counterfeit brands produce the knockoffs. As a result, Nvidia is seemingly working hand-in-hand with e-commerce platforms to remove the fake GeForce graphics cards.

It seems that Nvidia's efforts have immediately yielded positive results. Big-name Chinese platforms, including JD.com and Douyin, have cleaned house and restricted the sales of second-hand, refurbished, and falsely advertised GeForce RTX 20-series (Turing), GTX 16-series (Turing) and GTX 10-series (Pascal) graphics cards. Pinduoduo and Tmall have followed suit, implementing new second-hand and refurbished sales policies.

Nvidia shared three valuable tips to help consumers prevent buying counterfeit graphics cards. First, the chipmaker purportedly recommends users buy a GeForce RTX 40-series (Ada Lovelace) graphics card. The logic is that since Ada launched after the mining boom, there shouldn't be any recycled GeForce RTX 4090 or GeForce RTX 4080 mining graphics cards around. Secondly, consumers should buy from Nvidia's official partners, such as Asus, Gigabyte, MSI, and Zotac - to name a few. Finally, for consumers who understandably don't want to spend lots of money on a GeForce RTX 40-series product, the GeForce RTX 30-series (Ampere) is still pretty good. However, to avoid fakes, Nvidia ostensibly suggests consumers pick up a post-mining boom SKU that has enjoyed a facelift along the lines of the GeForce RTX 3060 Ti GDDR6X or the GeForce RTX 3060 8GB.

It's excellent that Nvidia has acted on fraudulent GeForce graphics cards. However, we hope to see the chipmaker's crackdown extend to the U.S. market because there are still some residues of graphics cards from 51RISC or MLLSE around.

Drivers are an integral part of the best graphics cards — without them, you only get baseline functionality, with no fancy 3D graphics, video encoding or decoding, upscaling, or all the other things we've come to expect. We compared driver download sizes for the latest GPUs in late January and found that Intel was strangely bloated. We must have caught someone's attention, as the latest 4255 drivers, which are also WHQL certified, are about half the size of January's 4090 beta drivers.

We mused at the time that maybe Intel was including some unneeded cruft, or maybe it just wasn't compressing things as much as it could. Whatever the case, there's been remarkable progress in just two months. I've got a collection of just about every Intel Arc driver release since launch. Here are the exact sizes of the downloads (which doesn't necessarily equate to uncompressed install size, but it's far easier to check), release dates, and other details.

There are a few things worth pointing out, like the jump in size when Intel went from just supporting the A380 (and various existing integrated graphics solutions) to the official Arc launch drivers. Why the extra 500+ MB? Part of it was merging the Arc Control Center into the main driver download, and there were bug fixes and other factors that likely played a role. Later in October, the size dropped by about 170 MB.

From then until February, the size of Intel's Arc drivers remained pretty consistent at around 1.2 GB. Note that late January was when we wrote the piece about how bloated Intel's drivers seemed to be compared to AMD and Nvidia. By March, the first driver release that month had lopped off about 100MB in size for the Raptor Lake-U laptop launch.

The next day, a different driver came out, reducing the size by 186MB, but that was only the beginning. The current 4255 drivers that came out last night dropped another 284MB in girth. The running total of weight loss since October is at 761MB, making Arc a serious contender for the Biggest Loser: they're 44% of the drivers they once were! While we're not entirely sure about all of the details, Intel's driver blog states has this to say:

"Good things come in small packages — the Intel Arc graphics driver package, specifically. This latest driver release punches above its weight, now down to 604 megabytes from nearly double that when the Intel Arc desktop GPUs launch in October. Our engineers put the old 1.3GB driver download on a diet with smarter compression algorithms. This means faster updates so you can Game On even sooner with less bandwidth consumed, all with zero compromises in performance or features made."

I've provided a more detailed analysis of what's happened with the old versus new drivers in this forum post. Intel mentions a compression algorithm change (it looks like it went from from Zip to 7-zip), but that alone isn't enough to get the 50% reduction. 7-zip probably gets 10–20 percent higher compression, but there are two other major differences.

First, the old drivers packaged the full "Intel Driver and Support Assistant Installer" (around 229MiB), while the new drivers use a smaller version that likely downloads and installs additional stuff (it's 42.7MiB now). The second big difference is that eight DLLs (igd11dxva32.dll, igd11dxva64.dll, igd12dxva32.dll, igd12dxva64.dll, igd9dxva32.dll, igd9dxva64.dll, igvkMedia32.dll, and igvkMedia64.dll) in the old drivers are about 36.6MiB smaller. That data appears to be repeated content that is instead moved into two new DLLs (media_bin_32.dll and media_bin_64.dll) that are each around 29.3MiB in size.

Overall, the file format changes appear to have reduced the driver size by 420MiB, and the new compression algorithm upgrade gets somewhere around a 3.16-to-1 compression ratio, compared to a 2.0-to-1 compression ratio before. Regardless, smaller downloads are a good thing for anyone with a data cap. Says the guy who downloaded over 300GB of Large Language Models while poking around at chatbots last week.

Update, 3/14/2023: Nvidia has now released an official WHQL driver, version 531.29, that contains the hotfix updates as well as addressing one or two other items. You can find the full 531.29 release notes if you want, but there's not much to add. Our original report on the hotfix release follows.

Nvidia has released GeForce Hotfix driver version 531.26. This hotfix is delivered to remedy a widely reported driver bug, which was an unwelcome guest within the latest WHQL certified download. The bug introduced an Nvidia software component which continued to hog 10-15% of your CPU time, even after you quit a game.

Earlier this week we reported that some, but not all, Nvidia GeForce driver version 531.18 users were suffering from overly high CPU usage. Double digit CPU utilization figures were being observed with the latest game-ready driver. It wasn’t clear why this problem occurred, and the fact that some expert users couldn’t reproduce the bug wasn’t helpful for understanding the cause. We asked Nvidia about it, and were told the problem typically occurred after a dozen or more game launches.

The GeForce Display Container Service has now been pinpointed by Nvidia as the source of the CPU utilization bug, but we already knew that, as it could be observed in Windows Task Manager. Nevertheless, we must be thankful that Nvidia has provided a hotfix driver so quickly.

Game Ready Driver version 531.18 was probably quite a popular release, as it is notable for adding RTX Video Super Resolution support, as well as optimizations for Atomic Heart DLSS 3, and The Finals closed beta. It also fixed some bugs, but somehow introduced an irksome CPU utilization issue.

Sean Pelletier, Senior Product Manager for GeForce Game Ready Drivers, Tweeted earlier today about the hotfix being ready. In a follow-up post he reminded potential users that hotfix drivers “have not gone through the full Game Ready testing process and are not WHQL-certified”. 

There’s a bonus for some GeForce GPU packing laptop owners in the GeForce Hotfix Driver version 531.26 that is also worth highlighting. The issue where a “random bugcheck may be observed on certain laptops with GeForce GTX 10/MX250/350 series GPUs,” has also been fixed today.

If your CPU has been taxed unnecessarily by the 531.18 bug, then you can grab the new hotfix 531.26 driver from the link in this short Nvidia support post. If you are wondering what the fuss is about, perhaps it is best to wait for the next WHQL game ready driver release.

No one loves the fact that motherboard prices keep going up (except probably the motherboard makers themselves). But there’s no denying that flagship boards still pack in the most complete and most interesting collection of features, along with high-end specs in an attempt to draw in those looking for the best of the best to build their new system around. And with AMD’s new Ryzen 7000X3D CPUs arriving with impressive performance, it’s likely that more people will be looking to build rigs based around AMD’s X670E chipset.

So here we’ll be looking at a few flagship AMD boards from Asus, Gigabyte and MSI. Similar to our Z790 flagship roundup on the Intel side, we’ll detail three (of four) flagship motherboards from the B670E platform. We’ll test and discuss the MSI X670E Godlike ($1,299), Asus Crosshair X670E Extreme ($999), and the Gigabyte X670E Aorus Xtreme in detail, and also cover the notably less expensive (but flagship nonetheless) ASRock X670E Taichi ($479.99). All boards offer a premium appearance, overkill power delivery, PCIe 5.0 slots, and M.2 sockets. Are any of these worth the price? Of these, which should you choose?

The biggest difference between these boards is the price. From the sub-$500 Taichi to the nearly $1300 Godlike, there’s a huge gap from least to most expensive. The problem (for the consumer) is that aside from just price, there isn’t a $700 difference between them on any front, including performance. While some VRMs are better, even the worst (and I use that term loosely) easily handles the Ryzen 9 7950X without issue. It comes down to how many and what speed M.2 sockets you desire, if you’re OK with last-gen flagship class audio or your want the latest and greatest, and finally, 40 Gbps USB ports. 

We’ll dig into all of this and more shortly to see if any of these high-end options can earn a slot on our Best Motherboards list. But it’s undeniable that the ASRock board has a fighting edge with its (comparatively) low price and impressive specifications. But first, here are the specifications for our first subject, MSI’s X670E  

Specifications: MSI X670E Godlike

Inside the Box of the MSI MEG X670E Godlike

The X670E Godlike, like it’s Z790 Godlike twin, includes a slew of accessories as value adds (like the M.2 Expander card, and M-Vision Dashboard display) and things to get you started like a USB driver stick and SATA cables. Below is a complete list of the included accessories.

