AMD's Ryzen AI Halo Developer Platform is finally here, with preorders for the powerful $3,999 mini PC now live in the U.S. through Micro Center. The company’s new developer platform is effectively an answer to Nvidia's DGX Spark, as both systems are designed as compact AI workstations capable of running large language models locally. First unveiled at CES 2026, the developer kit is available in two variants that feature identical hardware but ship with different operating systems- Linux or Windows 11 Pro.

Notably, the DGX Spark was originally introduced at the same $3,999 price point last year. However, Nvidia recently increased the price to $4,699 due to the ongoing global supply constraints affecting LPDDR5X memory and NAND flash. Apart from pricing, the Ryzen AI Halo Developer Platform offers native Windows support, whereas the DGX Spark is limited to Linux-based environments. This gives users greater flexibility, including the option to dual-boot and use both operating systems.

The Ryzen AI Halo dev kit is powered by the flagship Ryzen AI Max+ 395 processor, featuring 16 cores and 32 threads, with a base clock speed of 3 GHz and a max boost clock speed of 5.1 GHz. It also features 128GB of unified LPDDR5X-8000 memory, 16MB of L2 cache, and 64MB of L3 cache, onboard RDNA 3.5 Radeon 8060S graphics with 40 compute units, and a dedicated NPU capable of 50 TOPS.

The mini-PC measures 149 x 149 x 43.18 mm, has an aluminum chassis, and includes a 2TB M.2 SSD, while wireless connectivity options include Wi-Fi 7 and Bluetooth 5.4. As for I/O, it is strikingly similar to what Nvidia offers, including a 10 GbE LAN, four USB-C ports (one for power input), and an HDMI 2.1b port. Unfortunately, there is no NVLink, which, in Nvidia’s case, allows two DGK Spark machines to pair. To cool the Strix Halo chip inside AMD, it uses a unique cooling solution that includes a baseplate, direct-touch flat heatpipes, an aluminum channel heatsink, and two lateral airflow blower-style fans.

If you are a developer or simply interested in experimenting with local LLMs, Micro Center is currently accepting preorders for the Ryzen AI Halo Developer Platform, with local pickup available on July 10. Those looking for alternatives can also consider the Corsair AI Workstation 300, a slightly larger AI-focused system based on the same Ryzen AI Max+ 395 processor. It is also currently listed at a lower starting price of $2,699 for the 1TB storage model and $3,399 for the 4TB model.

Greg Kroah-Hartman, the Linux kernel's stable branch maintainer, who is widely regarded as second only to Linus Torvalds in the project's hierarchy, posted a photo to Mastodon on Friday showing the hardware behind his AI-assisted bug-finding tool, dubbed a "clanker."

The setup, which Kroah-Hartman has dubbed "gkh_clanker_t1000," is a Framework Desktop powered by AMD's Ryzen AI Max+ "Strix Halo" processor, running a local large language model to hunt down kernel bugs without relying on any cloud infrastructure, as first reported by Phoronix.

Since April 7, close to two dozen patches assisted by the Clanker T1000 have been merged into the mainline Linux kernel, addressing bugs across a range of subsystems, including ALSA, HID, SMB, Nouveau, and IO_uring. Kroah-Hartman first began testing the tool against the kernel's ksmbd and SMB code earlier this month, choosing that subsystem because it was straightforward to set up and test locally using virtual machines.

The patches carry a Git tag reading "Assisted-by: gregkh_clanker_t1000," and Kroah-Hartman has been up-front about the need for human verification, writing in the patch submission that the patches "pass my very limited testing here," adding "please don't trust them at all and verify that I'm not just making this all up before accepting them."

The tool doesn’t write kernel code but instead acts as a fuzzer, bombarding code with unexpected inputs to expose crashes, memory errors, and other latent bugs. Kroah-Hartman then reviews what it finds, writes fixes, and takes full responsibility for the submitted patches.

The Framework Desktop is a 4.5-liter Mini-ITX system built around AMD's Ryzen AI Max+ 395, which pairs 16 Zen 5 CPU cores with 40 RDNA 3.5 compute units and up to 128 GB of unified LPDDR5x memory accessible to both the CPU and GPU. That large shared memory pool makes it capable of running sizable language models locally, a task that would typically require either a high-end discrete GPU with substantial VRAM or a cloud API.

Kroah-Hartman has not disclosed any details about the software stack powering the Clanker T1000, and the emergence of the tool follows the Linux project’s formal adoption of an AI code policy earlier this month, which permits AI-assisted contributions provided developers use an "Assisted-by" disclosure tag and accept full personal liability for any code they submit.

Kroah-Hartman's workflow with the Clanker T1000 predates that policy but already conforms to it.

Minisforum has told Tom’s Hardware that its powerful AI NAS, the N5 Max, will launch on April 23, 2026. The firm describes this as a breakthrough device, and indeed its specs are pretty extreme for a consumer NAS, with an AMD Ryzen AI Max+ 395 delivering 126 TOPS of AI compute power under the hood. It boasts 64GB LPDDR5X RAM, support for up to 200TB of storage, and comes with OpenClaw pre-installed on its 128GB system drive. We’ve also been given the full specs sheet (see below) and told the price for this tasty tech concoction. It will retail at $2,899.

The Minisforum N5 Max isn’t actually retail listed yet, but you should expect it to appear on the firm’s Amazon store on or around April 23, at that $2,899 sticker price.

Previously, we have reported on the overall product concept and key specs. It was already clear that this was a true Max member of the Minisforum N5 range, which also includes the N5 Pro with Ryzen AI 9 HX Pro 370 for $1,750, or $899 for barebones, and the N5 Air, based on an AMD Ryzen 7 255, which is $519 for barebones.

Minisforum expects users to wield this AMD Strix Halo power for “self-owned computing, privately controlled AI.” To help users get started box-fresh, it includes pre-integrated one-click deployment tools to enable full OpenClaw capabilities on the local platform. More specifically, the mini PC stalwart suggests using the N5 Max's power for “AI-powered semantic photo search… AI smart editing… [and] AI Agent OpenClaw” to work smartly on your behalf.

Enough of that AI waffle, though, let’s look at the cold, hard specs. The table below should let you know exactly what you are getting for your nearly $3,000.

Minisforum N5 Max specifications

Some of the key specs we have now that weren’t disclosed in the prior PR or official blogs include the supplied RAM and storage specs/quotas. In the era we live in now, these are sensitive stats.

You can see that the N5 Max, as priced, is configured with 64GB of LPDDR5X-8533 memory, which is non-expandable. On the plus side, that's pretty fast system RAM, which the powerful iGPU will appreciate.

Though you can expand this device to 200TB of lovely storage with 5x HDDs and 5x M.2 SSDs, you don’t get much to start with. Minisforum is shipping this device with a 128GB system disk, according to the spec sheet. That uses one of the five M.2 slots.

There is probably sufficient I/O for many users, including a pair of 10 Gigabit Ethernet ports (but no wireless/BT). We also note that there are 3x USB4 ports, two of which are USB4 v2 with up to 80 Gbps.

We’ve seen Strix Halo systems without the NAS functionality around the $3,000 level, so this price is tolerable given that context. Has Minisforum chosen specs well with its fixed 64GB memory and minimal starter storage, given its product marketing? Buyers will vote with their wallets.

Corsair launched its AI Workstation 300 in the summer of 2025 with AMD's Strix Halo silicon on board to appeal to local LLM enthusiasts looking for a powerful yet compact mini-PC. At the time of our review in February, prices for the top-end Ryzen AI Max+ 395 variant had already risen to around $3,000 from a launch price of around $2,299, and we already criticized its subpar value proposition compared to rivals like the DGX Spark and its GB10 brethren.

Now, all configs of the AI Workstation 300 have gotten even more expensive. The top-of-the-line variant equipped with a Ryzen AI Max 395+ APU, 128 GB of LPDDR5X-8000 RAM, and 4 TB of storage has been bumped up to $3,399.99. Keep in mind that this is the only AI Max 395+ config available on the site, too, since the 1 TB SSD option is out of stock, and you can't down-spec the memory.

If the 1TB model was available, however, it would cost $2,699 now, up $700 from its $1,999 launch price. Finally, the Ryzen AI Max 385 config with 64 GB of RAM and a 1TB SSD has gone from $1,599.99 to $1,699.99, constituting just a $100 hike.

Cumulatively, those make for $1,100, a $700, and $100 price hikes across the board, but in the grand scheme of Strix Halo systems, these prices are pretty much the going rate.

For instance, another popular Ryzen AI Max+ 395 model, the Beelink GTR9 Pro, is going for $2,999 right now, and it comes with only a single 2 TB SSD. Adding another high-end 2TB PCIe 4.0 drive would cost you at least $350 right now, by which point you'd already be in the Corsair model's territory.

We reviewed the ever-popular GMKtec Evo X2 a couple of months ago as well, which is currently priced at $2,349 on Amazon for the same chip and a 2 TB SSD, but only 96 GB of RAM instead of 128 GB. If you want that much memory, you'll have to shell out $3,299, but you only get 2 TB of storage instead of 4 TB on the Corsair. Heck, even the Framework Desktop would be slightly more expensive when configured the same way.

Regardless of the context, Corsair's price changes came without an announcement, so we don't actually have an official reason for the price increase, but it's pretty easy to connect the dots. The ongoing component shortage caused by the AI boom is a likely root cause, and it's been made worse by escalating geopolitical tensions blocking off critical trade routes.

RAM and storage are extremely costly at the moment, even for vendors buying in bulk, so that added cost is going to get passed down to the customer. Recently, Ayaneo even had to halt orders for its upcoming Next 2 handheld featuring Strix Halo chips because the company simply could not get the parts at a reasonable price. But hey, if you're chasing tokens amid AI boom that started all this, maybe it doesn't matter how much pricier your Strix Halo machine gets.

Last month, Ayaneo unveiled the Next 2 — a new premium handheld console featuring AMD's Strix Halo chips and all the specs you could ever ask for. Windows-based handhelds are already niche enough, but launching a high-end one during a global component crisis was going to pose even more of a challenge. Sure enough, the company has just announced it's "no longer sustainable" to sell the Next 2, and pre-orders for the device are being suspended temporarily.

Ayaneo posted the update on its Indiegogo page, saying that it was already prepared to absorb a slight loss on the Next 2 due to ongoing shortages, but the situation has now slipped out of control. At the time of the initial announcement, storage prices were already high, but following the Chinese New Year, quotes from memory vendors had shot up many times, making it impossible to produce the device feasibly.

As a reminder, the Next 2 started at $1,999, but you could get it for $1,799 as part of early bird pricing, and that's with an AMD Ryzen AI Max 385 chip, 32 GB of LPDDR5X RAM, and 1 TB of storage. The higher-spec'd configs were much more expensive; the unit with the flagship Ryzen AI Max 395+ and 128 of RAM, along with a 2 TB SSD, was priced at $4,299 ($3,499 for early bird preorders).

Those price tags already look eye-watering, even with preorder discounts, so the fact that Ayaneo can't hit them anymore should signal just how bad things have gotten. The company claims the cost is approaching "twice the price" that was originally set, throwing away any profit and incurring a massive loss that a small OEM like Ayaneo can't absorb.

Still, Ayaneo is going to honor all the existing preorders, so if you already bought one via the Indiegogo campaign, you will receive the handheld. Going forward, though, all sales channels for the Next 2 are blocked, and no new preorders are being accepted, but the suspension is temporary. The company says it doesn't want to submit to price fluctuations, and the handheld should return to sale if the component crisis eases up.

Moreover, the after-sales customer support for the Next 2 is not affected, as promised by Ayaneo, with sufficient spare parts for any repairs or RMA. The Next 2 product line is not ending either; this is just a provisional pit stop, and the handheld is too spec'd out to just be left on the table. It features a 9.06-inch 2K 165 Hz OLED display and a 116 Wh battery, providing performance comparable to a laptop.

