Paying for your new games or in-game content like new levels and skins might be about to get even easier if a newly published PlayStation patent is anything to go by. The patent, titled “Video Game Controller-Driven Information Transfer,” details a system that would allow a PlayStation controller to accept payments whenever a gamer taps it with their credit card or smartphone. It would work with gift cards, too.

The patent was first filed way back in March 2025, but it was only made public on September 17 this year before being spotted by Dexerto.

The patent includes artwork that shows how the payment system would potentially work. One example shows a TV displaying the question "Do you wish to purchase this game?" with a PS5 DualSense controller beside a smartphone. Another shows a "Gift card received!" message, confirming that we may be about to see the end of entering long strings of text just to top up a PlayStation Store balance.

PlayStation's patent also suggests that gamers could be prompted to buy a game after playing a demo at a retailer or event. The console would handle payment once a card or phone has been presented to a controller, although it's unclear if gamers would need to sign in to their PlayStation account for the purchase to be completed. The patent does allow for both NFC and Bluetooth to be used, so it's possible the PlayStation smartphone app may play a part here.

However, those hoping the patent would hint at a future PlayStation console or controller will be disappointed to learn that is very much not the case. Imagery used in the patent shows a recognizable PS5 console and DualSense controller.

There have been rumors of a PS6 launch as soon as 2027, but given the current state of play with regard to component shortages and price hikes, that's far from confirmed at this point. A delay until 2029 has also been rumored.

Just because Sony has a patent doesn't mean this idea will hit real PlayStations. Companies apply for (and are granted) patents all the time. Not everything depicted in those patents turn into shipping products or features. There are no guarantees this new payment system will ever see the light of day.

Nvidia CEO Jensen Huang has a simple solution for frontier labs like OpenAI and Anthropic that have recently called for increased regulation for their models — "we have to shut the labs down," Huang said in an interview with The New York Times' Ezra Klein. Huang's comments are pragmatic ones, stating that if frontier labs can't contain their experiments, those experiments must end.

"Now, if they say the alternative, which is: There is no way to contain our experiments, there’s just no way; when we test our A.I. models, it will get out, and it will damage the world — then I think the answer is that we have to shut the labs down," Huang said. "Because the cost to humanity, the damage is too great. The shareholder, the liabilities — it could be civil liabilities, it could be criminal liabilities. I mean, the liability’s incredible."

The conversation around increased AI regulation comes on the back of a highly publicized incident where an OpenAI agent hacked popular model hosting platform Hugging Face — a platform that Nvidia now owns, and one that likely would've sought damages from OpenAI had it been under Nvidia's purview at the time. Following that incident, various frontier labs, including those at Anthropic and Alphabet (Google), have used the idea of rogue agents as equal parts marketing tool and call to regulatory action, likely in an attempt to be the first at the table to shape those regulations.

Huang has not been keen on those regulatory calls. Just days ago, the executive countered extreme claims that AI would destroy the world by 2030, and said that frontier labs "never should ship products before they're ready and deliver products that are unsafe."

Although the labs and their chief hardware vendor, Nvidia, have mostly been aligned when it comes to AI policy, the push for regulatory capture has created a rift. Multiple labs, including OpenAI, Anthropic, and SpaceXAI, agreed mutually to slow the pace of model development. Now, they're facing an antitrust lawsuit, with claims these plans are "self-serving."

Huang's comments to Klein shift the responsibility away from collective development and toward individual leaders. "These are companies with agency," Huang said. "If I believe that I'm about to launch a product that is unsafe, it is completely in my ability... to not launch that product. And so I can't buy into the idea that somehow, all of Americans, around 400 million of us, are pushing them to launch untested products that are unreliable."

Although one side of the argument (from the frontier labs) is seemingly pushing for more regulation, that doesn't mean Huang is against regulation. In his own words: "I’m saying that we have lots of laws and regulations. Apply it." The executive argued that labs, particularly those at OpenAI and Anthropic, have rapidly gone from research institutions to companies likely worth hundreds of billions of dollars, and that they should carry themselves as such. "I’m not against laws and regulations. I’m against, currently, the distraction," Huang said.