• M-Vision dashboard
• Wi-Fi antenna
• USB drive (drivers)
• M-Vision cable
• (3) SATA cables
• 1 to 2 RGB LED extension Y cable
• ARGB extension cable
• 1 to 3 ARGB Gen 2 extension cables
• (2) DP to Mini DP cables
• EZ Front panel cable
• Metal Dragon badge
• (4) M.2 screw/standoffs EZ M.2 clips
• Cable Stickers
• Quick Installation Guide

Design of the X670E Godlike

After removing the board from the retail packing, one of the first things I noticed outside of the stealthy black appearance is the sheer size and weight. It’s an E-ATX board nearly 12 inches wide, so you must ensure your case supports it. A backplate protects the back, doubling as a passive heatsink for the power delivery. The all-black appearance has a brushed aluminum finish on the heatsinks and shrouds, along with tastefully implemented branding on the IO area, chipset, VRM and M.2 heatsinks. The connections on the right side of the board are oriented horizontally, which helps with cable management.

Regarding RGB lighting, here again, only the IO area is illuminated. The colors are nice and saturated, but they won’t take over the inside of your chassis. If you need more, there are four headers to add lighting. Control over the RGBs is managed through the MSI Center Software and Mystic Light. Like its Z790 twin, I like the black-with-gold accents look. It matches most build themes and gives off that high-end vibe you want with a flagship-class motherboard.

Focusing on the top half of the board, starting with the left side above the IO area, is an Illusion Lighting effect that allows the board to display two distinct patterns (MSI Gaming Dragon and symbol) in the same space. I like this better than the dot-matrix look on the ROG Extreme. The VRMs connect via heatpipe with the top heatsink using a wavy fin design to increase surface area and dissipation properties.

To the right of the socket area and just above the DRAM slots are two 8-pin EPS connectors (one required) to power your AMD CPU. The four DRAM slots below support up to 128 GB of RAM, with speeds listed to DDR5-6666+(OC) (though that’s using one DIMM). During testing, we had no issues with our ‘base’ kits (DDR5-5600 and 6000). AMD tends to top out around 6200-6400 MHz, depending on the board and memory used.

Continuing to move right and in a unique location, we hit the first (of four) M.2 sockets. Hidden under a large flat heatsink, this CPU-connected socket runs at PCIe 5.0 x4 (128 Gbps) and supports up to 80mm modules.

In the upper-right corner is the first two (of nine) 4-pin fan headers. Each header supports DC and PWM modes with varying output. The CPU fan (auto mode by default) outputs up to 2A/24W, Pump fans up to 3A/36W, and the system fan headers at 2A/12W each (though the manual says 2A/12W). In short, there’s plenty of power to support your fans or a water-cooling pump for a custom loop.

Moving down the edge, we run into the first two RGB headers (in this case, two 3-pin ARGB), a 6-pin PCIe connector for the quick charging capability, 24-pin ATX to power the board, and finally, two front panel USB 3.2 ports with one 3.2 Gen 2 (10 Gbps) and the other 3.2 Gen 2x2 (20 Gbps). Oddly, there isn’t a 40 Gbps port on this board. Although many will not utilize that much speed, the lack of one of this platform’s new features on a flagship board just feels like an odd omission.

For power delivery, this Godlike, like its Intel-based twin, also has Godlike VRMs. Power travels down from the EPS connector(s) onto a Renesas RAA229628 controller to the 24x 105A Renesas RAA22010540 SPS MOSFETs in a teamed configuration. The 2,520A available for the CPU is the most I can recall seeing on an AMD motherboard and won’t get in the way of any type of overclocking, including using liquid nitrogen. You’re limited by the cooling method, just as you are with most mid-range boards. As awesome as this is for marketing, it raises the cost of the BOM, just for bragging rights.

On the bottom half of the board, we find the rest of the M.2 sockets, several headers, connectors, and even a few switches. The audio section on the left houses a Realtek ALC4082 codec and ESS ES9280AQ combo DAC/HPA. This is the flagship codec, with a DAC and Amp combo, and one of the better audio implementations for the AM5 platform.

Moving back to the M.2 sockets, three others are hidden below the heatsinks in this area. These M.2 sockets (M.2_2-4) connect through the chipset. Each socket supports up to 80mm modules with speeds reaching PCIe 4.0 x4 (64 Gbps). If you want/need more than the four onboard M.2 sockets, you’re in luck, as MSI includes the M.2 Expander-Z Gen 5 Dual AIC. The AIC supports up to two PCIe 5.0 x4 (128 Gbps) drives when connected to a PCIe 5.0 -capable slot (PCI_E2 on this board). The Xpander card has active cooling to keep hot-running drives within spec, and it even has an activity light so you can see when the device is in action.

Shifting focus past the chipset to the right edge, we see two USB 3.2 Gen 1 (5 Gbps) connectors and eight SATA ports. SATA ports 3/4 connect through the chipset, while SATA A1/A2 connect through the ASM1061 chip, and SATA B1-B4 connect through an ASM1064 controller. The X670E Godlike supports RAID0/1 modes for SATA devices on SATA P3/4, while the NVMe drives support RAID0/1/10 modes. 

If you plan on plugging in lots of drives, it’s worth mentioning here how things break down in terms of bandwidth. When the M2_4 (the PCIe 5.0 x4 socket) is populated, PCI_E2 (the second PCIe slot) is unavailable. If you’re using a SATA- or PCIe-based drive in M2_5, SATA ports 5-8 are unavailable, leaving you with just two functioning SATA ports available. 

The PCIe slots mixed in among the M.2 sockets are simple. Both are reinforced and PCIe 5.0. The top slot supports up to x16 speeds, but both run at PCIe 5.0 x8 speeds when the second slot gets populated. Last, along the right edge, are two USB 3.2 Gen 1 (10 Gbps) headers and the six SATA ports.

Across the bottom of the board are several exposed headers, all vertical. You’ll find the usual, and quite a bit more: additional USB ports, RGB headers, fan headers, custom water cooling headers (for flow and temperatures), power/reset buttons, and more. Below is a complete list from left to right.

• Front panel audio
• 4-pin RGB
• JDASH1 (M-Vison)
• 6-pin PCIe power
• LED and BIOS switches
• (4) 4-pin fan/pump headers
• 2-pin waterflow header
• (2) USB 2.0 headers
• (2) 2-pin temperature sensor headers
• Power/Reset buttons
• Front panel
• 4-pin RGB

The rear IO plate on the X670E Godlike comes preinstalled to the motherboard. It has a black background with light-grey labels on the ports, making them relatively easy to read. There are a total of nine USB ports:  two Type-C (20 Gbps) and seven 3.2 Gen 2 (10 Gbps) ports. On the left are three buttons: Clear CMOS, BIOS Flash, and the Smart button. The latter is programmable, letting you do things like reset the PC, boot into Safe Mode, control fans and lighting, etc. Around the USB ports are the Marvell 10 GbE and Intel 2.5 GbE ports. Last but not least are the Wi-Fi antenna connections and the five-plug plus SPDIF audio stack.

Firmware

MSI’s BIOS for the Godlike is not much different than the Z790 Godlike we reviewed previously. The menus reside on the sides, and information up top, with a black and red background. It’s a unique look compared to the competition. For X670E, the only updates were functionality for the new processors and chipset. The color scheme is still easy to read, and the informative Easy Mode displays various information about the system and allows for some changes (boot order and XMP enabled). Though different than other UEFIs, it is easy to get around; everything has a place, and most frequently used options are readily available, not buried within sub-menus.

Software

MSI has a single utility, MSI Center, that covers quite a bit of functionality. From hardware monitoring to RGB control with Mystic Light, there are many applets to choose from within the software and a one-stop shop to download all of the company’s utilities. I wish there were some overclocking options here, but for those looking to use the Gamebar feature, Super Charger, or any other utilities MSI offers, they will all be at your fingertips in MSI Center.

MORE: Best Motherboards

MORE: How To Choose A Motherboard

MORE: All Motherboard Content

Specifications: Gigabyte X670E Aorus Xtreme

Inside the Box of the Gigabyte X670E Aorus Xtreme 

Gigabyte’s X670E Aorus Xtreme, like it’s Z790 namesake, comes with plenty of accessories, including SATA cables, temperature sensors, Wi-Fi antennas, drivers, and even an external USB (Type-C) DAC for audio. Below is a complete list of the included accessories.

• (4) SATA cables
• Wi-Fi antenna
• PC speaker
• 4-pin RGB extension cable
• (2) temperature sensors
• User Guide
• Gigabyte Stickers
• Q-Connector
• ESSential USB DAC
• USB and Front Panel extensions
• USB drive (drivers)

Design of the X670E Aorus Xtreme

The X670E Aorus Master sits on a black 8-layer PCB covered in heatsinks and shrouds. Above the IO area is a large cover that displays the Aorus logo. Around the chipset area and PCIe/M.2 in the middle of the board is more Aorus branding. If you’re looking for RGB bling, there isn’t much here, just what you see above the IO area. However, you can connect your own to the integrated headers scattered around the board. Overall, the Aorus Xtreme has the premium look of a flagship board. The mostly black appearance gives way to grey and gold accents and fits well with most build themes.