Chinese hardware manufacturer Ayaneo, known for its handheld gaming consoles and mini PCs, has unveiled details of its upcoming flagship device. The Next 2 is a premium Windows handheld gaming console that features the largest built-in battery in its category, along with AMD’s powerful Strix Halo chips and a 9.06-inch OLED display. As with its previous projects, Ayaneo has listed the Next 2 for pre-order on Indiegogo, with prices starting at $1,799 for the 32GB RAM and 1TB SSD model and climbing to $4,299 for the fully specced configuration featuring 128GB of RAM, a 2TB SSD, and a top-of-the-line AMD APU.

The Next 2 will be available in white and black color options and feature a design that is larger than most current-generation handheld consoles, including the Asus ROG Ally X and Valve’s Steam Deck. It is also considerably heavier due to its beefy internals, weighing in at 3.14 lbs, more than double the weight of the Steam Deck.

Ayaneo claims it is the first Windows handheld to offer adjustable-torque TMR Hall-effect joysticks and also features Hall-effect linear triggers with a dual-lock switch on the rear. Additional controls include an 8-way floating D-pad, four customizable rear buttons, dual smart touchpads on either side of the display, and front-facing stereo speakers. I/O and connectivity options include two USB4 Type-C ports with charging and DisplayPort support, Wi-Fi, Bluetooth 5.4, a 3.5mm headphone jack, and a microSD card expansion slot.

The large custom OLED display features a 2400x1504 resolution and supports an adjustable refresh rate of up to 165 Hz, with selectable modes at 60 Hz, 120 Hz, 144 Hz, and 165 Hz. However, there is no clarity whether the panel supports variable refresh rate. The display also covers 100% of the DCI-P3 color gamut and delivers an impressive peak brightness of up to 1,155 nits.

Powering the Next 2 is AMD’s Ryzen AI Max+ 395, a performance-focused SoC featuring sixteen Zen 5 CPU cores and the Radeon 8060S iGPU, which has the potential to deliver gaming performance comparable to a laptop RTX 4060. The chip can be paired with up to 128GB of unified LPDDR5X memory running at 8,000 MT/s, shared between the CPU and iGPU, allowing users to allocate up to 96GB specifically for VRAM. Simply put, this chip should demolish nearly every handheld gaming console currently on the market. Ayaneo is also offering the Next 2 with the Ryzen AI Max 385, a more modest chip with 8-cores, integrated Radeon 8050S graphics, 32GB of memory, and a 1TB SSD.

To keep all that power in check, the handheld is equipped with a dual-fan cooling system paired with a massive heatsink to tame AMD’s Strix Halo chip under sustained loads. Powering the entire package is an enormous 116 Whr battery, an almost unheard-of capacity even among laptops that are limited to 99 WHr. For context, airline regulations typically allow lithium batteries up to 100 Whr without approval, while anything larger requires special clearance.

The only other handheld powered by a Strix Halo chip is the GPD Win 5, the first Ryzen AI Max+ 395–based handheld to actually make it to market. However, it relies on a detachable external battery rather than a built-in one, which can be a compromise for some users.

A few days ago, AMD added two new SKUs to its Strix Halo lineup at CES 2026. One of those was the Ryzen AI Max+ 392, serving as a cut-down version of the top-end AI Max+ 395 but with fewer CPU cores — 12 instead of 16. While the company teased the chips ahead of CES, the first real-world Geekbench listing for the AI Max+ 392 has just surfaced.

The new Strix Halo APU scored 2,917 points in the single-core test and 18,071 points in the multi-core test, which almost matches the 18,348 points achieved by the desktop Ryzen 7 9800X3D. This multi-threaded score is actually more than the 17,624 points listed for the flagship Ryzen AI Max+ 395 on Geekbench's database, but the sample size for that chip is much larger.

Apart from AI, these Ryzen AI Max+ chips are focused on gaming, which is more of a single-threaded task, so it's once again impressive to see the 392 land among the top performers. Once again, the Ryzen AI Max+ 395 itself has a recorded score of 2,781 points, so this result in particular is better. As for desktop counterparts, the Ryzen 9 7900X and 7950X3D are both within the margin of error as well.

The Geekbench listing also tells us that the score was extracted from an Asus TUF Gaming A14, a device we covered extensively at CES and even awarded "best in show" in the gaming laptops category. This unit, in particular, was rocking 64 GB of on-chip memory running at 8000 MT/s, and boosted up to 5.02 GHz, in line with the listed specs for the chip.

Speaking of, the Ryzen AI Max+ 392 is a 12-core/24-thread CPU with 40 RDNA 3.5 Compute Units (same as the 395). It has 76 MB of L3 cache, a boost clock of 5.0 GHz, along with a rated TDP of 45-120W — Asus caps it at 85W for this machine. Even though the general availability for new Strix Halo machines is Q1 2026, we'll have to wait and see just how much shelf space is populated with these. AMD's other, more mainstream Ryzen AI 400 series was recently listed for a January 22 release in China.

Intel didn't talk about Panther Lake performance in detail when it took the wraps off its first 18A processors a bit ago, but a possible flagship SKU from the lineup — the Core Ultra X9 388H — has just appeared on Geekbench, and the scores are quite impressive. They not only one-up Intel's outgoing chips, but match AMD's top-end Strix Halo offerings as well.

The Core Ultra X9 388H got 3,057 points in the single-core test and 17,687 points in the multi-core test, both figures that put it way ahead of the Arrow Lake-H Core Ultra 9 285H, but also on par with the beefier Core Ultra 9 275HX.

That all is impressive enough, but compared to Ryzen AI Max+ 395, AMD's best mobile chip right now, the X9 388H's multi-core score is within the margin of error, but the single-core numbers are 8.7% ahead. The Geekbench listing also shows the CPU boosting to 5.1 GHz. Check out the table at the end of the article for a detailed breakdown.

These are exceptional values for a CPU that fits into an even tighter thermal envelope than Strix Halo. The X9 388H has a 45W default TDP versus the Ryzen AI Max+ 395's 55W default.

For some more context, the X9 388H is the highest-spec'd model from Panther Lake we've seen so far, featuring 16 cores in a 4P + 8E + 4 LP-E config, with no Hyper-Threading. The SKU also has the "X" designation because it features 12 Xe3 iGPU cores, but that's beyond the scope of this leak, though we've covered Panther Lake graphics before.

Panther Lake isn't supposed to come out until early next year, but excitement for the release is already dimmed. As it stands right now, surging memory prices may result in insane markups on laptops and other mobile devices featuring Panther Lake parts, no matter how impressive the SoC itself may or may not be.

Also, keep in mind that each Geekbench listing is different, so we have to wait for more benchmarks to come out before an average can be formed, but the first impressions with this leak sure do look great.

AMD's latest Ryzen AI Max APUs might be designed primarily to crunch through heavy AI workloads, but they are also quite capable gaming beasts. Integrated graphics enthusiast RandomGaminginHD on YouTube benchmarked a Minisforum MS-S1 Max mini-pc boasting AMD's flagship Ryzen AI Max 395+ APU, and was astonished at the APU's performance in games. The chip's integrated Radeon 8060S iGPU was able to provide butter-smooth 1080p performance in all the games he tested.

The games the YouTuber tested featured Battlefield 6, Borderlands 4, Counter Strike 2, Cyberpunk 2077, Elden Ring, GTA 5 Enhanced Edition, Kingdom Come Deliverance 2, Marvel Rivals, Red Dead Redemption 2, and Outer Worlds 2. Graphics settings were varied based on each title, so not all games were tested at their highest quality settings. However, all games were tested at 1080p.

In Battlefield 6, at the high preset (with HD textures not installed), with FSR set to its "Native AA" mode, the Ryzen AI Max 395+ was able to pump out 86 FPS average with 0.1% lows just under 60 FPS. Borderlands 4 ran at 54.8 FPS average at the lowest quality settings with "Native AA" mode set for FSR. That might not sound like great performance, but it's worth mentioning that Borderlands 4 gets poor FPS at low settings even on discrete GPU hardware.

Counter-Strike 2 ran the best at over 250 FPS average at high settings with MSAA off. The only trouble the mini-PC had was in the 0.1 lows, which dipped all the way down to 66 FPS. But this game is renowned for its performance problems, so the low 0.1% lows are likely related to that rather than being hardware-specific.

Cyberpunk 2077 was the most demanding game the YouTuber tested (including graphics settings tested). At the high preset with high crowds enabled and high textures enabled, the integrated 8060S achieved 67FPS average with 36.4 FPS for the 0.1% low FPS. Switching things up a few notches with the RT Ultra preset turned on and FSR 3 quality mode activated, the game ran at 45.4 FPS average and 27.2 FPS for the 0.1% low frame rate.

The rest of RandomGaminginHD's testing suite ran well: Elden Ring achieved 60 FPS – with the game's 60FPS cap, mind you, at the game's high preset. GTA 5 Enhanced Edition ran at 77 FPS average at the game's very high RT preset, Kingdom Come Deliverance 2 ran at 90FPS average with the high preset and SMAA 2TX enabled. Marvel Rivals ran at 79 FPS average with very targeted settings, featuring the high preset but with global illumination and reflection quality turned off. FSR 3 was also set to "Native AA" mode in this title.

Red Dead Redemption 2 ran at 82 FPS average using Ultra Textures, everything else set to high, TAA on high, Geo LOD set to max, and the grass LOD set to 2 out of 10. Outer Worlds 2 ran at 65 FPS average at medium settings, and TSR set to native resolution.

The Ryzen AI Max series was not made specifically for gaming; it was a platform designed for power users and professionals who need a super-fast computing chip in an ultra-compact mobile form factor. It was also designed to run large language models, especially the 128GB LPDDR5x models, and allegedly has the performance chops to outperform an RTX 4090 by over 2x in tokens per second.

But, all of that high-performance hardware, by nature, turns the APU into a very good gaming processor (albeit an overkill gaming platform in some regards). We've already seen impressive benchmark numbers from the 395+'s integrated 8060S iGPU, with performance equivalent to an RTX 4060 in other devices. The 8060S' performance is a testament to its impressive specs, featuring 40 CUs and 2,560 shader cores – more CUs than a vanilla PlayStation 5, and 16 Zen 5 CPU cores paired to a huge (for a mobile APU) 256-bit interface with fast LPDDR5X memory. RandomGaminginHD's benchmarks further solidify the 8060S's impressive performance as an integrated graphics solution.

It is no wonder handheld gaming manufacturers are implementing this APU as a flagship gaming option, including GPD and its Win 5.

A small form factor PC aficionado with a penchant for passive, fanless systems has been documenting their process online (h/t Fanless Tech). So far, so ordinary, but the ‘Monochrome 2’ from TheJiral features the APU on everyone’s wish list, the AMD Ryzen AI Max+ 395 ‘Strix Halo’ processor. The project was flagged as completed a few days ago.

Our fan-averse hero started this black & white & passive build back at the end of June, before receiving the computing hardware. Nevertheless, as the forum postings show, TheJiral still managed to do a lot of preparation work ahead of receiving their Framework Desktop machine. They pondered over various supporting components, cabling, and so on. Monochrome 2 uses Framework’s relatively tinker-friendly device as its foundation, and you can follow the above link to check out our extensive review from earlier this month.

Chassis as the heatsink

Of course, a significant thrust of the Monochrome 2’s design is directed by the desire to use the chassis as a heatsink. We've seen this idea applied plenty of times before. Such a design lends itself to a compact build like this SFF, where case panels can be closely connected to computing cores for a dense build, bristling with fins for maximum convection opportunity.

One of the first non-custom components that was bought to complete this build was the HDPLEX 250W passive GaN AIO ATX PSU. Despite some concerns, this choice seemed to be validated when the GMKtec EVO-X2 mini PC (with the same processor) was confirmed to come packing a 230W PSU.

The Monochrome 2 seems to have first been booted with its Framework motherboard in situ a few days ago. TheJiral decided to make this a Linux (openSUSE) system. Thankfully, after some thermal interface concerns and tweaks, the system was announced “build complete” on Friday. Now for some benchmarks – with an eye on thermals, of course.