Reports of rogue AI agents continue to mount. Earlier this month, Anthropic disclosed a fourth instance where one of its models hacked an external system, and just days ago, Google confirmed a report that its Gemini models hacked three companies earlier this year.

Bookstores in Japan are enjoying a boom in sales right now. However, this welcome spurt in business, where used books are bought in the 100s or even by the ton, is causing mixed feelings among owners of these businesses, reports NTV Japan (machine translation). It is suspected that many of the well-read hardbacks and paperbacks, often directed to “a logistics center in Okayama Prefecture,” will be scanned and then pulped by one of the foreign AI tech giants.

We’ve previously reported on AI companies scanning and then destroying millions of books in the U.S. and Europe. Now reports indicate that the AI vandals are running similar schemes in Japan.

According to NTV, there has been a large swell of orders being placed with bookstores that sell via online marketplaces. In interviews, some stores have reported selling 100s of books per day via these channels, or even days when sales have multiplied fivefold.

The Japanese news site notes there are records of a 50-ton consignment of Japanese books being shipped to the U.S. to be churned through Anthropic’s destructive scanning process. It has also heard that a used bookstore in Western Japan has received an email from a foreign company inquiring about an order of tens of thousands of books.

There are several other AI-is-vacuuming-up-my-culture tells in the new NTV story. The Japanese news site says that while the multiple accounts are responsible for the surge in bulk orders, they all request shipment to a logistics center in Okayama Prefecture. Unfortunately, the logistics center operator wouldn’t comment or answer questions from NTV.

The genre purchasing pattern of the bulk buyers also hints at an AI scan and shred operation. A Tokyo bookstore owner told NTV that it used to be novels and comics that sold the most. However, the increased orders have recently been in genres like philosophy, history, political history, medicine, law, and topics such as “life and culture in the Edo period.”

There are mixed feelings among the used bookstore owners who were interviewed. There’s something of a boom to be enjoyed now, with sales rocketing. Clearly, though, if used books are fully removed from circulation, there will be impacts on the prospects of businesses – like stock availability and pricing.

There’s also the question of the removal of important cultural works from circulation if this trend continues. Another moral question concerns the use of copyrighted works for AI learning. This is still legal in Japan, unless it can be demonstrated that the scanning unfairly harms the interests of the copyright holder. With underhand bulk operations like these, it isn’t likely anyone will be asked for permission about their book being scanned.

The Blue Yeti launched back in 2009 and went on to become one of the most popular USB microphones for creators, podcasters, and streamers. Now, eight years after acquiring Blue Microphones, Logitech is finally launching its successor. The new Yeti 2 arrives as a more modern version with a cleaner design and features a 3D proximity sensor, AI-based noise reduction, and even a built-in display. Unlike the original Yeti, however, it will be available in only two colors — Black and White — at a starting price of $159.99.

• Grab the Logitech Yeti 2 microphone from Amazon

Logitech says the proximity sensor on the Yeti 2 is capable of tracking audio in 3D and can automatically adjust various settings in real time, including gain, equalizer, and pickup patterns. As for the AI denoiser, it can actively filter out unwanted background noise such as keyboard clacks, mouse clicks, and fan noise to ensure crystal-clear audio.

The Yeti 2 also retains the USB connectivity that helped make the original microphone popular, while adding several features aimed at making it easier to adjust settings on the fly. The integrated 1.3-inch colored LED display features an easy-to-use interface with an 8-bit Yeti mascot that provides quick information and helpful tips. It also allows you to tweak settings or switch between presets using the dial, without having to rely on dedicated software. Alternatively, the mic offers direct integration with Logitech’s G Hub software for deeper settings control, community-powered Blue VO!CE profiles, and custom avatar uploads.

In addition to its latest microphone, Logitech also announced some gaming-focused products under its G Pro series. The new gaming lineup includes the Pro X3 Superstrike mouse, Pro X Control mousepad, Pro X2 Rapid keyboard, and Pro X3 Lightspeed headset, all designed with input from professional esports athletes.