Zooming in on the top half of the board, we get a better look at the Aorus Eagle RGB animation. Unlike Z790 with its OLED screen (why not on this board?), this is the only RGB lighting. The colors are saturated, but not too bright. Next, we get a better look at the oversized VRM heatsinks that are finned (Gigabyte calls it Fins Array 3) with a nano-carbon coating. Above this area are two reinforced 8-pin EPS connectors that power the CPU.

To the right of the socket area, we run into four reinforced DRAM slots with locking mechanisms on both sides. The Aorus Xtreme supports up to DDR5-6666+(OC) for the AM5 platform. We can’t tell you if the memory works at that speed (AMD typically tops out around 6400 MHz), but our test kits (DDR5-5600 and 6000) worked without issue.

Just above the DRAM slots are the first three (of 10) 4-pin fan/pump headers. Each header supports up to 2A/24W, which should be plenty for most fans and water pumps. You can control the attached devices through the Gigabyte Control Center software. Continuing right, we then find three of the four RGB headers, two 4-pin RGB and one 3-pin ARGB. 

Moving down the right edge, we see the two-character debug port, 24-pin ATX connector, fan headers, and  2-pin temperature headers. We also run into an area to read voltages and the 4-LED debug. Unlike the Z790 Aorus Xtreme, all these headers are visible down the right edge. I would like to see the shrouds/covers seen on the Intel version make their way to this one.

Regarding the power delivery, the 18-phases dedicated to Vcore aren’t the most we’ve seen, but it’s up there and can easily support the flagship Ryzen 9 7950X. Power comes down from the EPS connector(s) onto the Renesas RAA229628 controller. From there, the signal is sent to 18x 105A Renesas RAA22010540 SPS MOSFETs. The 1,890 Amps available can handle any compatible CPU, even with sub-ambient cooling.

On the bottom half of the board, we’ll start on the left with the audio. Hidden under the heatsinks is a Realtek ALC1200-VB codec and an ESS ES91180AC DAC. While this isn’t the latest and greatest, it is the last generation’s flagship codec. While I’d like to see the ALC4082 on this board, most people will be satisfied with last-gen flagship audio. If not, you can always get powered speakers with their own DAC, like the AudioEngine A2+.

Next up are the PCIe slots. The Aorus Xtreme has three total full-length PCIe slots. The top slot is reinforced and runs at PCIe 5.0 x16 speeds (but breaks down to x8 when a device populates an M.2 socket, specifically M2B and M2C_CPU slots). The other two full-length slots connect through the chipset and support PCIe 4.0 x4 and PCIe 3.0 x2, respectively.

Located around the PCIe slots are four M.2 sockets (sorry folks, no add-in cards for more). The top socket, M2C_CPU, is PCIe 5.0-capable and runs at x4 (128 Gbps), supporting up to 110mm PCIe drives. The PCIe 5.0 slot breaks down to x8 when this socket is used. The three other M.2 sockets also run at PCIe 5.0 x4 (128 Gbps) and support up to 110mm devices.

Looking past the chipset heatsink and onto the right edge, we see the four SATA ports, two 4-pin fan headers, a reset button, and a jumper. Again, I’d like to see more shrouds covering things up.

The rear IO plate on the X670E Aorus Xtreme is also preinstalled on the motherboard. It sports a black background with white labels on the ports, making them easy to read. There are a total of 12 USB ports: two USB 3.2 Gen 2x2 (20 Gbps) Type-C ports, four USB 3.2 Gen 2 (10 Gbps) ports, and four USB 2.0 (480 Mbps) ports.  On the left are two buttons: Q-Flash and Ignition (the latter to pre-test liquid cooling systems, also allowing for a firmware update in this state). Video output comes from the HDMI and DisplayPort ports. Last are the two Wi-Fi antenna connections and the 2-plug plus SPDIF audio stack.

Firmware

Gigabyte’s BIOS for the X670E Aorus Xtreme retains the same formatting sporting the black-and-orange theme Aorus is known for. The most significant differences are just with the options available for the new AMD processors. Outside of that, the black-on-white/orange/yellow provides a good contrast and is easy to read. Like most BIOSes, the Aorus Master has an informational Easy Mode and an Advanced mode with headers across the top. 

Navigating the BIOS is easy, as many frequently used options (including overclocking) are generally on the same page or one level down. My biggest complaint with the BIOS is that page-down doesn’t work in all sections, and you have to type in some values manually or bang on the down arrow to scroll and select a value. Aside from that minor annoyance, it’s a solid BIOS.

Software

On the software side, Gigabyte’s primary tool for X670E (and Z790/B760) is the Gigabyte Control Center (GCC). GCC is a one-stop shop for many board-centric functions, including RGB lighting control (RGB Fusion), Fan Control, and performance/overclocking. You can also update and install drivers from within the application, too. It’s not as comprehensive as the App Center and some of its applications, but I doubt many will miss the old version, as GCC is quite capable.

MORE: Best Motherboards

MORE: How To Choose A Motherboard

MORE: All Motherboard Content

Specifications: Asus ROG Crosshair X670E Extreme

Inside the Box of the X670E Crosshair Extreme

Opening up the Asus ROG Crosshair X670E Extreme box, we’re greeted by the motherboard on top with a comprehensive accessory organized neatly below. Asus provides the ROG GEN-Z.2 (a PCIe 5.0 M.2 card), True Voltician (an Asus fan controller) along with a slew of cables, thermistors, screw packages, and more (don’t worry, we took the kitchen sink out for the review). Below is the full (and very long) list from the Asus website.

• 1-to-3 ARGB splitter cable
• 1-to-2 ARGB splitter cable
• (2) 1-to-4 fan splitter cable
• (3) ROG weave SATA 6G cables pack
• 3-in-1 Thermistor cables pack
• ROG Fan Controller
• Fan EXT PWR cable
• ARGB input cable 
• USB input cable 
• ROG Fan Controller 3M Mounting tape package
• Fan Controller manual
• ROG GEN-Z.2 with heatsink
• M.2 pad package for ROG GEN-Z.2
• (2) M.2 screw package for ROG GEN-Z.2
• PCIe 5.0 M.2 Card with heatsink
• M.2 screw package for ROG PCIe 5.0 M.2 Card 
• ROG True Voltician
• (4) cables for ROG True Voltician
• Thermal pad for M.2
• ASUS Wi-Fi moving antennas 
• Rubber Package for M.2 backplate
• Q-connector
• (2) M.2 Q-Latch packages for M.2 backplate
• ROG Graphics card holder
• ROG key chain
• ROG logo plate sticker
• ROG screwdriver
• ROG stickers
• ROG thank you card
• USB drive with utilities and drivers 
• User guide

Design of the X670E Crosshair Extreme

When taking the Asus flagship board out of the box, it’s instantly clear it’s a premium motherboard. From the sheer weight to the all-black motherboard that’s mostly covered in shrouds, it gives off a high-end vibe. The black VRM heatsinks are large and heavy, with mitered cutouts to increase surface area. Above the IO area, Asus uses the unique dot-matrix-like RGB display that rotates through a few default branding images. The right edge sports a shroud that covers all of the ugly headers and connectors along that edge. The bottom half of the board is also covered with heatsinks and shrouds. And above the chipset, another dot-matrix RGB area sports the ROG branding.

Starting with the top half of the board, we get a closer look at the RGB display and the massive heatpipe-connected heatsinks. Compared to the OLED screens we’ve seen on some flagship boards in the past, I’m not a fan of the dot-matrix look. However, the light it gives off is bright, and the colors are saturated, so it will still look good inside your chassis. Just above the VRM heatsink on the left are two 8-pin EPS (one required) connectors to power the processor. 

Also, on the board's top edge, we run into our first three (of eight) 4-pin fan headers. All headers support PWM and DC-controlled fans/pumps. All CPU, Chassis, and Radiator headers are Q-fan controlled by default and support up to 1A/12W output. The two water pump headers (W_PUMP+1/2) handle up to 3A/36W and run out of the box at full speed.

Just below the fan headers are four reinforced DRAM slots, with a single locking mechanism on top. On top of the 128GB capacity, Asus lists support of up to DDR4-6400+(OC), the lowest of the flagship-class boards we’ve looked at. That is, seemingly, the maximum on most AMD boards anyway, so I wouldn’t worry about anything faster in the first place. The Crosshair Extreme handled our DDR5-5600 and DDR5-6000 kits without issue during our testing.

Normally, these high-end Asus boards have the DIMM.2 slot for additional M.2 support. However, with this generation, Asus uses a proprietary slot called GEN-Z.2 that adds Gen 5 and Gen 4 M.2 slots inside a heatsink shell. The Crosshair Extreme supports five M.2 modules using the onboard sockets and the two add-in cards (the other is a more standard AIC that connects through the PCIe slot).