Some tests, but comparisons are hard to make

There’s a little bit of crossover in these tests with our Framework Desktop benchmarks, but sadly not that much. For example, our fan-cooled system scored 2,966 / 17,574 in single / multi-core Geekbench 6 tests. Scores for the passive Monochrome 2 were recorded as 3,116 / 20,785. The passive system does noticeably better, but there are OS and perhaps Geekbench 6.X version differences to spoil a direct comparison.

We both also did some Cyberpunk 2077 game performance testing. Our Framework machine achieved 22fps in the 1080p Ray Tracing Ultra preset. Meanwhile, the passive system is said to have achieved an average of 31fps at what looks to be the same preset. We have an inkling that some level of FSR was used by TheJiral, though.

While the benchmark comparisons we can see don’t make the passive machine’s relative performance crystal clear, we can perhaps conclude that the passive thermal solution is doing a decent job. Supporting this notion, the maker of Monochrome 2 shares thermal data and images from the machine with the APU set at 100W / balanced mode. They would later update with the system pushed to 120W TDP, which TheJiral decided was “too nerve-racking,” with the system only just managing to stabilize at 98.8°C after 2 hours of continued load.

A new portable gaming PC, which features a 14-inch OLED display and AMD Ryzen AI Max+ 395 processor, is being teased by One-Netbook on social media. We got an email from the firm, best known for its handheld gaming PCs and eGPUs, sharing a few other tantalizing details of this Microsoft Surface-like 2-in-1 design, which it has dubbed the OneXPlayer Super X.

The above video shares only a meager amount of specifications, but One-Netbook tells us the headline specs are:

• AMD Ryzen AI Max+ 395 processor & Radeon 8060S graphics (which One-Netbook claims delivers “RTX 4060 Ti level” performance in this instance)
• 14-inch 2.8K AMOLED HDR display with 100% DCI-P3 & VRR
• Up to 128GB LPDDR5x RAM, dual SSD expansion, and 83.5Wh battery
• And there’s an XDNA 2 NPU onboard delivering 50+ TOPS for local LLM execution

Now well past the mid-point of 2025, the Ryzen AI Max+ 395 field is starting to become crowded, with all sorts of devices. However, there’s already a tablet packing this potent SoC, from PCs and components Goliath Asus. It is, therefore, natural to want to compare the upcoming OneXPlayer Super X with the established Asus ROG Flow Z13 2025. Thankfully, One-Netbook produced exactly such a comparison chart for its Discord followers, and you can check through it below.

One-Netbook has had plenty of time to prepare a solid ROG Flow Z13 competitor, and according to this table, it can beat the Asus machine by every important metric. Probably the most pleasing ways it seems to outclass the Asus are in its screen, expandability, and battery capacity. However, it also boasts an impressive 50% greater TDP. On paper, of course.

Portable and handheld PC stalwart GPD today teased the first AMD Ryzen AI Max+ 395-powered handheld. Moreover, in a short video clip shared on X, we get to see the device in action, with someone playing Black Myth: Wukong. We don’t get a summary chart accompanying the video, but this handheld appears to run this popular game at between 170 and 212 fps for the duration of the gameplay clip.

AMD’s ‘Strix Halo’ Ryzen AI Max+ chips debuted at CES 2025, at the start of the year, with the Ryzen AI Max+ 395 being the halo product of the family. Its blend of RDNA 3.5 graphics and Zen 5 CPU cores – plus radical new memory tech, offering up to 256 GB/s bandwidth. It’s a natural for a premium handheld, but we’ve only seen rumors and suggestions of such handhelds in the making until now.

GPD’s “breaking news” post doesn’t reveal a lot about the GPD WIN 5 it is teasing. However, we can see a few interesting things straight away. The pocketable PCs specialist has chosen to show a few key stats overlaid on the gaming video: processor temperature, iGPU clock speed, CPU power consumption, and frames per second performance.

Watching these stats, we note that, during the brief 20s or so we get to see, the processor hovers around the 65 degrees C region, which looks good for a handheld. At the same time, the ‘CPU Power’ varies between 55W and 58W. The iGPU, the Radeon 8060S, also maintains a clock speed of around 2.5 GHz. As per our intro, Black Myth: Wukong runs from between around 170 and 212fps for the duration of the clip. Admittedly, there are a lot of stats and settings missing from our view, like the graphics quality settings used.

GPD WIN 5 controls were hidden in the shadows

After grabbing the best quality video we could from the source and fiddling with levels in Photoshop, the joyvpad and buttons layout of the GPD WIN5 comes into view.

There are no big surprises in the buttons, pads, and sticks plucked from the darkness of the promo video. On the left, it looks like there is a thumb stick at the top, with a D-pad below, then three buttons and what looks like a speaker grille.

Chinese ODM Shanghai Sixunited Intelligent Technology Co, Ltd. (via ITHome) has revealed the company's upcoming STHT1 motherboard. The STHT1 will leverage AMD's Ryzen AI Max 300-series (codenamed Strix Halo) processors.

The STHT1 conforms to the thin Mini-ITX standard, a specification that targets AIO systems. Size-wise, it still conforms to the Mini-ITX specifications, which are 170 x 170mm. However, the difference lies in the z-height, which is substantially lower, allowing these motherboards to fit into tighter spaces, such as AIO and small form factor (SFF) systems.

If you've been out of the loop, Strix Halo features AMD's latest Zen 5 cores and an improved RDNA 3.5 graphics engine. The core count on Strix Halo ranges between six and 16 cores. The Radeon 8060S unit rocks up to 40 compute units.

However, Sixunited hasn't disclosed which processor option will be available for consumers. The Strix Halo lineup presently consists of three models: Ryzen AI Max+ 395 (16 cores), Ryzen AI Max 390 (12 cores), and Ryzen AI Max 385 (8 cores).

MSI brought its first gaming handheld with an AMD chip inside, the Claw A8 BZ2EM, to Computex. The company also used Taipei tradeshow to launch the latest versions of its Crosshair and Cyborg laptops.

MSI's previous gaming handhelds, like its Claw 8 AI+, used Intel's Core Ultra processors — the only mainstream gaming company to do so. It doesn't seem that MSI is abandoning Intel, but is now making systems with chips from both partners.

Beyond using AMD's Ryzen Z2 Extreme chip, the Claw A8 will have other new features. The system, which is very green, has an Xbox-style controller layout, placing the right thumbstick below the A/B/X/Y buttons. MSI is using Hall Effect joysticks, which should help prevent stick drift. The Claw A8 also has Hall Effect sensors in the LT and RT triggers.

The company has rounded the corners and put in bigger grips, which should increase comfort over previous Claw systems.

The option for up to 24GB of RAM is impressive (especially fast LPDDR5x-8000 memory), but MSI hasn't said anything about what configurations will go to which markets, so we'll see how systems launch with that much memory. Asus used 24GB of memory in the ROG Ally X, one of the best PC Gaming Handhelds, which benefitted because the CPU and GPU share memory.

The 80 WHr battery on the Claw is also identical in capacity to the Ally X, which lasted far longer than the original ROG Ally with a smaller 40 WHr battery.

The top of the system features plenty of ports, including a headphone jack, microSD card reader, two USB 4 Type-C ports, as well as a volume rocker and a power button with a built-in fingerprint reader. While the microSD card reader is expected at this point, MSI is also using a standard M.2 2280 SSD, so in theory you could easily crack the system open and replace it with a standard-sized M.2 drive.

MSI has refreshed its MSI Center software to be accessible via Xbox Game Bar. This should take up less space and allow for extra customization, but it's still building on top of Windows 11, which hasn't proven to be the most intuitive operating system for gaming handhelds.

MSI hasn't announced pricing or a release date for the Claw A8.

Gaming Laptops

Like Acer and Asus, MSI brought its mid-range gaming notebooks to Computex, upgrading them to Nvidia's new RTX 5060 and the RTX 5070.

The new Crosshairs will come in 16, 17, and 18-inch sizes. Confusingly enough, using the 18-inchers as examples, the Intel models are styled Crosshair 18 HX AI, while the AMD versions are Crosshair A18HX. So AI means Intel Core Ultra 200 HX chips here, while the A-series systems use AMD Ryzen 8000.

Other than that, the systems have a ton of commonalities. They all use Nvidia GeForce RTX 5070 or RTX 5060 GPUs "and below" (whether that means an upcoming RTX 5050 or older generation chips remains to be seen) and go up to 115W on the GPU and 170W on the CPU. All of the systems have 90 WHr batteries, Wi-Fi 6E, and 2560 x 1600, 240 Hz screens.

MSI only sent along configurations for the Crosshair 18HX AI, which will start at $1,699 at Costco and features a Core Ultra 9 275HX, Nvidia GeForce RTX 5070, 32GB of RAM, and 1TB of storage. For $2,049, you'll be able to get 64GB of RAM.

MSI hasn't detailed full specifications, so while we know Intel will have slightly faster RAM, we don't know how much RAM many of these systems will ship with. We do know that they have a MUX switch on the motherboard, new crosshair logos on the hinge and touchpad, and, on the 18-inchers, a 24-zone RGB keyboard. MSI didn't provide pricing or release information for these systems.

There's also a refresh to the Cyborg line, which comes in 15 and 17-inch sizes. These systems use translucent plastic on the edges but still have a metal lid. The Cyborg 15 and Cyborg 17 will use Intel Core 200H chips and RTX 5060, 5070, "and below" graphics cards, while the Cyborg A15 AI and A17 AI (yes, here AI stands for AMD), will opt for Ryzen 200 series processors.

The 15-inch screens come in either 1080p at 144 Hz or 1440p at 165 Hz. if you opt for 17 inches, MSI has gone for a 1080p/144 Hz screen or a 1440p, 240 Hz option.

Again, we only have partial pricing here, and no availability. The Cyborg A15 AI will start at $1,149 at Costco with a Ryzen 7 260, RTX 5050 (which we were surprised to see in a sheet sent to press), 16GB of RAM, and 1TB of storage. For $1,399 at Best Buy, you can go up to a Ryzen 9 270 and RTX 5060. The 17-inch AMD model will be $1,549 at Costco with the Ryzen 9 270, RTX 5060, 32GB of RAM, and 1TB of storage.

MSI's other Computex announcements include some productivity laptops, including a new Stealth A16 AI+ and Prestige 16 AI+ notebook that further MSI's partnership with AMG Motorsport. Additionally, there's a new Prestige 13 AI+ "Ukiyuo-e Edition," a limited edition showcasing "The Great Wave off Kanagawa" that will be individually crafted by artists.

Computex 2025 is currently in full swing, and Asus is eager to show off its latest gaming innovations at the expo. Tom’s Hardware is on the ground, and our team looked through the company’s latest developments.

This includes some new peripherals, such as the ROG Falcata split keyboard and ROG Harpe II Ace mouse, and new GPU innovations, like the RTX 5080 Noctua, which features Noctua’s legendary cooling solutions, new BTF graphics cards that are now also compatible with non-BTF motherboards, and a Thunderbolt 5 GPU dock.

The ROG Falcata is a mechanical split-keyboard, allowing players to place the two halves of the keyboard in the most optimal position. This would make gaming for several hours much more comfortable, as your keyboard will now follow the natural angle of your wrists.

We’ve also tried the ROG Harpe II Ace mouse, which generally felt light and good, similar to other gaming mice from other top-tier manufacturers like Logitech. Hopefully, we can get our hands on these units in our testing lab to see how they stack up against our best gaming keyboards and best gaming mice.

Asus has several interesting new GPU-related innovations on show in Taiwan. It released an RTX 5080 version of its Noctua Edition GPU, which features three Noctua NF-A12x25 G2 120mm fans. This massive cooling solution allows the graphics card to run quieter, but it also makes it one massive GPU. The Asus GeForce RTX 5080 Noctua Edition is about four slots thick and over 360 mm long.