The Pro X3 Superstrike uses Logitech’s proprietary HITS technology and is claimed to deliver 50% improved power efficiency in a lightweight design. The Pro X Control mousepad measures 400 x 460 mm and features a knitted cloth surface, 6 mm foam cushioning, stitched edges, and a high-friction base for stability. The Pro X2 Rapid keyboard comes with magnetic switches, Rapid Trigger, Key Priority, and an OLED display that allows users to make quick adjustments. Finally, the Pro X3 Lightspeed headset features a 48 kHz microphone, customizable EQ, and extended battery life aimed at long competitive gaming sessions.

With RAMageddon, the storage apocalypse, GPU stagnation, and GPU price increases all happening more or less simultaneously, this is the worst time to build a PC that I can remember — and I built my first PC in the previous century (powered by an AMD K6-2). If you have a rig that meets your needs at the moment (even if just barely), current prices probably make you more interested in keeping it running than in considering major upgrades or, even more daunting, starting over from scratch with a new build.

But fret as we might about the future of our much-beloved component-swapping hobby, a few articles and experiences in recent weeks have given me glimpses of the PC-building future, and I don’t think it’s all bad — at least so long as you have time to research and tinker, you can live without the highest-end components, or you have enough of a build budget that price doesn’t matter much.

Old parts get a new life, with some patient tweaking

First off, we recently tested AMD’s BC-250, a 2021-era APU that started off as defective or cut-down PS5 silicon, repurposed as a cryptomining card. But more recently, a community-developed BIOS and drivers have turned this custom hardware into a Linux-based competitor to Valve’s much-hyped Steam Machine, while costing roughly half the price of a similarly configured system, once you source the power supply and other necessary parts. It even comes with 16GB of shared GDDR6, sidestepping the current RAM pricing issue altogether.

Now, this is more of a project than a PC. You have to source the card (from places like AliExpress, eBay, or Etsy), then research and figure out how to flash the board so that it will even boot, followed by getting Linux and the requisite drivers installed, before you can even think about playing games. But depending on your sample, you might also be able to unlock extra CPU cores and compute units for even better performance.

It’s far from a simple setup. But this kind of community-led hardware repurposing and enhancement reminds me of two distinct hardware eras: The very early days of PC kits (which I wasn’t really around for), and the early 2000s, when you could unlock AMD Athlon and Duron multipliers with a pencil (or later, conductive silver lacquer). Unlocking higher clocks and extra performance without paying for a much pricier chip is exactly the kind of achievable dream that the PC enthusiast hobby was founded on, and it’s good to see it making a comeback – even if it’s with this one very specific piece of hardware.

But as AI hardware development continues — and especially if (or when) specific areas of the hardware bubble pop or deflate – I expect to see more situations where communities of tinkerers figure out how to get discarded hardware to run games and other compute tasks, giving budget buyers hope for acceptable performance at reasonable prices. Ironically, AI tools like Claude Code will almost certainly make breakthroughs in hardware repurposing much easier.

To boldly go (to price brackets) where no PC has gone before

On the other end of the PC price spectrum, I recently assembled a PC using parts from Asus’ ROG Edition 20 line. The anniversary hardware, which debuted earlier this year at Computex, ignores affordability concerns entirely. Asus instead opted to celebrate two decades of its Republic of Gamers sub-brand by taking several of its already expensive parts and peripherals, adding a few (mostly aesthetic) features and some gold accents, and charging even more for them.

How much more? The finished PC, with about $2,000 worth of peripherals, a $3,300 motherboard and cooler combo, and a $6,000 RTX 5090, costs roughly $15,000 – and that’s with a modest 2TB SSD and 32GB of RAM. If you opt for a more fitting 128GB kit with tight timings and 8TB of solid-state storage, you could easily spend $20,000 on a build like this.

As tough as it is for many of us to swallow, there’s likely a large enough customer base of gamers and PC enthusiasts who can and will pay whatever it takes to own high-end, high-performance PCs to keep options like this afloat. Because while the cost of individual parts has certainly increased, the luxury PC market has existed for decades. I know because in the years before I joined Tom’s Hardware, I reviewed several PCs with custom cooling loops and multiple graphics cards (up to quad-SLI) from the likes of Digital Storm and Maingear. That market isn’t going away; it’s just getting more expensive.