The two-character debug LED and four-LED EZ debug features are on the right edge and help troubleshoot the POST process/non-boot situations. Moving down, we hit the start (power) and Flex buttons, a 3-pin ARGB header, two 4-pin fan headers, the 24-pin ATX power connector, the 6-pin PCIe PD connector for fast charging devices through the front panel Type-C port, and finally, a front panel USB 3.2 Gen 2x2 (20 Gbps) header that sits vertically.

Power delivery on the Crosshair Extreme is, fittingly, one of the more robust we’ve run across for the AM5 platform. Power comes from the EPS connector(s) onto the Digi VRM controller (ASP). From there, it moves on to the 20x SiC850A 110A SPS MOSFETs. The 2,200A available for Vcore is the second highest we’ve seen. You’ll only be limited by your cooling method on this board. It will handle ambient and sub-ambient overclocking without the VRMs holding things back.

On the bottom half of the board, we’ll start with the audio on the left side. Hidden under the heatsinks is a flagship-class Realtek ALC4082 codec. In addition to the audio separation line and dedicated audio caps, it also comes with an ESS SABRE9218C DAC and a USB Type-C to 1/8th inch DAC for non-USB headphones. If this audio solution isn’t good enough, you’re likely an audiophile with some seriously expensive equipment.  For most users, this is more than good enough and about the best you can get on a modern motherboard.

In the middle of the board are two full-length PCIe slots, an x4 size slot, and two M.2 sockets. Both full-length slots are reinforced and connect via the CPU to offer PCIe 5.0 bandwidth. The top slot runs at x16 speeds, and the second full-length slot has a maximum of x8. With both active, it splits to x8/x8. The bottom x4 slot connects through the chipset and runs at PCIe 4.0 x4 speeds.

Mixed in among the shrouds are two PCIe 5.0 x4 (128 Gbps) M.2 sockets that support up to 80mm modules. There is some lane sharing among the PCIe slots and M.2 sockets. If M.2_2 is enabled, PCIEX16_1 (top slot) runs at x8, and the bottom slot runs at x4 speeds. You’ll be able to run all six SATA ports and all five M.2 sockets concurrently; you’ll just lose some bandwidth on the PCIe slots.

The rear IO plate on the Crosshair Extreme is busy. It comes attached to the board out of the box, which is what we expect from most higher-end boards these days. There are 10 total USB ports out back, with two USB 4 ports Type-C ports (JHL8540 USB4 controller), a 3.2 Gen 2x2 (20 Gbps) Type-C port, and a 10 Gbps Type-C port. In addition to the Type-C ports are eight other USB 3.2 Gen 2 (10 Gbps) ports. On the far left are the BIOS Flashback and reset CMOS buttons, while on the right are the Wi-Fi antenna connectors and the five-plug plus SPDIF audio stack (each analog port is backlit with matching color RGB). Last but not least, mixed in among the USB ports, are the 2.5 and 10 GbE ports.

Firmware

Asus’ BIOS on the X670E Crosshair Extreme as we’re used to with other X670E. 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.

Software

Asus’ software suite is still the same as the Armoury Crate and AI Suite, but like the firmware, it gets updated to support the AMD processors. There are several applications 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.

MORE: Best Motherboards

MORE: How To Choose A Motherboard

MORE: All Motherboard Content

Test System / Comparison Products

We’ve updated our test system to Windows 11 64-bit OS with all updates applied. We kept the same Asus TUF RTX 3070 video card from our previous testing platforms but updated the driver. We also updated to F1 22 for our games and kept Far Cry 6. Unless otherwise noted, we use the latest non-beta motherboard BIOS available to the public. The hardware we used is as follows: 

Test System Components

 EVGA supplied our Supernova 850W P6 power supply (appropriately sized and more efficient than the 1.2KW monster we used previously) for our test systems, and G.Skill sent us a DDR5-5600 (F5-5600U3636C16GX2-TZ5RK) memory kit for testing. 

Benchmark Settings

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. For this baseline testing, the Windows power scheme is set to Balanced (default), 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 either stability or performance, though, and those settings can impact some testing.

Across our synthetic benchmarks, all these boards performed well, ending up around the average when it’s all said and done. In several instances, the Gigabyte X670E Xtreme was the faster of our group, though again, not by much.

Timed Applications

In our timed applications, both of the ‘Extreme’ boards led the MSI Godlike. Although the difference wasn’t big, it was still a trend in these timed applications (and several others below). That said, none of the results, slower or faster, are significantly out of line, and all of the boards were performant in these tests.

3D Games and 3DMark

Starting with the launch of Zen 4, we’ve updated one of our test games, F1 21 to F1 22, while keeping Far Cry 6 in our testing suite. 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.

Looking at the results from our gaming tests, our flagships did well here again, posting average or above-average results in the 3DMark tests and F1 2022, but merely average among all results in Far Cry 6. Like the Z790 flagships, these are competent gaming boards too.

Overclocking

When overclocking, we aim to increase the power and add stress to the VRMs. We do so by increasing the clock speed and voltage until we’re at the thermal threshold for the processor during stress testing. However, where these CPUs are allowed to run with ‘boost’ speeds, those speeds are closer to the limit than ever before.

Overclocking with all three boards was simple and easy. We plug in our values, set the voltage to manual, adjusted LLC if needed, and off we went. Once in Windows, we applied the stress test load and successfully finished our short test on all these boards. No fiddling was needed outside of the usual suspects (voltage, multiplier, and LLC adjustments).

On the memory side, we dropped our Kingston Fury Beast DDR5-6000 kit in the systems, enabled XMP, and tested it without issue, as expected. The boards list speeds above this, but your mileage may vary, as reaching speeds past this depends on the memory kit, the quality of the IMC, and the motherboard/BIOS. Quite frankly, most boards top out around DDR5-6400 anyway. 

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 use only the stock power use/VRM temperature charts, as this section aims to ensure the power delivery can handle the chip even when overclocked. Since we’re using less power to get more clocks, those datasets are more novel than useful. We’re also temperature limited on the processor, so adding more power isn’t possible without increased throttling. 

The power consumption difference between our three review boards was around 10%, with the Godlike becoming our first AM5 board to average over 200W between idle and a stress test load. The Godlike idled higher (103W compared to 84W and 82W) and used slightly more power under load (300W compared to 284W and 283 for the Crosshair Extreme and Aorus Xtreme, respectively) to reach this distinction.

VRM temperatures on the boards were more than acceptable, keeping the overkill power bits underneath running well within specification. In fact, none of the boards broke 50 degrees Celsius in our stress testing, which is quite impressive. These boards won’t have any issue driving AMD’s processors with any cooling option.

Bottom Line

Right out of the gate, all of these flagship boards offer the best of the best for the current AMD platform. Each board includes overkill power delivery, 10 GbE ports and Wi-Fi 6E, PCIe 5.0 slots and M.2 sockets, and a high-end appearance. From there, we find a few differences, with most (but not all) boards implementing flagship-class audio and some using 40 Gbps USB ports. Still, which is better for you is generally going to come down to price and appearance, unless you absolutely need those features.

That out of the way, let’s address the elephant in the room: pricing. At the time of this writing, The MSI X670E Godlike at $1299.99 is, by far, the most expensive AM5 board. Asus’ ROG Crosshair X670E Extreme checks in at $999.99, while the Gigabyte sits at $699.99. A dark horse in the race is ASRock’s flagship, the X670E Taichi, at $479.99. Frankly, the Taichi is arguably the most well-rounded and is, by far, the least expensive option. It has everything the more expensive boards do, including 40 Gbps USB ports, the latest generation audio codec, overkill power delivery, and even the look.

So which flagship board should you choose? That’s a tough one, as each has redeeming qualities. The MSI X670E Godlike and its six total M.2 sockets are on the table if budget isn't a concern, but that price needs to come closer to $1000 as there’s nothing it offers more than the other boards outside of additional M.2 sockets. The Asus is a great option for those looking for all the bells and whistles but unwilling to pay for the MSI. Gigabyte’s Xtreme is a great option as it’s reasonably priced (for a flagship) but does not have 40 Gbps ports or current-gen flagship audio. Chances are that’s not a deal breaker for most users, however. If you’re looking to get the most bang for your buck out of these flagships and don’t require more than four M.2 sockets, ASRock’s X670E would be my weapon of choice in the battle of these behemoths. That’s one of the reasons we’ve already given it a full review and placed it in the number two spot on our best motherboards list.

MORE: Best Motherboards

MORE: How To Choose A Motherboard

MORE: All Motherboard Content

Nvidia's latest driver, version 531.18, appears to have a bug that causes high CPU usage after exiting games. However, the company admits that the problem exists, so expect Nvidia to issue a hotfix or a new driver version that fixes the annoying bug. 

As it turns out, Nvidia's Display Container Service causes 10% - 15% higher CPU usage after a game is closed, which can be observed in Windows Task Manager. Some Reddit users believe the bug was caused by Nvidia's Game Session Telemetry Plugin version 531.18 (NvGSTPlugin.dll).  

It should be noted that not all systems are affected by the bug. For example, our colleague Andreas Schilling tried to replicate the issue but failed. 

"I tried to recreate the issue on two systems with the latest 531.18 drivers and was not able to get the NVIDIA container task going crazy like this," he wrote in a Twitter post. 