The company also released the next generation of its BTF GPUs, which now come with an adapter. The first version of these BTF GPUs had extra gold connectors for power that made it impossible to plug into a non-BTF motherboard.

But this time, Asus recessed those extra connectors, allowing you to plug them into a non-BTF motherboard and then use the usual 12V 2x6 power connector. And when you finally upgrade to a BTF mobo, you can just attach the included adapter, allowing the GPU to get power straight from the motherboard instead of the connector.

Aside from looking much cleaner, it also has the potential to solve the melting issues that many high-powered GPUs have with the 12VHPWR and its successor connector. Hopefully, this implementation is far more robust (especially as the adapter is made from metal), but only time will tell if it runs into issues.

Asus also showed off the ROG XG Station 3 — a Thunderbolt 5 external GPU dock that delivers up to 80Gbps of performance and is compatible with the latest GPUs from AMD and Nvidia.

• Keep up with everything Computex 2025 with our live blog.

AMD is reportedly going big on core-counts with Zen 6 mobile next-generation, as suggested by a new rumor from HXL, backed by several other leakers. Codenamed "Medusa Point", these APUs will reportedly carry up to 22 hybrid cores, based on Zen 6, with classic, dense, and low-power options. Since Zen 6 is at least a year off, and mobile versions might not arrive until early 2027, we need to be careful about putting too much faith in this leak.

Medusa Point is slated to be the follow-up to AMD's current Zen 5-based Strix Point APU series. We probably won't see a direct shift to Medusa Point as AMD is reportedly working on the Gorgon Point (Strix Point refresh) family, planned as an intermediate step.

That being said, architecturally, Medusa Point will switch to the Zen 6 architecture, which should be detailed by AMD sometime around Computex next year. The graphics engine will allegedly adopt the updated RDNA 3.5+ design, though RDNA 4 would've been the ideal choice for many. It's probable RDNA 4 won't make its way to the APU landscape, since AMD's next-generation graphics architecture, UDNA 1 / RDNA 5, is projected for release during the same timeframe.

The mainstream Ryzen 5 and Ryzen 7 offerings from Medusa Point have been purported to feature up to 10 hybrid cores, divided across four classic Zen 6 cores, four dense Zen 6c cores, and two new LP (Low Power) cores. These LP cores are very likely smaller than their Zen 6c siblings, with their Voltage/Frequency operation tweaked for maximum efficiency. This is complemented by an eight Compute Unit equipped RDNA 3.5+ based graphics engine, similar to the Radeon 860M. The iGPU is a downgrade from the current 16-CU design on the Radeon 890M, but this was likely done to free up space on the chip for other components.

Zotac is preparing to launch new mini-PC offerings at Computex next week, featuring AMD's beastly Strix Halo APUs and desktop Blackwell GPUs from Nvidia. So far, the company has showcased two designs with Nvidia hardware, featuring the desktop RTX 5060 Ti and the RTX 5070, and one AMD-based design with a Ryzen AI MAX CPU. Details on exact specifications and pricing have not been shared, but we'll hear more from the team at Taipei in just a few days.

Under its ZBOX Mini PC offerings, Zotac's Magnus-E lineup features compact yet powerful mini-PCs targeted at enthusiasts and content creators. The latest Strix Halo addition should extend their appeal to AI/ML developers as well. The upcoming Magnus One is equipped with an 8.48-liter chassis, housing a desktop RTX 5070 Ti 16GB along with a Core Ultra 7 processor from Intel, potentially from the Core Ultra 200S series.

The Magnus EN follows with a compact 2.65-liter design, home to the RTX 5060 Ti 16GB (desktop). This is accompanied by a mobile Core Ultra 7 processor, which can either be the Core Ultra 7 265H or its HX counterpart. With Arrow Lake, the distinction between Intel's H and HX grade processors goes beyond TDP and core counts. Arrow Lake-H features rebadged SoC Tiles from Meteor Lake, which can potentially offset performance gains, as shown in a previous review.

The Ryzen 5 7533HS is a new Rembrandt-R-based APU that AMD quietly introduced to the market last year and is just now appearing in budget laptops via realVictor_M at X. To clarify, this isn't exactly a brand-new chip; it is simply a rebadged Ryzen 5 7535HS, with lower boost clocks and presumably a more attractive price tag for partner(s).

The Ryzen 5 7533HS breaks away from AMD's original naming convention for the Ryzen 7000/8000 series, where the last digit was used to indicate versions within an architecture ('0' for the lower model and '5' for the upper model), like Zen 3 (7530) versus Zen 3+ (7535). This isn't entirely surprising as AMD has a history of abandoning naming schemes, evident with their latest shift to "Ryzen AI". The Ryzen 5 7533HS is tagged under the Rembrandt-R family, with Zen 3+ cores and an RDNA 2-based iGPU.

In terms of specifications, the Ryzen 5 7533HS is nothing special. It wields a six-core / twelve-thread layout, along with 3MB of L2 and 16MB of L3 cache. The included Radeon 660M iGPU offers six RDNA2-based Compute Units. There's a 150 MHz drop in boost clocks compared to the Ryzen 5 7535HS. As the 'HS' modifier indicates, the APU has a configurable TDP between 35W and 54W, with support for DDR5-4800 (SODIMM) and LPDDR5-6400 (Soldered RAM) memory types.

Die shots of AMD's 16-core Zen 5c CCD used in its latest EPYC 9005 series server processors have been exposed, revealing clear differences compared to AMD's previous-generation Zen 4c CCDs. Posted by HXL on X, photos of AMD's 16-core 3nm Zen 5c CCD expose a long row of two banks of cores flanking a 32MB layer of L3 cache in the middle of the die.

AMD's Zen 5c dies are substantially longer than its vanilla Zen 5 CCD, measuring 5.7 mm by 14.83 mm. Zen 5, by contrast, measures 7.4mm by 11.26mm. The extra length was added to accommodate the additional cores each Zen 5c CCD contains, while simultaneously enabling all cores to share a single unified L3 cache, making Zen 5c AMD's first compact core architecture to sport a single-CCX design.

Click "see more" to see the die shots:

Stability AI, the company behind Stable Diffusion, has released ONNX-optimized models that run up to 3.3x faster on compatible AMD hardware, including select Radeon GPUs and Ryzen AI APUs on mobile.

Amuse is a platform developed by AMD and TensorStack AI that allows users to generate images and short videos locally, on AMD hardware. The latest Amuse 3.0 release not only supports these updated models but also introduces a range of new features, including video diffusion, AI photo filters, and local text-to-image generation. Amuse 3.0 and AMD optimized models require the Adrenalin 24.30.31.05 preview drivers or the upcoming Adrenalin 25.4.1 mainline release.

Over the past year, AMD has partnered with several OSVs, OEMs, and ISVs to optimize AI applications from the ground up, incorporating hardware optimizations, efficient drivers, compilers, and optimized ML models, among other enhancements. Building on this partnership, Stability AI has launched Radeon-optimized versions of its Stable Diffusion family, which include Stable Diffusion 3.5 (SD3.5) and Stable Diffusion XL Turbo (SDXL Turbo). First-party metrics report a 3.3x speedup with SD3.5 Large, dropping to 2.1x for SD3.5 Large Turbo and 1.5x for SDXL Turbo compared to the base PyTorch implementations.

AMD released ROCm 6.4, bringing along several improvements, including framework enhancements, broader OS support, and enhancements to several performance and profiling tools. What it continues to lack is enablement for RDNA 4 GPUs, which may discourage developers from shifting to AMD's latest architecture.

When RDNA 4 hit shelves last month, day-one ROCm compatibility was highly anticipated, as was teased by AMD's VP of AI Software. ROCm is AMD's open-source GPU programming platform, rivaling Nvidia's CUDA ecosystem to power HPC and AI applications. ROCm is chiefly focused on AMD's MI Instinct Accelerators and its prosumer Radeon Pro family, but support has slowly been trickling down to consumer-grade Radeon GPUs, just not as fast as CUDA.

RDNA 4 brings several architectural improvements to the table, doubling FP16 operations per cycle, with an x8 increase in INT4 operations with sparsity. Furthermore, with support for FP8, RDNA 4 can dish out an eightfold increase over RDNA3's FP16 capabilities, again with sparsity. Without official ROCm support, these enhancements are effectively dormant. And, ironically enough, the first mention of Navi 48 (the GPU that powers the RX 9070 series) traces back to a ROCm patch last year.

In any case, AMD introduced several changes with ROCm 6.4, which can be summarized as:

• Support for CPX mode with NPS4 memory mode.
• Support for PyTorch 2.6 and 2.5.
• Support for VP9 with rocDecode/rocPyDecode.
• Several improvements to the ROCm Compute Profiler.
• Support for Oracle Linux 9 and the Radeon PRO W7800 GPU.

Despite recent pushes to challenge the CUDA moat, AMD's hardware support remains perpetually behind Nvidia's. ROCm support for consumer-grade Radeon GPUs on Windows, which began in 2022, now extends across almost all GPUs from the RDNA 2 and RDNA 3 families, excluding HIP SDK support for the RX 6600 to RX 6750 XT range. I said "almost all" as the list lacks the RX 7650 GRE and the RX 7900 GRE. The Linux side is even more dire, compatible with only four Radeon GPUs.

It's not all bad news, as you can now use ROCm on AMD's Strix Halo family of APUs featuring up to 128GB of memory, making them great options for powering AI and HPC workloads on the go. AMD generally releases ROCm updates on a monthly cadence, so we could see RDNA 4 support spring up with the next release. However, this inconsistency could sway developers towards Nvidia, which offers better and more predictable compatibility.

AMD has officially announced its Ryzen 8000HX Series of mobile CPUs today, intended for gaming and other high-end applications. Codenamed Dragon Range Refresh, the chips serve as a refresh of AMD's Ryzen 7000HX family, AMD's highest-performance laptop CPU line that has been in production since mid-2023.

Coming in four unique SKUs, all aiming for high-end performance, the Ryzen 8000HX series comes in up to 16 cores and 32 threads, clocked at up to 5.4 GHz. The chips come with up to 80MB of total cache and are paired with Radeon 610M onboard graphics, a two-core graphics solution meant to be supplanted by a dedicated GPU.

Dragon Range Refresh was leaked several times by AMD partners before its announcement today, with MSI and Asus prematurely announcing the chips in their new products several times since January. Originally expected to be announced at CES 2025, the Ryzen 9 8940HX was prematurely leaked in a listing for the ROG Strix G16 2025 yesterday, confirming its stats and speeds.

As a refresh of the Dragon Range 7000HX series, the 8000HX doesn't offer anything consumers haven't seen before, besides a refined polish on a successful laptop CPU family. Built on the 5nm node, Dragon Range Refresh uses AMD's Zen 4 architecture to bring chiplet tech to high-end gaming laptops. Reviews of the 7000HX series found it to be a fierce competitor against lacking Intel competition, and the 8000HX is expected to be a continuation of the same impressive run.

The refresh doesn't seem to provide consumers with any serious boosts or differences in performance over Dragon Range's first go-around, as can be seen in the chart below. The only change visible from these basic speeds and feeds is a 100 MHz increase in boost clock to the Ryzen 9 8940HX from the 7940HX.

Part of Dragon Range Refresh seems to be consolidation. Where the Ryzen 7000HX series offered 7 SKUs, the refresh cuts this down to 4. We will be interested to see if 8000HX ever offers an X3D model like the Ryzen 9 7945HX3D, but for now it does not seem like the case.

The refresh was also carried out with relatively little fanfare; where Intel's lackluster 2024 Raptor Lake Refresh was highly advertised and marketed, Dragon Range Refresh was heralded by a simple, graphic-less press release and a few new product pages on the AMD website. Clearly, the 8000 HX series is not seeking to be a bold new step for AMD, but a simple sidegrade, likely at some cost advantage thanks to reusing old wafer designs.