The murky future of mainstream desktop gaming PCs

So what about the mainstream PC gaming market? Things are tougher there, and I suspect that space will continue to shrink due to, well, everything I linked to up top. Those looking for gaming options in the sub-$2,000 range that “just work” without much tinkering are more likely to make do with consoles (which are, of course, also getting more expensive), gaming laptops or handhelds, or console-like PCs like Valve’s Steam Machine.

Intel and AMD are reviving or extending the lifespan of chips and platforms that support last-gen DDR4, but older memory isn’t getting any cheaper either. And our own testing has shown that DDR4 is increasingly holding back gaming performance.

There was a brief hope that Chinese memory makers like CXMT might help bring down DDR5 prices. But U.S. approval for that seems unlikely, especially as the FCC is busy banning humanoid robots and robot vacuum cleaners built overseas. And, of course, companies like CXMT aren’t looking to be PC saviors. They want to sell their RAM to AI hyperscalers for the same high prices as SK hynix, Micron, and Samsung. And even if DDR5 somehow suddenly became more affordable, GPU prices are also continuing to climb, and storage (SSDs and hard drives) faces similar pricing issues.

Can software save the day?

Like it or not, we’ll certainly see further advancements in AI upscaling and other VRAM tweaks and tricks that make modern gaming possible on older and current-gen cards with 8GB or even 4GB of VRAM before the price of RAM, storage, and GPUs drop significantly.

Will it be enough to keep PC gaming enticing and affordable enough to prop up the mainstream PC gaming market? That’s tough to say with any certainty at this point, but if being a PC enthusiast since the 1990s (and a PC gamer since the late 1980s) has taught me one thing, it’s that this hobby is resilient. Some will pay what they need to pay for high-end hardware. Some will find suitable solutions by tinkering with repurposed silicon. And others will forge a middle ground that works for them.

I transitioned from a PC enthusiast to a full-time professional covering the PC and related hardware space in 2008. Even back then, people were telling me that the desktop was dead or dying. Instead, it has evolved and survived for nearly 20 years, through the smartphone revolution, past soldered-on laptop components, through two cycles of Steam Machines, and the death of physical game media. Meanwhile, game stores like Steam rake in record profits, indicating that the demand for PC gaming remains high. All this tells me that PC building and PC gaming will need to continue to evolve, but it’s not quite on the endangered hardware list yet.

A Chinese tech firm’s crowd-funded project has touched off a firestorm over U.S. export controls after an analyst spotted a heavily restricted, military-grade AMD chip in its design selling for pennies on the dollar. The discovery — and the analyst’s subsequent call for a treason investigation — puts a sharp spotlight on how easily dual-use silicon can bypass strict U.S. national security sanctions through third-party supply chains ... despite corporate denials of direct sales.

Dylan Patel, the head of top semiconductor analysis company SemiAnalysis, has discovered that China's Puzhi Electronic Technology is crowdfunding development of a system based on AMD's Zync UltraScale+ XCZU47DR RFSoC, which is a U.S. export-controlled dual-use component. He called on the government to investigate AMD for treason.

"AMD found a business opportunity in dumping U.S. military chips for ¼ of the cost in China," Patel wrote in an X post. "U.S. list price is $36K for this chip, with $4K – 5K to U.S. companies at volume, but it is getting quoted $1K in China to crowdfunding campaigns. AMD needs to be investigated for treason. It is not just export violations, but it is literally selling military end use components to an adversary for less than your home country."

In a statement to Tom's Hardware, AMD denied selling or shipping the chips directly into the restricted region. The company says it investigates suspected diversion cases to determine whether any violation of law occurred, although it did not explicitly confirm that it had opened an investigation into this particular case.