Some Reddit users suggest blocking/removing NvGSTPlugin.dll to fix the problem manually, but it is unclear what side effects this remedy may cause, so this solution can hardly be recommended. Therefore, the easiest fix is to use the previous 528.49 WHQL driver that may not support some of the latest games, but at least it works without significant problems. 

While Nvidia formally admits the problem, the company didn't reveal when it plans to release a hotfix or a new version without this bug. Meanwhile, given that the bug is rather annoying, it is likely that Nvidia is inclined to release a hotfix sooner rather than later, so perhaps it makes sense to wait a little bit before rolling back to a previous version of Nvidia's driver.

It takes time to reproduce bugs and find solutions, especially rare ones. However, after a couple of weeks of investigation, AMD finally found a solution for Radeon graphics card owners who faced an Adrenalin WHQL driver bug that corrupted their systems.

AMD's Adrenalin 23.2.1 driver delivered decent performance improvements and new features to users. But, lamentably, the driver also brought some users headaches since it bricked their Windows installation. 

The issue isn't limited to the Adrenalin 23.2.1 driver, either. Brad Chacos, the executive editor at PCWorld, encountered the same bug even with the latest Adrenalin 23.2.2 driver that recently left AMD's oven. Chacos worked hand-in-hand with AMD engineers to discover the root cause of the problem and a temporary workaround.

"We have reproduced an issue that can occur in an extremely small number of instances if a PC update occurs during the installation of AMD Software: Adrenalin Edition, and we are actively investigating. We recommend users ensure all system updates are applied or paused before installing the driver, and that the "Factory Reset" option is unchecked during the AMD driver installation process. We are committed to resolve issues as quickly as possible and strongly encourage users to submit issues with AMD Software: Adrenalin Edition via the Bug Report Tool," one AMD spokesperson told PCWorld.

Judging from AMD's statement, the software has a flaw that appears to conflict with Windows Update. However, you can argue that it's a Windows bug, not an AMD bug. Windows shouldn't be updating a driver of any kind if there's a current installation in progress. It has been an issue with all sorts of other drivers for quite some time. It's the reason why some users prefer to disable automatic driver updates.

In the meantime, AMD has advised users not to use the "Factory Reset" option while installing the Adrenalin drivers. The aforementioned option removes all traces of the previously installed driver, similar to what popular third-party Display Driver Uninstaller (DDU) does, and requires a restart. Our theory is that Windows Update simultaneously installs the Adrenalin driver during the standard driver installation because the system momentarily lacks a display driver when restarted.

It's good practice to always do a clean install of your display drivers with every update. However, the feature is currently borked on AMD's Adrenalin drivers, so heed the company's advice and make sure you untick the "Factory Reset" option before installation. Also, for good measure, disable or pause your Windows updates before installing the Adrenalin drivers. AMD believes it's an uncommon bug, but we don't recommend taking any chances until the chipmaker provides a final solution.

DDU, for example, has a specific option that prevents Windows Update from downloading drivers when the program is running. AMD could adopt a similar strategy to ensure that Windows Update and the Adrenalin software don't perform a parallel installation.

Windows recovery mode is the answer for those unlucky users that have already gone down this road. However, it may not be easy to get there.

According to Chacos, AMD recommended hitting the power button to force Windows into automatic recovery. It's easier said than done, though. The trick is to hit the power button right after the BIOS options disappear but before the Windows circle. Unfortunately, Chacos noted that pushing the power button during the BIOS options would turn off the system and at the Windows spinning circle only lead to a black screen. It took Chacos 15 attempts to succeed, so perseverance is key to recovering your system.

If you're not the type to keep regular data backups, you should probably do so before installing AMD's latest Adrenalin drivers. It's a rare bug, and unticking "Factory Reset" seemingly prevents it from happening, but do you want to take that chance?

Intel’s Arc Alchemist GPUs launched toward the latter part of 2022, vying for a spot among the best graphics cards. You can see where they land on our GPU benchmarks hierarchy as well, but that's only part of the story. Intel has been busily updating drivers on a regular basis since launch, and rarely does a fortnight pass without at least one new driver to test.

And that's the crux of the story: How much better are Intel’s latest drivers compared to the original launch drivers? The company made a lot of noise about its improved DirectX 9 performance, claiming 43% higher performance on the A750 with recent drivers compared to the original 3490 launch drivers — and at the same time dropping the price of the A750 to $250 down from $290. What about newer games, though?

We set about testing (and retesting) every Arc graphics card, including the Arc A770 8GB using an ASRock model that we haven't quite got around to reviewing just yet, due to all the retesting that's been going on.

We began our testing a couple of weeks back, right after the 4123 Intel drivers became available. Last week, version 4125 drivers came out, but a quick check on one of the Arc cards indicates the only changes are related to some recent game launches and that the performance otherwise remains the same as the 4123 drivers for our test suite.

Update: Intel just released the 4146 drivers on February 28, 2023. These fix a problem that we noticed with Red Dead Redemption 2 crashing on the Arc A750, among other things. We have included the results for that one game from the newer drivers (as before we couldn't get a result).

Plugging the Holes in the Arc

Arc arrived with what were arguably the best GPU drivers Intel has ever created – but that’s not saying much. A few years ago, a lot of games simply refused to work at all on Intel’s integrated graphics solutions — and even those games that worked would often perform poorly. Intel’s DG1 helped pave the way for more frequent driver updates, but things were still iffy back in 2021.

Since the first dedicated Arc A380 cards started showing up in China — which meant they also started getting shipped to individuals around the world — Intel has been cranking out updated drivers on a regular basis. There are presently eleven different versions of Arc drivers available from Intel: five WHQL (Windows Hardware Quality Labs) certified and six beta drivers, but there were other “hotfix” or beta drivers that are no longer listed.

Many of the updates have been targeted at one or two specific games — there were two different drivers that addressed problems with Spider-Man Remastered performance on the Arc A380, for example. Other updates have had much further reaching ramifications, with the biggest change being DirectX 9 optimizations.

DirectX 9, 20 Years Later

Microsoft’s DirectX 9 API became publicly available back in 2002, but it continues to see quite a bit of use even today. Some of that goes back to the venerable Valve Source engine used for Half-Life 2, which saw widespread adoption for various mods and spinoff games like Counter-Strike: Global Offensive, Left 4 Dead 2 and Team Fortress 2; StarCraft 2 and League of Legends likewise continue to use DX9.

Considering how old the API is, it’s easy to understand Intel’s decision to “focus on modern [DX12 and Vulkan] APIs” for the launch of the Arc graphics cards. Most recently released games use DirectX 12 or DirectX 11, or sometimes Vulkan. Why put a bunch of effort into trying to optimize the drivers for older games? For DX9 compatibility, Intel opted for support via DX12 emulation, which probably sounded like a good idea at some point.

The problem is that a lot of older games remain immensely popular. Token support for DX9 via DX12 emulation might have sounded fine on paper, but in practice it was a glaring indication that Intel’s GPUs and drivers weren’t as good as AMD and Nvidia drivers. That’s even though most DX9 games aren't particularly demanding, especially when compared with recent DX11/12 releases, performance in DX9 is only as good as the emulation.

Simply put, at launch, the DX9 via DX12 emulation for Intel Arc wasn't very good. It was very prone to fluctuations in frametimes, leading to stuttering in games. People complained and Intel decided to rethink its DX9 strategy. The solution Intel came up with was to leverage more translation layers to turn DX9 API calls into either DX12 calls or Vulkan calls. The DX12 emulation still comes from Intel’s internal work on the drivers, or perhaps Microsoft’s D3D9On12 mapping layer, but now with a bit more tuning.

The bigger change appears to be the use of DXVK — DirectX to Vulkan — from Steam for “some cases.” Intel hasn’t detailed exactly which games use DXVK and which use DX12 emulation, but whatever the changes, frametimes and average performance have increased substantially. This was all rolled out with Arc’s December driver update.

This is great news if you want to buy an Arc GPU for a modern PC to run old games. Remember: PCIe Resizable BAR is basically required for Arc GPUs, which means you generally need a 10th gen or later Intel CPU, or Ryzen 5000 or later AMD CPU. Otherwise, you’ll lose about 20 percent of the potential performance, based on testing, at which point you should just buy an AMD or Nvidia GPU.

So Intel can't simply ignore older games, as weak performance in games that millions of people still play on a regular basis makes all Intel GPUs look bad. But how much has really changed?

Arc Performance Update: Test Setup

Given all the changes, we pulled out a collection of fifteen different games, including two DX9-based games, to check how the various cards perform. The fifteen games and their respective APIs consist of Borderlands 3 (DX12), Bright Memory Infinite Benchmark (DXR, or DirectX Raytracing), Control Ultimate Edition (DXR), Counter Strike: Global Offensive (DX9), Cyberpunk 2077 (DXR), Far Cry 6 (DX12), Flight Simulator (DX11), Forza Horizon 5 (DX12), Horizon Zero Dawn (DX12), Mass Effect 2 (DX9), Metro Exodus Enhanced (DXR), Minecraft (DXR), Red Dead Redemption 2 (Vulkan), Total War: Warhammer 3 (DX11), and Watch Dogs Legion (DX12).