Major laptop brands will likely begin offering laptops containing the Ryzen 8000HX series soon. We already know from their leaks that MSI and Asus will be among the first to bring Ryzen 9 8940HX to market. Those looking for the best in laptop performance will likely seek out 8000HX, but those not needing the best and brightest can turn to now-last-gen 7000HX machines as long as they hang around.

AMD is allegedly planning a successor to its Strix Point lineup of APUs for laptops, codenamed Gorgon Point per JNtechreview (via Harukaze). The reviewer shared several slides on their Naver Cafe channel from the LG Gram X Ryzen launch event that seems to be under embargo, revealing a previously unannounced APU family from AMD that reportedly targets a 2026 launch. There are also talks of a new AI Max series succeeding Strix Halo under the "Medusa" lineup. Even if the slides seem real, it's best to approach them with caution, as AMD has not officially confirmed these products.

Gorgon Point allegedly serves as a drop-in replacement for existing Strix Point designs, supporting the same FP8 package. There aren't many architectural changes, as Gorgon Point seemingly sticks with the same Zen 5/5c CPU cores, an RDNA 3.5-based integrated GPU, and an NPU based on XDNA 2. The entire lineup screams Strix Point Refresh, apart from a new Ryzen AI 3 addition, likely to be reserved for entry-level laptops.

Per the slides, AMD's mobile offerings will continue with Hawk Point (Ryzen 200) for mainstream devices, while on the premium end, Gorgon Point is expected to supersede Strix Point next year. Following a two-year cadence, we anticipate Zen 6 to launch sometime later in 2025. However, the slides mention no Zen 6-based APU (likely to be named Medusa Point) for 2025 or 2026, although we could be reading too much into what's essentially leaked information. As a side note, Intel's competing Panther Lake is slated for release later in 2025 and is expected to feature Cougar Cove and Darkmont cores alongside an integrated GPU based on Xe3 (Celestial).

Moving over to performance, and AMD's internal projections depict single-digit performance increments versus incumbent Strix Point APUs, if these slides are genuine. With the respective Ryzen AI 300 (Strix Point) APU at 15W as the baseline (100%), we've found that the most impressive wins land at just 5-6%, largely due to the faster clock speeds and improved binning.

It was expected for AMD to target Gorgon Point (Strix Point Refresh) for the mainstream market and a potential Medusa Point for premium notebooks next year, similar to how Ryzen 200 (Hawk Point Refresh) chips recently launched as an entry-level option. However, that doesn't seem to be the case, at least according to this leak.

Furthermore, the source also mentions that the successor to the beastly Strix Halo will be available under the Medusa series. AMD will probably incorporate its then-latest architectures, such as Zen 6 and RDNA 4 or the first iteration of UDNA, if we're lucky, with these chips. Sadly, an expected release window has not been specified, although we could be looking as far off as late 2026 or early 2027.

During a Q&A session with LG and AMD, it was supposedly mentioned that there are no plans to bring RDNA 4 discrete GPUs to laptops. This much was already said by AMD in broad terms previously, where it indicated RDNA 4 desktop GPUs are their first priority. All in all, there are still some gaps in AMD's next-generation mobile portfolio, though we're hopeful that time will bring more clarity.

China's GMKTec has announced the EVO-X2, which it claims is "the world's first AI mini PC equipped with the AMD Ryzen AI Max+ 395 processor." The device was shown off at the AMD Greater China Channel Summit today, with top red-team execs like CEO Lisa Su and SVP & GM Jack Huynh in attendance. GMKTec was lucky enough to get Dr. Su to sign one or more of the new EVO-X2 mini PCs, and the source blog hints there is a batch of Signature Editions.

Detailed hardware specs of the EVO-X2 are not readily available, and the product page has yet to be published by GMKTec, so you'll have to settle for a brief outline for now. As far as the internals go, all we are sure about is that this compact desktop will pack an AMD Ryzen AI Max+ 395 processor. However, its performance in DeepSeek R1 tests is more than 3X that of an RTX 5080 desktop GPU, according to purported GMKTec hardware demonstrations.

To recap the AMD Strix Halo APU's specs, it features a Zen 5 architecture CPU configured with 16 cores and 32 threads (two CCDs), operating at a maximum frequency of 5.1 GHz. As an APU, this is joined by a beefy iGPU, namely the RDNA 3.5 architecture Radeon 8060S with 40 compute units. Rounding off the processing package, there is an XDNA2 NPU which is capable of 50 TOPS. AMD has also shoved 80MB cache onboard, but buyers will have a choice of the RAM quota (up to 128GB of 256-bit LPDDR5X-8533 RAM) at time of purchase.

Back to pixel peeping to unearth more details about the GMKTec EVO-X2, and one of the first comments we can make is that it is pretty sizable for a mini PC. In the image featuring Jack Huynh, you can see the AMD SVP & GM holding a Geekom machine, which appears to be a traditional '4 x 4' NUC-a-like. With the EVO-X2 in close proximity, we Photoshop-estimate that the EVO-X2 is a '5 x 4.5' and larger in the remaining dimension.

Images shared by GMKTec on social media provide further ideas about the shape and form of the EVO-X2. Also, we can clearly see a power button to the front, alongside an SD card reader, a Thunderbolt (Type C) port, twin USB (Type A ports), and a headset jack. Around the back are three more USB (Type-A) ports, another Thunderbolt, HDMI, DP, an RJ45 network port, another headset jack, and a barrel power jack.

Enthusiasts are excited about the new generation Strix Halo APUs from AMD. So if GMKTec can get its product on shelves ahead of the likes of the HP Z2 Mini G1a and Framework Desktop - at an attractive price - it will have won a minor coup. We don't know how many 'Lisa Su' signed models there are, but the source blog hints there is a batch.

As for who will be first with an AMD Strix Halo, Asus originally slated its ROG Flow Z13 for release in February, but it still seems to be in pre-order purgatory, and it's not a mini PC. The Framework Desktop machines with Ryzen AI Max chips aren't expected until Q3 this year. The other machine we have heard about, the HP Z2 Mini G1a, is set to be released in 'spring.'

AMD's fire-breathing Ryzen AI Max+ 395 allegedly crushes Intel's latest efficiency-focused Lunar Lake CPUs in AI benchmarks. An AMD blog post claims the new Zen 5 + RDNA 3.5 chip is up to 12.2x faster than the Core Ultra 7 258V.

AMD benchmarked the Ryzen AI Max+ 395 and Core Ultra 7 258V (with Arc 140V graphics) in a variety of large language models and LLM configurations, including DeepSeek R1 and Llama. Model sizes were restricted to 16GB to offer a fairer comparison against Lunar Lake-powered laptops with 32GB of memory (the highest memory configuration available for these devices). An Asus ROG Flow Z13 64GB was used for the Ryzen AI Max+ 395 test system, and an Asus Zenbook S14 32GB was used for the Core Ultra 7 258V test system.

In DeepSeek R1, the Ryzen chip was up to 2.1x faster (measured in tokens per second) than the Intel counterpart using Distill Qwen 1.5b, up to 2.2x faster using Distill Qwen 7b; up to 2.1x faster using Distill Llama 8b; and up to 2.2x faster using Distill Qwen 14b. In Phi 4 Mini Instruct 3.8b, the Ryzen chip was up to 2.1x faster than the Intel Lunar Lake chip; up to 2.2x faster in Phi 4 14b; and up to 2.1x faster in Llama 3.2 3b Instruct.

In the same LLM configurations but benchmarking the "time to the first token," the Ryzen AI Max+ 395 was up to 12.2x faster in DeepSeek R1 Distill Qwen 14b. The Zen 5 chip's least-performant dominance was in Phi 4 Mini Instruct 3.8b and Llama 3.2 3b Instruct, where the AMD chip was "only" 4x faster than the Core Ultra 7 258V.

AMD showed similar dominance in AI vision models using the same "time to the first token" benchmarking technique. In IBM Granite Vision 3.2 2B, the 395 was up to 7x faster than the 258V, up to 4.6x faster in Google Gemma 3.4b, and up to 6x faster in Google Gemma 3 12b.

AMD's benchmarks show complete dominance of its Ryzen AI Max+ 395 against the Core Ultra 7 258V in AI benchmarks. This is all thanks to the Ryzen AI Max CPU's significantly more powerful integrated graphics chip (which rivals discrete graphics with its 40 RDNA 3.5 CUs), eight more CPU cores, and its significantly higher configurable TDP (rated up to 120W). Even though it consumes significantly more power than the Core Ultra 7 258V (which has a max turbo power of 37W), both chips operate in the same market, and are compatible in the same thin-and-light category of laptop PCs.

It will be interesting to see how the new AMD mobile APUs shape up against Nvidia's RTX 50-series mobile GPUs, which are reportedly facing supply chain issues, delaying their launch in upcoming RTX 50 series gaming laptops. On a pure performance level (not considering form factor), these new Nvidia-powered systems will be AMD's primary competition.

AMD is allegedly well on its way to handling discrete GPU competition, as it already advertised superior AI performance on the Ryzen AI Max+ 395 against Nvidia's RTX 4090 laptop GPU.

AMD is swinging back at Nvidia with new DeepSeek benchmarks that claim its monster 48GB RDNA 3 GPUs can outperform Team Green's previous-generation RTX 4090.

David McAfee, AMD vice president and general manager of Ryzen CPUs and Radeon graphics, posted on X that the Radeon Pro W7900 and Pro W7800 48GB cards can outperform an RTX 4090 by up to 7.3x in DeepSeek R1.

McAfee shared a graph of the three GPUs benchmarked in several iterations of DeepSeek R1 using LM Studio 0.3.12 and Llama.cpp runtime 1.18. The DeepSeek R1 iterations consisted of Distill Qwen 32B 8-bit, Distill Llama 70B 4-bit, Distill Qwen 32B 8-bit, and Distill Llama 70B 4-bit. The former two were configured to output conversational prompts (with 20 tokens) and the latter summarization prompts (with 3017 tokens).

Click See more to see the benchmark results:

In DeepSeek R1 Distill Qwen 32B 8-bit, the RTX 4090 allegedly produced 2.7 tokens a second, the Pro W7800 48GB produced 19.1, and the Pro W7900 48GB produced 19.8 tokens per second. In Distill Llama 70B 4-bit, the RTX 4090 produced 2.3 tokens a second, the Pro W7800 48GB 12.8, and the Pro W7900 48GB 12.7 tokens a second.

In Distill Qwen 32B 8-bit, the RTX 4090 produced 2.5 tokens per second, Pro W7800 48GB 15.7 and Pro W7900 48GB 16.2 tokens per second. In R1 Distill Llama 70B 4-bit, the RTX 4090 produced two tokens per second, Pro W7800 48GB 10.1 and Pro W7900 48GB 10.4 tokens per second.

AMD's benchmarks claim the Radeon Pro W7800 or Pro W7900 48GB GPUs are up to 7.3x faster in Distill Qwen 32B 8-bit, 5.5x faster in Distill Llama 70B 4-bit, 6.5x faster in Distill Qwen 32B 8-bit, and 5.2x faster in Distill Llama 70B 4-bit compared to the RTX 4090.

David McAfee claims the 48GB trims of the WPro W7800 and W7900 have enough VRAM to run the largest DeepSeek R1 models. VRAM is one of the most critical aspects of processing large language models; parameters for LLMs are stored directly in VRAM and are directly proportional to the model sizes. Thus, the larger an LLM is, the more VRAM you need. But with the extra VRAM capacity comes very high prices.

The W7900 48GB costs a whopping $3,500 — $1,500 over the RTX 5090's $2,000 MSRP and $2,000 over the RTX 4090's $1,500 MSRP (though hardly any 4090's were sold at that price). But on the flip side, the 48GB RDNA 3 GPU is less than half the price of the closest current-generation 48GB Nvidia GPU you can buy today, the RTX A6000 Ada.