"AMD is committed to full compliance with all U.S. and global export regulations and has a clear policy requiring our distributors, resellers, and customers to do the same," a spokesperson told Tom's Hardware. "Any shipment into restricted regions is a direct violation of our policy. This recent instance is unrelated to any direct AMD sales or shipments. We investigate all suspected diversion cases and if we find any violation of law, we will report it to relevant authorities and cease all business with the company."

AMD told Patel the same, according to another X post he made as a follow-up on the matter.

AMD's Zynq UltraScale+ XCZU47DR RFSoCs pack four general-purpose Arm cores, two Cortex-R5F real-time cores, an FPGA with 930,000 gates, and DSP engines with 4,272 DSP slices. It also has eight 14-bit analog-to-digital converters (ADC) capable of up to 5 GSPS and eight 14-bit digital-to-analog converters (DAC) operating at up to 9.85 GSPS to receive, process, and generate RF signals in real time without relying on separate data converters.

Devices based on Zynq UltraScale+ RFSoCs are used for software-defined radios, radar and phased-array systems, wireless and satellite communications, electronic test equipment, and other applications that require high-bandwidth, low-latency signal processing. Puzhi Electric officially positions its upcoming systems based on the Zynq UltraScale+ RFSoCs for 'professional engineers' and 'advanced researchers in universities,' though such systems can be used for completely different applications.

AMD's Zynq UltraScale+ XC-series RFSoCs are designed for commercial and industrial applications, but AMD also offers separate XQ-series defense-grade RFSoCs for aerospace and defense applications, although there the XCZU47DR does not have a direct XQZU47DR counterpart. Yet, the closely related ZU48DR is available in both XC and XQ versions. The latter adds defense-oriented qualification, ruggedization, and support rather than fundamentally different processing capabilities.

Therefore, while the XCZU47DR is technically aimed at civil, commercial, and industrial applications, the chip is a dual-use component featuring a 3A001.a.14 ECCN classification under the U.S. Commerce Control List (CCL) and therefore its exporting, re-exporting, or transferring normally requires a BIS export license to non-allied destinations (e.g., China) on anti-terrorist and national security concerns. This means that any shipment to China requires a license, but in practical reality, that license will almost certainly be denied. However, there is a license exception that allows shipping such items to commercial companies from the Group B countries, which means legitimate exports to dozens of states from where such goods could get to China illegally.

This might explain how Puzhi obtained its XCZU47DR devices, but it does not explain the enormous apparent difference between the prices available to Chinese and American buyers. The $1,000 figure, however, is Patel's reported price for the chip rather than a publicly advertised retail offer.

Xilinx, which AMD later acquired, introduced its 3^rd Generation Zynq UltraScale+ RFSoC family, including the XCZU47DR, in February 2019, and the devices reached production status in 2020, meaning that they have been on the market for almost six years now. Xilinx and then AMD have sold the unit to tens, if not hundreds, of companies in different countries, so determining the origin of the units that Puzhi has might be difficult.

It should be noted that lead times for XCZU47DR at various resellers like Avnet and DigiKey are 40 to 52 weeks, which indicates that the components are in high demand in the U.S. and Europe. Meanwhile, Puzhi claims it can ship its system based on the RFSoC on November 6, 2026.

GMKtec already has an AMD Strix Halo mini PC in its stables, in the form of the Evo-X2. Now, just six months later, we have another AMD Ryzen AI Max+ 395 processor-packing mini PC from GMKtec in the labs. The obvious difference is that the Evo-X3 has a brand new chassis and cooling design, it also comes with 128GB of LPDDR5X-8000 as standard. Those are welcome changes. Half a year later, though, the price has basically doubled as seemingly every electronics product with RAM and NAND chips inside has been affected by the ongoing component shortage.

The result is a desirable computer that will offer strong appeal to enthusiasts who want machines with a large pool of fast, unified RAM, in a competently revamped case. But you will have to have quite a compelling reason to part with the $3,799.99 our 128GB/2TB configuration costs in mid-2026.