All of the games were tested at "medium" and "ultra" settings, except for Mass Effect 2 — it only has three graphics options, and it's old enough that we simply enabled all three options for the "ultra" testing. The three A7-class cards were tested at 1080p, 1440p, and 4K resolutions, except for in ray tracing games where we dropped 4K testing (for what will become obvious reasons in a moment). The A380 dropped the 4K testing completely, and also dropped 1440p testing for ray tracing games.

Each game and setting were tested at least three times, dropping the high and low result. Where there was more variance (looking at you, CSGO), the tests were done five times and we took the median result. This is an important point, because sometimes the first run was very poor (again, CSGO, particularly on the 3490 drivers).

All these tests were done using the same Core i9-12900K PC that we used for the previously published Arc GPU launch reviews, so that we can look at how performance has changed. How would the old drivers differ from the launch performance and now? That's where things get a bit messy.

There have been multiple major updates to games in our test suite over the past few months, and we've been working to shift all of our testing to a new system with a Core i9-13900K, plus adjusting our test suite and settings as needed. Our older test system is now effectively (slightly) out of date, and while we've retested some cards with the updated games, it was best to just hit the reset button.

What we know for certain is that Cyberpunk 2077, Flight Simulator, Forza Horizon 5, and Total War: Warhammer 3 all had updates that changed performance and/or settings quite a bit — on some GPUs more than others. Cyberpunk now has DLSS 3 and FSR 2.1 support, Flight Simulator added DLSS 3 and DLAA, and Forza Horizon 5 added DLSS 2 and FSR 2.2 along with TAA. All three of those games also added support for Nvidia Reflex. As for Total War: Warhammer 3, a couple of months back there was a major update that improved performance by roughly 20% on virtually all GPUs.

And those are just the changes that were readily visible. Almost all of the other games in our test suite have also received various updates, and keeping track of what has changed and what remains consistent is difficult. Regardless, we'll have the original launch performance data in our charts, with an asterisk indicating it's from a different version of the game and may not reflect current performance. Just for good measure, we also applied all the latest Windows 11 updates and flashed our motherboard BIOS.

And somewhere along the way, we ran into a problem. Windows 11 defaults to Virtualization Based Security (VBS) being on, and we previously turned it off. Some time in the past few months, it got turned back on (likely when I was poking around at Stable Diffusion and tried using WSL2). You can still turn this off, but we didn't realize it was on until after testing was completed. So, if you saw this article earlier and we talked about how performance dropped about 7% on average since our launch testing last year, that's the explanation. We retested the A750 with VBS disabled and found that, in general, performance matched or exceeded our October 2022 results and was around 7% faster than our current VBS-enabled testing.

Intel Arc Graphics Performance: 3490 Launch vs. Current 4123 Drivers

We have two results for each of the test cards in the following chart: 3490 are the launch drivers for the A770 and A750 — the Arc A380 was available several months earlier with 3259 drivers, and we saw 3267, 3268, and 3276 drivers released before our review was finished. But we're sticking with the 3490 drivers for all of the Arc "launch" drivers to keep things consistent. 4123 are the "current" drivers (4125 drivers are "Game On" for Company of Heroes 3, The Settlers: New Allies, Atomic Heart, and Wild Hearts; more recently, 4146 are "Game On" for Destiny 2: Lightfall and Wo Long: Fallen Dynasty). Let's start with the 1080p medium test suite.

We should also discuss some of the anomalies we'll see before we get to the charts. First, Red Dead Redemption 2 using the Vulkan API consistently crashed our test PC with the A750 card — and it crashed a different test PC as well. It was mostly stable using the original 3490 launch drivers, enough so that we could complete a few test runs before the PC would restart, but every later driver we checked had stability problems. Intel was able to replicate the problem and it’s limited to the A750; a fix was made available in the latest 4146 drivers, and we've included those results here (even though they're still labeled "4123" for our charting purposes).

Next, Bright Memory Infinite Benchmark, a ray tracing title, has a rather severe VRAM memory leak problem — or at least, that's how it appears. We could complete one benchmark run at 1080p with the Normal preset on the 8GB cards at 1080p, but higher settings (and the A380) would go from being somewhat smooth to a stuttering mess partway through the benchmark, with minimum framerates of 1 fps. This is another bug that Intel has confirmed and is working to fix — it's not clear if this affects the game as well, or only the standalone benchmark (which is far more demanding and better looking than the actual game, if you were wondering).

Finally, Minecraft performs very poorly on all Arc GPUs right now. When the cards first launched, you couldn’t even turn on ray tracing in the game, as apparently it was coded with a “whitelist” of cards that support DXR. An update to the game in December finally allowed the Intel GPUs to run with ray tracing enabled, but the results remain much lower than expected. As a point of reference, in other demanding ray tracing games like Cyberpunk 2077, the A770 is only 10 percent slower than an RTX 3060; in Minecraft, the RTX 3060 is currently nearly triple the performance. In fact, even the RX 6600 outperforms the fastest Arc GPU by 25 percent, where the A770 is 60 percent faster in Cyberpunk. Intel is also looking into the Minecraft performance, so hopefully that improves with a future driver.

Disclaimers aside, here are the results of our testing.

Overall performance changes between the 3490 and 4123 drivers at 1080p medium are… not a lot. That's pretty much expected. The three A7-class cards are 1–2 percent faster, while the A380 is 5% faster.

We don't have a "medium" result from Mass Effect 2, and CSGO appears to be mostly CPU limited to around 360 fps (with fluctuations as it's a non-deterministic test sequence — I played against harmless bots on the Mirage map, running the same route each time, basically a loop to the right through bomb site A as a counter terrorist). That means two of the biggest potential gains are out of the picture, except on the A380 where CSGO performance was 16% higher on the updated drivers.

Flipping through the individual charts, the only major differences are the somewhat larger than expected drop in performance in Control on the two Intel Arc Limited Edition cards — the ASRock A770 8GB and A380 didn't show such behavior — with modest gains elsewhere. That's mostly the same pattern we'll see at the other test settings.

Moving the quality setting to "ultra" results in bigger gains overall for the latest drivers, but much of that comes from the two DirectX 9 games. Overall, we're looking at a 5–10 percent aggregate improvement in performance with the 4123 drivers compared to the 3490 launch drivers. However, if we drop the two DX9 games from the geometric mean calculation, then the net improvement is only 3–6 percent.

There are positive and negative changes once again, with some of the biggest deltas coming in DXR games where they're often meaningless — A380 benchmarked at 6.1 fps compared to 4.5 fps, is a 34% increase in performance… at completely unplayable framerates.

There are a few cases where we see up to a 10% improvement in games that aren't using DXR or DX9, like Total War: Warhammer 3, but overall the changes are mostly within the margin of error, or at least close to it. But look at the two DX9 games… or at least the one DX9 game.

Mass Effect 2 performance improved by anywhere from 40% on the A380 to as much as 90% on the A770 8GB. That's for average fps, but the 99th percentile fps shows even bigger gains. The low fps results are 60% faster with the 4123 drivers on the A380 and up to 135% higher on the A770 8GB.

What about CSGO? The A770 16GB card shows basically no change, which makes sense given the large amount of VRAM. The A770 8GB improved by 9%, and the A750 by a few percent, but the A380 performance was slightly slower. Perhaps Intel focused its DX9 optimizations more on the bigger chips than on the ACM-G11 used in the A380? There's also the question of what map to use and how to test, so perhaps other CSGO maps would show different performance results, or maybe playing online rather than against bots.

Moving to 1440p ultra testing, the A380 is now mostly not a factor to consider outside of playing older DX9 games. Overall, we again see a 5–10 percent improvement in performance for the three A7-class Arc cards, with the A750 showing the biggest overall increase. Of course, if we omit the DX9 games, it goes back to being almost margin of error, 2–3 percent, with the A750 still improving by 9% thanks to a big increase in the Cyberpunk result, from an unplayable 10 fps to a still-unplayable 14 fps.

The gains in CSGO are still pretty muted, 3% on the A770 16GB, 10% on the A770 8GB, and 17% on the A750 — and the A380 performance again dropped, this time by 6%. Mass Effect 2 on the other hand shows gains of 50% on the Intel Limited Edition cards, and 70% on the ASRock A770 8GB — but only 22% on the A380, which still manages to break 100 fps on the 13-year-old game.

There's not much more to add. Outside of the two DX9 games, most of the changes aren't something you'd really notice as a gamer. It's worth mentioning that improvements in minimum FPS on the DX9 games are often even larger than the average FPS gains — CSGO had a lot of stuttering on the first couple of test runs with the 3490 drivers, and that's mostly gone now, but our use of the median result of five runs partially obscures this fact.

Last we have the 4K testing results, though we didn't do any DXR tests so the overall chart is down to 10 rasterization games. The three ACM-G10 based graphics cards show 3–7 percent aggregate gains in performance using the full test suite, but only a 1–5 percent increase if we omit the two DX9 games.