AMD's marketing looks great, but we have seen this before. AMD previously shared benchmarks of its RX 7900 XTX outperforming the RTX 4090 (mostly) in DeepSeek R1 benchmarks. However, Nvidia responded by showcasing benchmarks of the RTX 4090 (and RTX 5090), drastically outperforming the flagship RDNA 3 GPU with the same DeepSeek R1 configurations.

AMD also neglected to share any benchmarks comparing Nvidia's newest flagship, the RTX 5090, against its RDNA 3-based 48GB workstation-focused graphics cards. It will be interesting to see if Nvidia will follow up with another round of benchmarks to combat AMD, particularly since AMD has more VRAM on its 48GB cards than even the RTX 5090 with its 32GB of GDDR7.

Tom’s Hardware has tested dozens of processors from AMD and Intel to find the best CPU for gaming. Our list of 2026 CPU gaming benchmarks currently comprises 17 of the most demanding titles available on the market, which we run each gaming processor through to see the chips that come out on top. We select our picks based on the data in our CPU benchmark hierarchy, so all of the CPUs below are backed by hundreds of hours of real-world, hands-on testing where we gather extensive data on how a CPU performs and behaves while gaming. If you want a broader look at the CPU market, our AMD vs. Intel article shows you where the current CPU duopoly stands, while our CPU buying guide can help you narrow down the best processor for you.

Now in the back half of the year, we don’t expect major new releases from AMD or Intel. AMD has been on a tear with refreshes, particularly among X3D CPUs. We recently reviewed the Ryzen 9 9950X3D2 Dual Edition, AMD’s first CPU with 3D V-Cache on both CCDs, and it’s the most powerful chip from Team Red currently available. It’s earned a spot on this list, though the (much cheaper) Ryzen 9 9950X3D isn’t far behind in overall performance.

We’ve also seen the Ryzen 7 9850X3D, which is technically the fastest gaming processor on the market, as you can see in our Ryzen 7 9850X3D review. However, we’ve kept the base Ryzen 7 9800X3D as our top recommendation for the best CPU for gaming due to its price. It’s only marginally behind the refreshed model (about 3% on average), and much cheaper. For most gamers, the Ryzen 7 9800X3D makes more sense.

Intel has seen a recent boost in gaming performance with Arrow Lake Refresh, and both the Core Ultra 5 250K Plus and Core Ultra 7 270K Plus have earned spots on our list. AMD dominates in gaming at the moment, however. Our sights are set on Intel’s next-gen Nova Lake chips for a big gaming boost from Team Blue, as well as the rumored ‘Raptor Lake Next’ lineup that’s supposedly arriving early next year.

For the rest of the year, we have the Ryzen 7 7700X3D and Ryzen 7 5800X3D 10th Anniversary Edition to look forward to, which are arriving in July and June, respectively. We don’t anticipate they’ll make a major impact on our rankings here, but we plan on reviewing both CPUs as soon as they’re available.

In addition to the fastest CPUs from AMD and Intel, we’ve included a few DDR4 options on this list. The price of DRAM and NAND flash has made building even a budget PC prohibitively expensive, so DDR4 platforms are a great way to save money. Vendors are signaling a shift back toward DDR4 platforms at the moment, so we may reconsider some older CPUs for our rankings as pricing and availability allows.

Here are the gaming CPUs we recommend buying. We have a shortlist of the top options and some alternatives below, but you can click the ‘More’ links to read our thoughts about a particular CPU and where it stands in the current market.

Best CPU for Gaming in 2026 at a glance (more info below):

The list below is for the best CPUs for gaming, while the best budget CPUs can help you find a cheap chip. Processors benefit from the best thermal paste, so check out our guide if you're shopping for a new processor. But if you're after the best CPU for gaming, you're in the right place.

Best CPU for Gaming Benchmarks

We rank all the Intel and AMD processors based on our in-depth CPU benchmarks hierarchy. You can see some of those numbers in the charts above, including CPU overclock performance results (marked as PBO for AMD processors). We're currently retesting all of these processors with the Nvidia RTX 5090, but only the first four slides have that testing. The remainder are historical testing results with the RTX 4090, which we'll remove once we have fully retested all of the gaming CPUs with the RTX 5090 for our benchmarks. This group of results comprises only the chips that have passed through our newest test suite. Additionally, the tables in our CPU benchmark hierarchy include rankings based on past CPU benchmarks and breakdowns of single- and multi-threaded performance in productivity applications across a broad spate of processors. Finally, be aware that the pricing in the charts above can fluctuate.

Quick Shopping Tips

When choosing the best CPU for gaming in 2026, consider the following:

• You can't lose with AMD or Intel: As noted in our AMD vs. Intel feature, AMD tends to make the best all-around CPU for gaming for mainstream PCs lately, but both offer compelling performance options at any given price point.
• Eight cores is sufficient for gaming: If you’re looking at a pricey flagship, you’re likely wasting some money if gaming is your primary focus. You can game on as little as a quad-core CPU, but performance scaling really falls off past eight cores.
• Budget platform costs: You never want to pair a strong CPU with a weak GPU, RAM, and storage. Right now, it’s especially important to consider platform costs, however. DDR5 prices are peaking, and you’ll need to factor in the cost of DDR5 and a new motherboard if you’re coming from an older socket like AM4.
• Overclocking isn’t for everyone, but if you follow our How to Overclock a CPU guide, you can scrape out extra performance gains.

Best CPU for Gaming 2026 - $300 to $400

The Ryzen 7 9800X3D technically isn’t the fastest gaming chip on the market any more. That title goes to the new Ryzen 7 9850X3D, though the victory is marginal. As you can read in our Ryzen 7 9850X3D review, AMD’s latest X3D offering pushes ahead by 3.3% on average. Despite a minor uplift, we’re still recommending the Ryzen 7 9800X3D. With prices as they currently are, the Ryzen 7 9850X3D is only 3.3% faster despite costing around 6% more than the Ryzen 7 9800X3D.

This chip really has no peer in the market outside of the Ryzen 7 9850X3D — the Ryzen 7 9800X3D delivers outstanding gaming performance, beating Intel's fastest gaming chip, the $469 Core i9-14900K, by 30% in our test suite. The 9800X3D is also almost unbelievably 35% faster than the current-gen Intel flagship, the $560 Core Ultra 9 285K. The stock Ryzen 7 9800X3D's 1% low frame rates (a good smoothness indicator) also deliver an exceptionally smooth gaming experience, benefiting gamers even in GPU-limited scenarios.

The Ryzen 7 9800X3D has eight cores and 16 threads that operate at a 4.7 GHz base and 5.2 GHz boost clock rate. The chip employs AMD's 3D V-Cache tech with a new spin, which places a 3D-stacked SRAM chiplet underneath the die to deliver an incredible 96MB of L3 cache to great effect. AMD moved the L3 cache chiplet from the top to the bottom of the compute die this generation. That gives the integrated heat spreader (IHS) direct access to the compute die, allowing for more thermal headroom, and in turn, higher clock speeds. The end result is a comparatively low-power chip that delivers incredible gaming performance and comparable productivity performance to other eight-core models on the market.

3D V-Cache previously came with trade-offs in the productivity department, but that’s not the case with the Ryzen 7 9800X3D. Still, 3D V-Cache doesn’t provide a performance benefit in every game, and the performance benefit is less pronounced as your display resolution climbs.

The Ryzen 7 9800X3D has much lower power consumption than the Intel competition, making it a far cooler processor that won't require as expensive accommodations, like a beefy cooler, motherboard, and power supply. It also takes particularly well to undervolting, which is easy to accomplish with AMD’s Curve Optimizer. That means the 9800X3D delivers top-notch gaming performance and a cooler, quieter, and less expensive system than you'll get with an Ultra 9 or Core i9.

Read: AMD Ryzen 7 9800X3D Review

The $305 Ryzen 7 9700X had a rough initial product launch, but AMD's targeted firmware and operating system improvements have changed the picture tremendously, allowing the chip to place much higher on our CPU benchmark hierarchy (head there for the most up-to-date gaming benchmarks). Combined with lower-than-launch pricing, the Ryzen 7 9700X is a strong contender, tying Intel's Core i9-14900K in gaming and beating the Core i7-14700K. That's not to mention that it beats Intel's entire lineup of Arrow Lake processors as well. Now, all of those processors offer faster performance in heavily-threaded productivity applications than the 9700X, but when it comes to a pure gaming experience, the 9700X either ties or beats all current Intel competitors.

The Ryzen 7 9700X has eight Zen 5 cores with 16 threads that operate at a 3.8 GHz base and 5.5 GHz boost clock. The chip has a 65W TDP, though AMD retroactively added a 105W TDP option you can select in the BIOS that helps boost performance in productivity applications. It's covered by the warranty, as well. With either setting, the 9700X has comparatively tame power consumption, so it is an easy chip to cool. You'll have to buy your own cooler for the processor, though.

The Ryzen 7 9700X drops into socket AM5 motherboards, and B-series motherboards make the most sense for this class of chip. B850 and B840 motherboards get AMD's latest chipset with features like mandatory PCIe 5.0 support on the top M.2 slot and better availability for features like Wi-Fi 7. However, the Ryzen 7 9700X will still work with the older B650 chipset if you can find a board on sale.

Read: Ryzen 5 9700X Review

The $350 Core Ultra 7 270K Plus performs like a flagship CPU, but it costs about half as much. In games, it narrowly outclasses the Core i7-14700K and offers a 2.4% boost over the competing Ryzen 7 9700X. AMD’s last-gen Ryzen 7 7800X3D still offers around a 10% boost over the Core Ultra 7 270K Plus, but it’s also around $50 to $80 more expensive depending on sales.

It’s a solid gaming CPU, and certainly a better recommendation than the Core i7-14700K given prices right now. Compared to the Ryzen 7 9700X, things are tighter. The Core Ultra 7 270K Plus gains an edge with productivity performance. Short of the 9950X, it’s at the top of our multithreaded performance rankings, more than doubling the performance of the Ryzen 7 9700X.

On the gaming front, it supports Intel’s new Binary Optimization Tool, which offers an average of an 8% improvement in gaming performance based on our testing. It’s only available in a limited number of games at the moment, but Intel says it plans to support the feature with updates in the future.

For specs, the Core Ultra 7 270K Plus is close to the 285K. It comes with 24 cores and threads, split across eight Lion Cove P-cores and 16 Darkmont E-cores. The P-cores boost up to 5.4 GHz and the E-cores can climb to 4.7 GHz. Across the CPU, you get a total of 76 MB of combined L2 and L3 cache. It comes with a 125W TDP and 250W MTP. Critically, the Core Ultra 270K Plus also comes with a 900 MHz boost in die-to-die frequency and 400 MHz boost in fabric frequency compared to stock Arrow Lake chips.

The Core Ultra 7 270K Plus slots into existing 800-series motherboards with the LGA 1851 socket. This is an unlocked chip, so if you want to get the full benefits of overclocking, you’ll need a Z890 board. However, it’ll still work with H- and B-series motherboards, just without CPU overclocking support.

Read: Core Ultra 7 270K Plus Review

Mid-Range Best CPU for Gaming - $200 to $300

Intel has returned to gaming prominence with its Arrow Lake Refresh CPUs, and nowhere is that clearer than with the $220 Core Ultra 5 250K Plus. It’s priced like a budget CPU at $220, but it can perform as well (and sometimes even better) than chips that cost twice as much. It doesn’t dominate the gaming charts in the same way as AMD’s X3D offerings, but at this price, it doesn’t need to. It offers marginally better performance than AMD’s competing six-core Ryzen 5 9600X in games while running the tables with application performance.

On average, the Core Ultra 5 250K Plus is 1% faster than the Ryzen 5 9600X at 1080p, and 9% faster than the 245K. It’s functionally identical, but Intel’s new iBOT feature allows the chip to hold some solid leads in certain titles. For instance, it’s 10% ahead of the 9600X in Cyberpunk 2077. Even in a non-iBOT title like Doom: The Dark Ages, the 250K Plus leads by 12%. There are still some games that struggle with the unique Arrow Lake architecture like F1 2024, but the losses are less pronounced with the souped-up Arrow Lake Refresh chips compared to the stock offerings.