Design of the GMKtec Evo-X3 AI Mini PC

GMKtec has thoroughly revamped its Evo-X family with the new X3. Gone is the short-but-stout design, where we saw a plastic inner case sandwiched by some brushed aluminum. Now GMKtec has gone almost entirely metal, with the exception of a textured and vented soft-touch green plastic cap. This towering new design, which is over 14-inches tall with the base attached, features a metal CNC chassis. That body feels very strong, the panels seem thick, and the base is made from a matching material. Where the mini PC meets the desk, there’s a pair of broad padded rubberized feet that you stick on after unpacking and fixing the metal base on the device.

While still in mini PC territory, the Evo-T3 is pushing it a bit at roughly 2.7 liters. Its flat base is about 4-inches wide, which helps it stay very stable on the desk. There are no side feet, which makes running it on its side inadvisible. Users will also have to hide away the large power brick if they are after the cleanest minimal workspace.

GMKtec Evo-X3 specifications

Ports and Upgradeability

GMKtec keeps the front of its mini PC simple with just three ports and the power button. In the top third of the tower’s facade, there’s a central green power button with a white LED ring lit when it is powered up. The middle third of the front is left plain, but the lower third has one USB 3.2 Gen 1, one USB 2.0, and a 3.5mm audio combo jack. I wish there was a Type-C port on the front, preferably USB4 or better.

Around the back it has a good variety of ports, but I wouldn’t describe them as numerous. You will find a USB‑A 3.2 Gen 2 port, a USB4 Type‑C (PD 3.0 / DP-alt) port, an HDMI 2.1 port, an OCuLink PCIe Gen 4 ×4 port, an RJ‑45 2.5G LAN port, and a DC‑IN barrel jack. Ideally, I’d like to have a full size DP port, as well as a second USB4 port, plus more USB-A ports. If I were to use this as a daily driver I think I’d need to connect a fast USB docking solution for the oodles of peripherals I like to have ready and connected.

Wireless connectivity seems adequate enough, with an M.2 2230 card inside providing Wi-Fi 7 and Bluetooth 5.4. I had no problems with either during the test period.

Access to the components inside the GMKtec Evo-X3 is straightforward, but takes time as there are eight Torx-6 screws securing the side panel in place. Once they were all unscrewed I needed to use a guitar pick and a bit of gravity to tilt the heavy panel off. Removing of the panel allowed access to the two M.2 PCIe 4.0 x4 slots, one of which was already populated with a Crucial P310 2TB M.2 NVMe SSD fitted with a slim heatsink. In the opposite corner of the chassis I saw an M.2 2230 socketed MediaTek wireless card. There are no other internal user upgrade or repair options as the 128GB of LPDDR5X-8000 RAM is soldered to the motherboard.

Test System Setup

Productivity Performance on the GMKtec Evo-X3

GMKtec describes the Evo-X3 as a machine that targets AI workloads, but first we ran through our standard mini PC benchmark comparison suite, which targets more general computing usage.

A brief glimpse at the tech specs of the X3 reveals that it has lots of modern fast CPU cores, supported by one of the best iGPUs, and plenty of RAM and storage. It should make light work of our productivity benchmarks, and indeed its 16C/32 core AMD Ryzen AI Max+ 395 CPU left its mini PC rivals eating dust, most strikingly in multi-threaded tests.

As we expected, the single core tests show the GMKtec Evo-X3 is a competent performer in the context of the competition in our charts. It isn’t the only Zen 5 machine in this race, so it doesn’t get a chance to dazzle yet.

Moving onto the multi-core tests is where the Ryzen AI Max+ 395’s 32 threads show their potential for real-world data crunching performance. For example, in Y-cruncher it completed its Pi compute in around 60 seconds. That’s twice as fast as its nearest competitor in our mini PC charts, the Beelink SER10 Max.

Rendering in Cinebench is also strong. In Cinebench 2026, the X3 scored roughly 7,300 points in multi-core. That’s over 40% faster than its nearest competitor, the Arctic Senza AI 370. Video transcoding performance is also in a new class, with the Evo-X3 crushing the competition, achieving around 24 frames per second.

For some added AI benchmarking stats, we did a few tests in Geekbench AI and LM Studio Bionic to weigh the GMKtec Evo-X3 against the Evo-T2. These are both machines with large unified memory pools by default and which the makers claim to be local LLM champs.