Mass Effect 2 now shows improvements of 10–35 percent, with the 16GB card benefiting the least, and the factory overclocked ASRock card benefiting the most. CSGO does show modest gains as well this time on the two Intel Limited Edition cards, while the ASRock card only shows a 3% improvement. Again, there was a lot of variability between runs in CSGO, and the minimum fps does show much better results with 35–50 percent higher 99th percentile framerates.

Playing Intel Arc, the Iteration Game

It's been an interesting six or so months since the Arc A380 first appeared. Given all the claimed performance gains, we were hoping to see more… consistency across our test suite. The latest drivers generally do outperform the launch drivers, but some issues remain, and that's on a still relatively limited test suite of 15 games.

None of our findings are particularly surprising, including the continued oddities. AMD and Nvidia have been playing the drivers game for decades, and even they have occasional problems. Intel has had graphics drivers for integrated graphics solutions for decades as well, but the performance on tap was so low that there often wasn’t any pressure to even try to optimize for newly released games. Creating dedicated graphics solutions changes the user expectations, and now Intel is playing catch up.

The most important things to note with the driver updates coming out of Intel are that there’s a regular cadence, and that older DirectX 9 games got some much-needed TLC to get them running more smoothly. It’s rare that more than two weeks passes without some new driver being announced, and Intel has also been doing better on getting Game Ready drivers out for bigger launches, including "Game On" drivers for the Hogwarts Legacy and Atomic Heart launches over the past couple of weeks.

If Intel can keep that up, plus add the occasional larger overhaul that provides more universal improvements, in a couple of years we hopefully won't even need to have a serious discussion about Intel’s GPU drivers. And of course, we're only scratching the surface with our game performance testing.

The games, settings, resolutions, and even test sequences can absolutely affect performance and potential gains from newer drivers. Maybe a different CSGO map would have yielded bigger performance improvements from the newer drivers. Or maybe the 12900K CPU was holding us back a bit more than the 13900K Intel used in its testing. We're looking into our CSGO results and may do some further testing and update this article as appropriate.

Regardless, let’s hope that Intel’s GPU and driver teams get the needed time to continue working on future drivers and architectures. Right now, it doesn’t sound like many people are biting on Arc graphics cards. Intel recently dropped the official price of its A750 Limited Edition card to $249, down $40 or 14 percent from the launch price. Big companies don’t do that if parts are flying off the shelves. How long can Intel continue to bleed money on consumer GPUs? Perhaps more importantly, can Intel afford to not invest more money into consumer GPUs?

There’s a lot riding on Arc Battlemage, which is currently slated to arrive in 2024. Intel's Raja Koduri has also said that the company is more interested in competing with mainstream GPUs than worrying about halo cards, with $200–$300 being the sweet spot. We’ll have to wait and see if Intel can narrow the gap between Arc and its AMD and Nvidia competition come next year — which, incidentally, is when we expect we'll see the next generation Nvidia Blackwell and AMD RDNA 4 GPUs launch.

• MORE: Best Graphics Cards
• MORE: GPU Benchmarks and Hierarchy
• MORE: All Graphics Content

AMD has thoughtfully shared a quickstart guide outlining how to set up a PC system with one of its alluring new Ryzen 9 7950X3D or 7900X3D processors. This guide is good for users, and good for AMD, as it could help reduce community discontent due to sub-optimal install and setup procedures. AMD’s guide isn’t hard to follow, but it could be time-consuming if followed to the letter - particularly going through the recommended Windows 10 or 11 fresh install.

Firstly, an updated version of Windows 10 or 11 is recommended. Specifically, AMD recommends a fresh install of at least “Windows 11 version 21H2 build 22000.1455 or Windows 10 version 1903 build 19044.2546.” If you don’t mind going through this drastic start afresh option, you might as well go for the latest releases. When you install the new OS, AMD notes that Windows Virtualization-based security (VBS) may be turned on, but that may mean L3 cache reporting is inaccurate. Microsoft has previously given the green flag to gamers wishing to speed up Windows by disabling VBS (and HVCI), so don't lose any sleep over disabling this security feature. 

Another essential preparatory step for new AMD Ryzen 7000X3D users is going to be updating the motherboard BIOS to the latest version, to ensure proper support. Users should look into motherboard compatibility and readiness before purchasing these CPUs. Luckily many modern motherboards (particularly expensive Ryzen 7000-supporting boards) allow BIOS updates without even having a CPU installed, so in many cases, this step can be completed without BIOS / CPU compatibility issue concerns.

AMD has prepared Ryzen chipset drivers so that its latest generation processors enhanced with 3D V-Cache can work optimally with your system. Look for AMD Chipset Driver 5.02.16.347 (or newer) for the best chance of optimal performance and stability. Please note that “using an older version of this driver will not enable X3D / Ryzen 9 game performance optimizations.” Some users have already noticed that if you switch between AMD Ryzen 7000X3D CPUs, it is necessary to reinstall the chipset driver to get the optimizations working again. Of course, you'll want to restart your system after the driver install completes.

The first boot after a fresh platform assembly can take two to three minutes, even on a fast modern system, so be patient with the new system as it performs first-boot memory training and other tasks in the background.

AMD also provides some Windows accessory software guidance for optimal Windows gaming performance. It asks that users enable Windows' Game Mode, as well as install and update the Xbox Game Bar App (version 5.823.1271.0 or newer). Some may have disabled or removed these apps previously, but they should follow AMD’s sage advice, as these Microsoft apps work with AMD’s new thread targeting tech to run gaming processes on the fastest cores, while parking the slowest cores if not needed. Our own initial tests of this adaptive technology didn’t unearth any problems.

AMD’s guide is based on its testing with both AMD Ryzen 7 7800X3D and Ryzen 9 7950X3D PCs using an AM5 Reference Motherboard, 32GB DDR5-6000 RAM, an Artic Liquid Freezer II cooler and Windows 11. The company made sure AMD Smart Access Memory technology was turned on, and VBS was off. A long list of modern games was tested.

Nvidia has released a new Game Ready driver for its GeForce graphics card users. Probably the most important thing Nvidia’s version 531.18 WHQL driver adds for content consumers is support for RTX Video Super Resolution. There are slimmer pickings for gamers, with just two headlining additions: some additional DLSS 3 optimizations for Atomic Heart, and support for The Finals closed beta. However, there are some worthwhile looking bug fixes for popular games and Adobe apps.

We provided some VSR testing and details earlier today, but put simply, VSR adds AI upscaling and enhancement smarts to streaming video consumption. It's tangentially similar to DLSS, though with different inputs and a completely separate trained network. Anyone with an RTX 30-series or newer GeForce card can give this tech a try. Just grab this new driver and the latest release of Edge or Chrome. (RTX 20-series support will come at a later date.)

Nvidia’s other headlining game picks aren’t awe inspiring. Firstly, the 531.18 WHQL driver “further optimizes DLSS 3” in Atomic Heart (where’s the ray tracing, Nvidia?). Second, a two-week run of a closed beta for The Finals (no, we're not going to put it all in caps) is getting support in this driver release. We haven’t mentioned this game before, so briefly: it's an upcoming free-to-play shooter set in a highly destructive world, where players team up in threes for tournaments. This new game will feature Nvidia technologies such as DLSS, Reflex and RTXGI. See it in action below.

Nvidia boasts that games supported by its GeForce Experience One Click Optimal Settings now exceed 1,000. If this feature works for your system, you might be happy to hear that seven new games have been added with this driver: Atomic Heart, Company of Heroes 3, Hogwarts Legacy, Perish, Sons Of The Forest, The Settlers: New Allies, and Warlander.

Bug fixes are always welcome, and in its release notes Nvidia admits to some juicy bugs being quashed. For example, freezes affecting Forza Horizon 4 (Steam) gamers have been eradicated, and stability issues in Call of Duty MW2 have been banished. Content creators may be more interested in the ironing out of wrinkles in Adobe After Effects and Premiere Pro applications.

You can grab the new 531.18 WHQL driver now manually, or leave it up to the GeForce Experience accessory app to prompt you. Whatever you choose, this is an 897.24MB download.

After being forsaken for over two months, AMD's Radeon RX 6000-series graphics cards, which are still some of the best GPUs, can finally welcome a new driver. Making good on its promise, AMD has deployed the Adrenalin Edition 23.2.1 driver for RDNA 2-powered graphics cards. It's not just for the RX 6000-series, though, as the unified driver also supports the RX 7000-series as well as RX 400-series and later GPUs.

The Adrenalin Edition 23.2.1 driver brings official support for Forspoken and Dead Space. But, of course, it wouldn't be a driver update without performance improvements. AMD claims that the Adrenalin Edition 23.2.1 driver offers up to 7% higher 4K performance on Forspoken and up to 4% on Hogwarts Legacy compared to the previous Adrenalin Edition 23.2.1 driver released in December 2022. Again, F1 2022 showed the most significant uplift, with AMD measuring a performance increase of up to 19%.

AMD utilizes the Radeon RX 6950 XT for the performance comparison between the two drivers (the 7000-series drivers previously received many of these updates), so your mileage may vary. In addition, the chipmaker focused more on Vulkan in this release. It also added support for the IREE compiler using the MLIR interface on Vulkan and several Vulkan extensions.