The application performance is what really stands out with the 250K Plus, however. With 18 cores, it outpaces the Core i7-13700K, nearly matches the Core i7-14700K, and more than doubles the performance of the Ryzen 5 9600X in multithreaded applications. In single-threaded applications, it beats the Ryzen 5 9600X by 6%.

Although you get 18 cores, they’re split between six Lion Cove performance cores and 12 Darkmont efficient cores. The P-cores climb up to 5.3 GHz, while the E-cores top out at 4.6 GHz. The CPU comes with a combined 60 MB of L2 and L3 cache, along with a TDP of 125W and a MTP of 159W. Like all Arrow Lake chips, it doesn’t support Hyper-Threading, so you get 18 total threads.

The Core Ultra 5 250K Plus slots into existing motherboards with the LGA 1851 socket. It’s unlocked for overclocking, so a Z-series motherboard is an ideal pairing. However, Intel increased the die-to-die frequency and the fabric frequency out of the box, and you’ll see those improvements in action on B- and H-series motherboards, as well. It’s locked to DDR5 memory, unlike Raptor Lake and Alder Lake platforms, and it officially supports speeds up to 7200MT/s.

Read: Core Ultra 5 250K Plus Review

The $230 Ryzen 5 7600X3D is currently the best value gaming CPU you can get right now, though it trades performance in other areas to reach that status. It's just 4.5% slower than the Ryzen 7 7800X3D based on our testing, giving you most of the performance of AMD's coveted 3D V-Cache in games without the extra cost.

In games, it outclasses more expensive CPUs with ease, including the Ryzen 7 9700X, and averaged just 65W of power draw during our gaming tests. Outside of games, however, the Ryzen 5 7600 X3D struggles. The Core Ultra 5 250K Plus is more than twice as fast in multithreaded performance, and in single-threaded performance, even the base Ryzen 5 7600X is around 13% faster.

The lagging productivity performance makes sense. The Ryzen 5 7600X3D is a six-core / 12-thread chip, so it has limited multithreaded potential, and it only clocks up to 4.7 GHz. The limited specs give AMD room to cram 102 MB of combined L2/L3 cache on the die, however, which comes with a sizeable boost in gaming performance. Compared to the base Ryzen 5 7600X, the X3D version is 22% faster despite coming in at lower peak clocks and power draw.

You can slot the Ryzen 5 7600X3D into socket AM5, which is available on 600- and 800-series motherboards, though the latter may require a BIOS update. Memory and CPU overclocking is available on both B- and X-series chipsets; however, the Ryzen 5 7600X3D has a locked multiplier, so the only overclocking you can access is through AMD's Precision Boost Overdrive, or PBO.

Read: Ryzen 5 7600X3D review

Highest Performance Best CPU for Gaming - $400+

How do you improve upon a CPU that already claims a dominating position in gaming and productivity workloads? You add more cache, of course. The Ryzen 9 9950X3D2 is powerful, expensive, and hungry for wattage, but it’s the best of the best if you want top-shelf gaming and application performance. It throws value out the window, and it’s only marginally better than the Ryzen 9 9950X3D, but it is still better.

Based on our testing, it’s about 3.9% ahead of the Ryzen 9 9950X3D in multithreaded applications, and in lockstep in gaming at 1080p. Compared to Intel’s Core Ultra 7 270K Plus, the Ryzen 9 9950X3D2 is 9% ahead in multithreaded performance and 23% ahead in average gaming performance. The Ryzen 7 9800X3D delivers a better value on the gaming front, and the Core Ultra 7 270K Plus is a monster productivity chip at a third of the price of the 9950X3D2. But the magic trick of this chip is that it can do both without breaking a sweat.

Under the hood, the Ryzen 9 9950X3D2 is similar to the Ryzen 9 9950X3D. It’s a 16-core / 32-thread chip packing AMD’s Zen 5 architecture, and it tops out with a 5.6 GHz boost clock; just 100MHz behind the 9950X3D. As the name suggests, this processor is unique because it uses AMD’s 3D V-Cache on both CCDs. Both eight-core CCDs have 32 MB of onboard cache, plus an additional 64 MB chunk placed under the cores, giving you a total of 192 MB of L3 cache.

The extra cache slightly accelerates multithreaded performance overall, though only by around 4%. There are specific workloads where the advantage is more present, with some data science workloads showing performance gains in the realm of 26% over the Ryzen 9 9950X3D. Those specific workstation-class workloads are where the Ryzen 9 9950X3D2 earns its stripes.

Otherwise, it’s the chip to buy because you simply want the best, no matter what the cost or how marginal the improvements are. It slots into existing AM5 motherboards, and it’s best suited for newer 800-series chipsets. AMD officially supports memory speeds up to DDR5-5600, though we find that DDR5-6000 is the sweet spot for Zen 5 CPUs.

Prices have dropped since release, though the 9950X3D2 is still expensive. It launched at $1,000, but you can find the chip for around $900 now.

Read: AMD Ryzen 9 9950X3D2 Review

The prior-gen $469 Core i9-14900K is now selling for all-time low pricing, primarily because the newer $560 Core Ultra 9 285K has arrived to take its place. However, the Core Ultra 9 285K is actually slower than the 14900K in gaming, so it isn't a suitable replacement. The new Core Ultra 7 270K Plus is marginally slower based on our testing, as well. Even in the face of Intel’s 200S Boost update, which was meant improve gaming performance, the competitive landscape remains unchanged. In our testing, the Core Ultra 9 285K gained an average of 7% from the update, which means it’s still slower than the Core i9-14900K.

You should be aware that the much more economically-priced 14700K (listed above) is only 2% slower than the 14900K in gaming but costs over $100 less. The Ryzen 7 9700X, also listed above, is also less expensive and effectively ties the 14900K in gaming.

However, there are Intel fans willing to pay extra for the absolute most gaming performance they can get from an Intel platform. Also, the 14900K does offer more multi-threaded horsepower than the 14700K and 9700X, which could be useful if you game, stream, and record simultaneously or do other heavy multi-tasking while gaming. Just make sure that your use case justifies the extra cost.

The 14900K sports leading-edge connectivity, supporting DDR4-3200 or up to DDR5-5600 memory, along with 16 lanes of PCIe 5.0 and an additional four lanes of PCIe 4.0 from the chip for M.2 SSDs.

The chip comes with eight P-cores that support Hyper-Threading and 16 single-threaded E-cores for a total of 32 threads. The P-cores have a 3.2 GHz base, and peak frequencies reach an amazing 6.0 GHz with Turbo Boost Max 3.0 (this feature is only active on P-cores). Meanwhile, the E-cores have a 2.4 GHz base and stretch up to 4.4 GHz via the standard Turbo Boost 2.0 algorithms. The chip also has 36MB of L3 cache and 32MB of L2.

This 14900K has a 125W PBP (base) and 253W MTP (peak) power rating, but we recorded considerably lower power consumption than its prior-gen counterpart. You'll need to buy a capable cooler for the chip, and you'll also need either a 700-series or 600-series motherboard. Like other Raptor Lake Refresh chips, you can find DDR4 and DDR5 motherboards, though you’ll need to go with a DDR5 board for the highest performance.

The lower price of DDR4 might entice some gamers, but you'll lose anywhere from 5-8% of gaming performance with higher-end Intel chips. You can step up to the much more expensive DDR5 if you need access to more memory throughput and, thus, every bit of performance possible.

Beyond specs, the Core i9-14900K was at the center of a years-long controversy concerning instability. An error in the microcode (CPU firmware) meant the Core i9-14900K would degrade faster than expected, starting with instability in games. Intel has rectified the issue with microcode 0x12F, so make sure you update your BIOS immediately if you pick up Intel’s last-gen flagship

Read: Intel Core i9-14900K Review

Best Budget CPU Pick - $100 to $150

The $164 Ryzen 5 7600X is an attractive budget CPU at its new price, forced down by Intel's new Arrow Lake Refresh chips. It’s marginally slower than the Ryzen 5 9600X, but also marginally cheaper – the Ryzen 5 7600X offers about 90% of the performance of the Ryzen 5 9600X for 94% of the price. It’s a slightly worse value, but it’s still a good option to keep in mind, especially if you find it on sale. The Ryzen 5 7600, sans X, is available at around the same price. We’ve yet to see it drop below the Ryzen 5 7600X, however.

With the 7600X, you get six cores and 12 threads based on the Zen 4 architecture, clocked at 4.7GHz with boost speeds up to 5.3GHz. Unlike the Ryzen 5 9600X, the Zen 4-based version comes with a TDP of 105W. Cooling it shouldn’t be an issue, and you’re free to run in AMD’s 65W Eco mode through the Ryzen Master software.

The Ryzen 5 7600X slots into AM5 motherboards, including 600- and 800-series chipsets, and it supports PCIe 5.0. DDR5 is required, which is a tough pill to swallow at this bang-for-your-buck price point, but it’s hard to avoid soaring RAM prices.

Read: Ryzen 5 7600X review

The $135 AMD Ryzen 5 5600 delivers a solid blend of performance in both gaming and productivity applications, bringing a new level of value to the Zen 3 lineup. If you're fine sticking with a previous-gen AM4 motherboard, the Ryzen 5 5600 makes a great base for a budget build. The primary trade-off for the AM4 platform is that you're limited to DDR4, and you don't have access to PCIe 5.0. You also have a limited runway for upgrades, as the fastest gaming CPUs on AM4 – the Ryzen 7 5700X3D and Ryzen 7 5800X3D – have reached end of life.

The Ryzen 5 5600 also makes an absolutely unbeatable budget chip if you're updating a first-gen Ryzen system. The 5600 unseats the Ryzen 5 5600X, a long-time favorite. The 5600X is only a mostly imperceptible ~1% faster in gaming and multi-threaded PC work than the non-X model, but provides a 4% advantage in single-threaded work.

Our testing shows that the Ryzen 5 5600 generally matches the gaming performance of its more expensive sibling, the ~$230 Ryzen 7 5800X. That makes the 5600 an incredibly well-rounded chip that can handle gaming well, from competitive-class performance with high refresh rate monitors to multi-tasking gaming workloads like streaming, while also serving up more than enough performance for day-to-day productivity apps. As with all AMD CPUs for gaming, you can fully overclock the chip.

The Ryzen 5 5600 has a 3.7 GHz base and 4.6 GHz boost clock. The chip also has a 65W TDP rating, so it runs cool and quiet. Existing AMD owners with a 500-series motherboard will be happy, as the 5600X drops right into existing 500-, 400-, and 300-series motherboards. If you need a new motherboard to support the chip, AMD's AM4 motherboards are plentiful and relatively affordable, with the B-series lineup offering the best overall value for this class of chip.

Prices for the Ryzen 5 5600 have drifted upward as stock depletes, but that’s offset by platform costs. In addition to low prices on AM4 motherboards, the Ryzen 5 5600 is limited to DDR4. High DDR5 prices are a significant roadblock to opting for a newer chip, as prices continue to surge. So it’s hard to recommend a newer budget CPU, even if it’ll net you higher performance. In the event you already have a kit of DDR5, the $189 Ryzen 5 7600 is a compelling option, and it includes AMD’s Wraith Stealth cooler.

Read: AMD Ryzen 5 5600 Review

Entry-Level Best CPU for Gaming - For gaming on integrated GPUs

The $268 Ryzen 7 8700G, AMD's flagship desktop APU, delivers the fastest socketed performance on the market from integrated graphics, bringing passable 1080p gaming to the desktop PC without a discrete graphics card, but its high price point relegates it to a niche audience.

In contrast, the $191 Ryzen 5 8600G delivers 90% of the 8700G's performance but for ~$80 less, making it a solid alternative for gaming systems that don't use a discrete GPU.