Using Geekbench AI with the ONNX framework and DirectML backend the Evo-X3 achieved a half precision score of ~37,100. On the Evo-T2 using the same options, a score of ~23,060 was achieved.

However, machines like the GMKtec Evo-X3 and T2 might be picked up by consumers to work with local LLMs considerably larger than most consumer GPUs can load. With this in mind, I ran LM Studio Bionic chat sessions powered by the new Qwen3.8 27B Q4_K_M model on both machines. To use this 27 billion parameter model as intended, a machine needs around 20GB of ‘VRAM’ of shared memory to avoid undesirable side effects like out-of-memory errors and slowdowns.

I prompted the same tasks on both machines then analyzed the logs and found prompt processing was pretty similar between the platforms, ranging either side of 200 tok/s. However, the AMD Strix Halo Evo-X3 far outpaced the Panther Lake Evo-T2 by generating responses at up to 28 tok/s, compared with around 8 tok/s at best, respectively. It would be possible to tune these LLM speeds for each machine, but I didn’t tweak settings for the comparison tests. With 128GB of unified memory in the Evo-X3, allowing up to 96GB to be dedicated to the GPU, even larger models can be utilized for the most capable local AI chatbots, agents, generation workloads.

During our demanding tests, the GMKtec Evo-X3 stayed admirably cool and quiet. All our tests were done with the default 85W balanced profile, as the system shipped. Owners can also choose to run this system in Silent Mode (54W) or push it to Performance Mode (peak 140W).

Gaming and Graphics on the GMKtec Evo-X3

The GMKtec Evo-X3 is pretty good at gaming, thanks to its powerful CPU and iGPU with lots of fast memory onboard. Remember, the Ryzen AI Max 395+ has 16 Zen 5 CPU cores with 32 threads, and its Radeon RX 8060S iGPU features 40 CUs and 2,560 shader cores – more CUs than a vanilla PlayStation 5. The capacious LPDDR5X-8000 memory pool has a 256-bit interface.

Before moving onto our standard synthetic and AAA PC gaming tests, I decided to run around in Star Wars Outlaws. I chose the High preset, at 1440p, and kind of enjoyed this game, experiencing a performance level of mostly between 60 and 90 frames per second. The graphics weren't anything special, but I noticed the game was using about 9GB of VRAM which could have meant dips and stutters on systems packing 8GB GPUs and using the same graphics settings. While the CPU usage was mostly under 20%, the iGPU was nearer to 90%.

Next up, our benchmarks include a trio of 3DMark synthetics and two iconic PC gaming standards. In the 3DMark Solar Bay test, which seems most appropriate to the mini PCs we have tested in 2026, the Evo-X3 left the competitors in the dust with over 50,000 points. It was almost twice as fast as the Panther Lake B390 iGPU, and about three times as fast as the Radeon 890M machines that came with fast dual-channel RAM (like the aforementioned Arctic and Beelink mini PCs).

In 3DMark Steel Nomad we got a respectable score of 2,132, and the frame rate reached 21 FPS. It was a similar story in 3DMark Speed Way, where the Evo-X3 managed a score of 2,046. For some context this Speed Way result still lags behind my Asus Zephyrus G16 laptop with RTX 4060 Laptop GPU, which scored up to 2,425.

Moving on to real AAA games, let’s check the data provided by the multi-scene benchmarks built into Shadow of the Tomb Raider and Cyberpunk 2077. The former scales well with processor resources and can still often look and run great on humble iGPUs. As for Cyberpunk 2077, it can really push even the strongest modern GPUs, so here we've used two settings to show how the GMKtec Evo-X3 can perform. I thought both these titles looked better than Star Wars Outlaws, too.

The Shadow of the Tomb Raider High 1080p test resulted in an impeccably smooth 138 FPS on average, with the low of 126 FPS, on the Evo-X3. That’s a great score for a mini PC without a discrete graphics card. I was inspired to re-test at 1440p High, and the X3 still achieved an average 96 FPS / low 86 FPS.