Adrenalin Edition 23.2.1 Performance Improvements

Radeon RX 6900-series owners will be happy to know they're finally receiving the features that Radeon RX 7900-series users have been enjoying. AMD delivers H.264 and H.265 streaming optimizations on the Adrenalin Edition 23.2.1 driver and a merited update to AMD Link. The chipmaker highlighted that users of Radeon RX 7900-series graphics cards would also need to download the new version.

The newest version of OBS Studio will come with an optimized encoder molded to AMD's Radeon RX 6000- and RX 7000-series graphics cards. In addition, the software brings AV1 encode acceleration support for the latter.

AMD fixed many issues with the Adrenalin Edition 23.2.1 driver, such as the Fortnite performance drops or stuttering in Sea of Thieves. Some known bugs remain, though. For example, the high idle power on Radeon RX 7000-series graphics cards related to high-resolution and high refresh rate monitors persists. Some users may still experience performance degradation on virtual reality games or applications. Again, you can find more information within the release notes.

Google has released an ‘Early Stable Update for Desktop’ which heralds the arrival of Nvidia’s RTX Video Super Resolution to the Chrome browser. This new video stream upscaling feature, using the same kind of AI that has made Deep Learning Super Sampling (DLSS) such a success in the games sphere, now only needs a new GeForce driver to hit the PC mainstream. However, before you read further, we think it is important to point out that Nvidia is only supporting this new AI video scaling tech on its GeForce RTX 30 and RTX 40 GPUs.

Nvidia RTX Video Super Resolution was announced at the CES 2023 last month. In brief, it takes in video at resolutions between 360p and 1440p (inclusive) and uses AI derived from deep learning and neural networks to upscale the video up to 4K. Of course, the special sauce delivered by Nvidia’s AI is a more impressive upscaling quality than what could previously be achieved by your hardware / software combo.

Above you can check out Nvidia’s infomercial for RTX Video Super Resolution, which was shared during CES 2023. In the video you see some game footage from Apex Legends captured at 1080p using H.264 at the 8mbps bitrate. Nvidia shows some illuminating split screen, and zoomed in detail comparisons - contrasting the gaming footage at 1080p and AI-upscaled to 4K. Gamers may appreciate that the RTX Video Super Resolution feature will support up to 144Hz video, too.

The results look good in the video, and there is no reason why they shouldn’t look impressive, with the similar DLSS having progressed through several major versions and refinement steps over the years. Applying the RTX Video Super Resolution to game footage might even show the tech at its best, and makes one wonder how well it would work in the wide range of online video genres to which it might be applied; nature documentaries or sports videos would present the Ai with very different upscaling challenges to something like Apex Legends.

VideoCardz spotted that Google’s latest stable version of Chrome (110.0.5481.77), now has the ‘NvidiaVpSuperResolution’ feature turned on by default – something that was previously a beta-only feature. The general release of the new stable Chrome 110 is expected tomorrow, February 7. As mentioned in the intro, Nvidia has yet to release a GeForce game ready or studio driver with support, but with any luck it should also arrive tomorrow, too. It would be fortunate to see Microsoft Edge updated for support simultaneously to give users a choice from day one.

With any upscaling technology, the proof of the pudding is in the watching, as all the major graphics chip vendor firms have been boastful about their various upscaling technologies previously – even when they weren’t anywhere as good as they have become recently. Fingers crossed we will be able to test Nvidia RTX Video Super Resolution, with the wide variety of streaming video content out there, very shortly.

AMD promises to release new drivers for its Radeon RX 6000 series and prior generations of graphics processors within a couple of weeks. The statement comes after the company did not release drivers for its previous GPUs over the past two months as it focused on the new Radeon RX 7000 series.  

"We are working on new AMD Radeon drivers for [Radeon RX] 6000-series and prior generation cards," said Frank Azor, AMD's Chief Architect of Gaming Solutions & Marketing, in a Twitter post. "Aiming to release them within the next two weeks. Will provide another update if we run into any delays and as we get closer to posting them. Thank you for your patience."

AMD launched its Radeon RX 7900 XT and RX 7900 XTX graphics cards in mid-December but had to fix bugs that inevitably emerged with the release of new hardware for existing games. As a result, AMD assigned as many software engineers as possible to support the hardware launch and drivers for its new RDNA 3-based products.  

Unfortunately, such a shift affected the company's ability to release new drivers for its existing Radeon RX 6000-series graphics cards. As a result, drivers for these products have not been updated since November, which has outraged many graphics card owners. The good news is that updated drivers are incoming. The bad news is that those cards will sit without new drivers for about three months by the time the new drivers emerge.

AMD's Radeon RX 6000-series boards are still among the best graphics cards money can buy today. While they may not be outright performance champions in ray tracing games, they are inexpensive and widely available. Furthermore, since AMD's Radeon RX 7900-series offerings only target the $899 to $999+ segment, people who are not willing to spend that much but want to stick to an AMD graphics board now have to buy something from the Radeon RX 6000 series. Therefore, timely driver releases for these products are essential to retaining customers reliant on robust customer support to play the latest games.

The Intel Arc Alchemist architecture brought a third competitor to the best graphics cards, and while it can't top the GPU benchmarks in terms of performance, there's certainly a strong value proposition. At the same time, there are areas where the drivers still need tuning — Minecraft with ray tracing being one example that comes immediately to mind since it's one of the games in our standard test suite. But Intel hopes to encourage further adoption with the latest announcement of a $40 price cut to the Intel Arc A750, bringing it down to just $250.

It's difficult to say exactly how many people have purchased Intel Arc graphics cards — mobile or desktop — since they first became available in the spring of 2023. I've seen claims that Intel dedicated GPUs accounted for up to 4% of total sales in Q4 2022, but so far Arc GPUs don't show up as individual entries on the Steam Hardware Survey (which will be updated with January data shortly, so maybe that will change). Certainly, dropping the price of the A750 by 14% can't hurt.

It's not just about lowering the price, either. Since launching with the 3490 drivers back in October 2022, Intel has delivered three WHQL drivers and at least four beta drivers. The latest beta, version 4090, became available last week, and improving performance and compatibility has been a key target for all of the driver updates.

DirectX 9 performance, an area that Intel hadn't really focused on prior to the Arc launch, has been one of the biggest benefactors of the newer drivers. Intel claims that, across a test suite of thirteen games, average framerates at 1080p have improved by 43%, and 99th percentile fps has improved by 60%. At 1440p, the average fps increased by 35% while 99th percentile fps improved by 52%.

Granted, the test suite for DX9 games isn't so much about making games that ran poorly suddenly run well. The worst performing of the suite, Stellaris, looks to have performed at about 75 fps with the launch drivers, whereas it's now getting more like 130 fps. And Half-Life 2 went from just under 400 fps to about 600 fps. Even so, the overall experience has improved, and framerate consistency and frame times are much more stable.

It's also interesting that Intel continues to show the Arc A750 as an RTX 3060 competitor, mostly ignoring (in charts) AMD's own RX 6600. That's probably because AMD has a much stronger value proposition, with the RX 6600 regularly selling for $225, give or take. Our testing puts it slightly below the A750 (and RTX 3060), but the price cut does make up for the higher power use on Intel's GPU.

Intel is also continuing to push its XeSS (Xe Super Sampling) AI upscaling algorithm as an alternative to Nvidia's DLSS and AMD's FSR technologies. The adoption rate isn't nearly as high, but considering how new Intel is to the dedicated GPU arena, getting 35 games to support XeSS in the first six months or so is pretty decent.

Another feature of the Arc GPUs that's more than just "pretty decent" is the video encoding and decoding support. Arc was the first modern GPU to offer full AV1 support, and the quality of the Quick Sync Video encoding goes head to head with Nvidia's best (with AMD trailing on previous generation GPUs, though we still need to look at the latest RDNA 3 chips).

But it's not all sunshine and flowers. Our own testing of Bright Memory Infinite (using the Bright Memory Infinite benchmark on Steam) and Minecraft shows there's still room for improvement. Another interesting aspect of the Arc GPUs that we've discovered is that, using a Samsung Odyssey Neo G8 32-inch monitor, the DisplayPort connection can only run at up to 4K and 120 Hz, while Nvidia's RTX 20-series and later (using DP1.4a) all support 4K and 240 Hz via Display Stream Compression.

Ultimately, lowering the price of the A750 by $40 probably won't change the minds of millions of gamers, but it does make the overall package more attractive. Intel has also added Nightingale and The Settlers: New Allies to the software bundle for anyone who purchases a new Arc graphics card or system equipped with an Arc GPU. As we've noted before, Intel may not have the fastest cards on the planet, but the value proposition is certainly worth considering.

Which does bring up an interesting question: What's happened with the Arc A580? That's supposed to have the same 8GB of GDDR6 as the A750, but with 24 Xe-cores instead of 28 Xe-cores (3,072 shaders vs. 3,584 shaders on the A750). It also has a lower TBP of 175W compared to 225W and a Game clock of 1700 MHz, or at least that's the theory. With the new price on the A750, the space for an A580 continues to shrink, but maybe Intel could still release something in the $199–$219 range. We're still waiting...

The full slide deck from Intel is included below, for reference.