In Cyberpunk 2077 the Evo-X3 managed to achieve 96 FPS using the 1080p High preset, with 1% lows at 78 FPS. This is certainly another candidate for moving up to 1440p and/or higher quality presets. CD Projekt Red also provides lots of upscaling tech (DLSS, FSR, XeSS) settings to adjust in this futuristic RPG, to try and get your sweet spot performance / resolution / quality setting dialed in.

Power, Heat, and Noise when using the GMKtec Evo-X3

When I left the GMKtec Evo-X3 to idle, my power meter measured wattages around 16W from the wall for the whole system. When the PC was pushed with simultaneous CPU and GPU benchmarks running, power draw seemed to peak at a little under 150W. GMKtec shipped this device with ‘Balanced Mode’ set in the BIOS, allowing the processor to consume up to 85W. The supplied 230W power brick had plenty in reserve, and it might need it in ‘Performance Mode’ where GMKtec says that the processor alone will demand up to 120W with peak power spikes to 140W.

System tools showed that during light activities like browsing the web, and typing reports like this one, the CPU temperature stuck around 42 degrees Celsius (in a 22 C room). Playing games for 30 or 40 minutes, the reported CPU temperature seemed to stay between 75 C and 82 C most of the time. The peak CPU temperature I saw was 85 C.

GMKtec’s powerful mini PC was a quiet operator at idle and under load. My measurements from about half a foot in front of the device show noise levels under 30 dBA at idle, which is very quiet and under the measurement threshold of my meter. Stressing the system with multiple CPU and GPU loads I managed to get nearly 37 dBA out of the system. This seemed an exceptional result during my testing, as most of the time the max noise level during benchmarks was nearer 33 dBA.

Software and Warranty on the GMKtec Evo-X3

The GMKtec Evo-X3 ships with Windows 11 Pro installed. Unlike its Evo-T2 brother, I found no pre-installed AI apps and tools on the X3. However, I note that the product page mentions GMKtec Claw and Wrangler as features. I installed LM Studio Bionic and Comfy UI for some AI testing, though.

GMKtec sells the Evo-X3 with a one-year warranty.

GMKtec Evo-X3 configurations

GMKtec sells only two Evo-X3 configurations in the U.S. Both machines feature the same processor and generous amounts of RAM. The only configuration choice is between a 2TB SSD (our model) and a 4TB SSD equipped model.

At the time of writing our 2TB configuration is listed at $3,799 but the product page says the price is normally $4,100. The 4TB model is currently $3,999, discounted from $4,349. We don’t know when the prices might get adjusted. Pricing for computers and devices with lots of RAM and storage is volatile right now.

Bottom Line

I was impressed by many aspects of the GMKtec Evo-X3 mini PC. Computing on a machine with so much RAM felt quite liberating. Especially when so much of it can be allocated to the system's iGPU. Moreover, the CPU and iGPU components of the SoC are plenty powerful for the compact form factor. I was also impressed by the X3's sleek design, solid build quality, and quiet operation.

No machine is perfect, though. For example, despite having a variety of ports available including USB4 and OCuLink, I felt there were not enough ports in total. To use this as my daily driver I’d feel pinched by the Evo-X3’s design offering just a single USB4 port, and one HDMI out.

The price is high, even considering the spec, but that's unfortunately common during this component shortage.

GMKtec sells its new Evo-X3 direct at the offer price of $3,799, as mentioned and linked above. At the time of writing we note that our GMKtec Evo-X3 128GB/2TB configuration is also available from Amazon U.S. at the same price, if you prefer.

If you like what you’ve seen of the Evo-X3, the nearest competitors that we have experience of are the Evo-X2, also from GMKtec, and AMD’s first-party, turn-key Strix Halo mini-PC. However, we know there are several next-gen Ryzen AI Max 495+ machines coming to retail soon. Depending on market pricing, configurations, design, and port selections of these refreshed ‘Gorgon Halo’ models, they might be worth waiting for. We saw Minisforum, Lenovo, GMKtec, Chuwi, and AceMagic with AI Max 495+ designs at IFA 2026 earlier this month.