Naturally, you'll have to accept lower fidelity settings and be realistic about which titles can play at 1080p resolution. Still, AMD's Hyper-RX suite of features, which includes in-driver Radeon Super Resolution upscaling tech, frame generation with AMD Fluid Motion Frames (AFMF), Anti-Lag+, and Radeon Boost, helps boost performance at a slight cost to image quality. This new feature set, a first for AMD's iGPUs, is a boon for budget gamers.

The Ryzen 5 8600G has six Zen 4 CPU cores and the RDNA 3 GPU engine with eight CUs. The Ryzen 5 8600G drops into the AM5 platform, with value-focused B650 and A620 motherboards being the obvious best combination. These systems offer a new level of connectivity for AMD's APU processors, which were previously on the aging AM4 platform but require DDR5 memory. That adds some cost, so do a value analysis before selecting this processor. If you're looking for the lowest entry price possible with an APU, the Ryzen 5 5600G listed below slots in as the value alternative.

The Ryzen 7 8600G only supports 16 usable lanes of PCIe 4.0 connectivity, while other processors on the AM5 platform support PCIe 5.0. However, we don't feel this will impact this class of system.

More: AMD Ryzen 7 8700G and Ryzen 5 8600G Review

The Ryzen 5 5600G steps into the arena as the value champ for APUs, which are chips with strong enough integrated graphics that they don't require a discrete GPU for light gaming—just be sure you're willing to accept lowered quality settings.

The Ryzen 5 5600G gives you 96% of the gaming performance on integrated graphics than its more expensive sibling, the Ryzen 7 5700G, but for 25% less cash. Our testing shows that its level of performance makes it the best value APU on the market. As long as you're willing to sacrifice fidelity and resolution and keep your expectations in check, the Ryzen 5 5600G's Vega graphics have surprisingly good performance in gaming.

The 5600G's Vega graphics served up comparatively great 1280x720 gaming across numerous titles in our tests, but options become more restricted at 1080p. Of course, you can get away with 1080p gaming, but you'll need to severely limit the fidelity settings with most titles.

With eight cores and 16 threads that operate at a 3.9 GHz base and boost up to 4.4 GHz, the Ryzen 5 5600G also offers solid performance for its price point in standard desktop PC applications. The chip also comes with a bundled Wraith Stealth cooler, sweetening the value prop, and drops into existing 500-series and some 400-series motherboards, though support on the latter will vary by vendor.

The Ryzen 5 5600G is nearly five years old, and as a result, it’s hard to find it in stock at a reasonable price. AMD updated this model with the Ryzen 5 5600GT in early 2024, which features identical silicon and a slight boost to clock speed, and you’ll generally find it for less at around $150.

If your budget is tight and you're looking to build a system for modest gaming, you should check out our Best Cheap CPU feature. Some of those chips can deliver passable gaming performance without a graphics card, and their prices start at just $55 (£40).

Read: AMD Ryzen 5 5600G Review

• MORE: CPU Benchmark Hierarchy
• MORE: AMD vs Intel
• MORE: Best Cheap CPUs
• MORE: Best CPUs for Workstations

Yesterday, it was reported that AMD's RDNA 4 GPUs will not get ROCm support at launch. Unsurprisingly, this upset many developers who were planning to upgrade their setups and hoped for a seamless transition. However, ROCm support might not be far behind the launch, as AMD's Vice President of AI Software has teased what appears to be an AMD GPU from the RX 9070 series running in a dedicated ROCm environment.

ROCm is an open-source software stack for GPU programming, enabling graphics cards to be used for much more than rendering graphics. ROCm enables select GPUs to power HPC and AI applications, including professional Instinct Accelerators, prosumer Radeon PRO GPUs, and a handful of consumer-grade Radeon GPUs as well.

It seemed lacking for AMD to miss out on the opportunity of providing day-one support to developers. Ironically, the initial traces of Navi 48, the chip beneath RX 9070 GPUs, surfaced in ROCm updates last year. Anush Elangovan reported that RDNA 4 is "ROCm'ing on just fine," demonstrating a potential RX 9070 series GPU under rocm-smi (ROCm System Management Interface).

Developers have repeatedly bemoaned AMD's slow adoption of new ROCm features and support on mainstream hardware. As of writing, AMD officially supports the RX 7000 and RX 6000 families of GPUs and the Radeon VII on Windows. Select RDNA 2 GPUs can only use the HIP runtime and don't have access to the full HIP SDK. Linux support, which is notably better and more consistent, only extends to the RX 7900 series and the Radeon VII. There are a few workarounds to get unsupported hardware up and running with ROCm, which is reportedly more successful in Linux environments.

Nvidia's CUDA, on the contrary, can be run using GPUs dating back to 2006. Of course, newer architectures offer updated feature sets and fancier instructions. Given the recent advances in AI, 90% of Nvidia's revenue is a result of its data-center accelerators, driven by CUDA at their heart. This shows how critical ROCm is for AMD if it wants to compete against the mammoth that Nvidia has become in the AI space.

To broaden ROCm support, the next addition will likely be Strix Halo, or the Ryzen AI MAX 300 series. With as much as 128GB of unified memory, Strix Halo is much more than a gaming laptop; it's more tailored for LLM inference and development, tackling Apple's M-series silicon.

When Framework debuted its Desktop earlier this week, most attendees of its San Francisco event didn't get to tear the machine all the way down to the studs. There's an exception, however: iFixit posted a full teardown of the 4.5L Mini-ITX machine, removing the cooler and showing off the AMD Ryzen AI Max+ 395, also known as Strix Halo.

Teardown tech Shahram Mokhtari removed the four Phillips screws to free the large cooler (co-engineered by Framework, Cooler Master, and Noctua) to reveal the die shot everyone was waiting for. (You can see it around 8:38 in the embedded video).

Strix Halo hasn't changed; we've seen the die shots before. But here, we do see that the massive, seemingly custom cooling solution is a unit that covers both the processor and the RAM, which is soldered to the motherboard. Framework CEO Nirav Patel wrote in a blog that Framework "spent months working with AMD to explore ways around this but ultimately determined that it wasn’t technically feasible to land modular memory at high throughput with the 256-bit memory bus."

To cool the chip and RAM, Framework isn't using a standard desktop CPU cooler. It attaches to a 120 mm fan; you can pick from options from Cooler Master or Noctua, or pick your own. This is an engineering board, so it's possible changes will come to the final product, but it seems unlikely that there will be an option for someone to put Strix Halo under water cooling (but if you figure it out, get in touch).

Mokhtari was able to remove the board after opening the case and removing the front and side panels. He removed the fan from the cooler first and unplugged the board. He did, however, have to unscrew more of the case than expected, possibly due to this unit being an engineering sample. The Flex ATX PSU stayed in the case the whole time.

The Framework desktop is available for pre-order now and is set to release in Q3 of this year. The base model, with a Ryzen AI Max 385 and 32GB of RAM, starts at $1,099, while the top-end machine with a Ryzen AI Max+ 395 with 128GB of RAM begins at $1,999. Framework is only doing "DIY" editions here, so you'll have to get your own storage drive and bring your own operating system (the company is calling it "the easiest PC you'll ever build"). The mainboard on its own will be available from $799.

Framework is expanding its laptop lineup and getting into desktops. In an event announcing what the company is calling its "second-gen" products, it detailed three new computers: an updated Framework Laptop 13 with AMD Ryzen AI 300, a 4.5-liter Mini-ITX desktop powered by Ryzen AI Max, and a colorful, convertible Framework Laptop 12 designed with students in mind. Framework refers to the latter as a "defining product for us."

It's a large expansion of Framework's existing lineup, which until today consisted of two clamshell notebooks: the Framework Laptop 13 and 16.

Framework Desktop

Perhaps the biggest swing for Framework is its entry into the desktop market, which is already known for its customization options.

"We want to make this space as accessible as we possibly can by building a desktop that is simultaneously small and simple and incredibly powerful and customizable," Framework wrote in a blog post. "Everyone should have the opportunity to experience the culture around PCs and PC gaming first-hand."

The Framework Desktop is a 4.5L Mini-ITX machine using AMD's Ryzen AI Max "Strix Halo" chips with Radeon 8060S graphics. While this is a mobile chip, Framework says putting it in a desktop chassis gets it to 120W sustained power and 140W boost "while staying quiet and cool." Framework says this should allow 1440p gaming on intense titles, as well as workstation-class projects and local AI.

Much of Framework's desktop uses existing PC standards, including a mini-ITX motherboard, "with ATX headers, a PCIe x4 slot, and a broad set of rear I/O (including 2x USB4, 2x DisplayPort, HDMI, and 5Gbit Ethernet)," as well as a "semi-custom" 400W power supply in a Flex ATX form factor. The 120 mm CPU fans are part of a cooling system "co-developed" with Noctua and Cooler Master. You could swap out the fans for whatever you want, or even take the mainboard out and stick it in your own case. There are two PCIe NVMe M.2 2280 slots (for up to 16TB of storage) and Wi-Fi 7 via an RZ717 Wi-Fi module (which you can also find in plenty of AMD laptops).

One area where you won't be able to make changes is memory, as the LPDDR5x RAM is soldered to the board. Framework wrote in a blog post that "we spent months working with AMD to explore ways" to enable the 256GB/s memory bandwidth with modular memory, but ultimately decided this wasn't possible with the 256-bit bus. The company says that it is "being deliberate in making memory pricing more reasonable" because you can't upgrade it down the line.

There's a bunch of options for aesthetics, too, including black and translucent side panels, an RGB fan, and an optional carrying handle. Framework's front panel consists of 21 "color-customizable" tiles, and you can 3D print your own to make your own design. There are two ports up front for Framework's Expansion Cards (as seen on the Framework Laptops) to pick your own ports, which are on the bottom of the case.

AMD's Frank Azor took to the stage alongside Framework CEO Nirav Patel to detail performance claims with FSR and HyperRX to improve frame rates and rivalling "60-70 class discrete graphics" on an integrated GPU. Patel also pushed the Desktop as way to run AI locally, should you use the version with 128GB of RAM.

The base model, with a Ryzen AI Max 385 and 32GB of RAM, starts at $1,099, while the top-end machine with a Ryzen AI Max+ 395 with 128GB of RAM begins at $1,999. Framework is only doing "DIY" editions here, so you'll have to get your own storage drive and bring your own operating system (the company is calling it "the easiest PC you'll ever build"). The mainboard on its own will be available from $799. Pre-orders are open now, and Framework expects to ship sometime in Q3.

For the price, this is likely to be niche. It could make a good console competitor, but even a PS5 Pro starts at $699, even if it is far less customizable and repairable. We're looking forward to seeing just how much of Framework's DIY DNA has made its way into this desktop to differentiate itself from other desktop PCs.

Framework Laptop 12

Framework's other new project is the Laptop 12, a 12.2-inch convertible 2-in-1 aimed primarily at competing in the entry-level market popular among young students. A company blog post says that entry-level machines "tend to be janky, locked-down, disposable, underpowered, and frankly, boring machines."

The Laptop 12 is designed to bring the flexibility from the Laptop 12 but make it smaller, cheaper, and in more colors (with an optional stylus to match). These machines are made of ABS plastic molded in thermoplastic polyurethane, all around a metal frame. Framework says that it's "our easiest product ever to repair," but that more information on that will come closer to its launch in mid-2025. I'm really looking forward to this repair guide. It comes in five colorways: lavender, sage, gray, black, and bubblegum.

The laptop will come with 13th Gen Intel Core i3 and i5 processors, which aren't the latest, but better than entry-level junk. You'll get up to 48GB of RAM, 2TB of storage, and Wi-Fi 6E. It has a 1920 x 1200 touch screen that the company claims will surpass 400 nits of brightness.

There's no pricing information yet, and Framework says there's more to share on pricing and specs later in the year. Pre-orders will open in April ahead of the mid-year launch. I haven't seen a cheap, education-ready laptop with repairable options since the Asus Chromebook C202, so I'm excited to see those details when Framework is closer to launch.

Framework Laptop 13

The Framework Laptop 13 is getting a significant refresh with AMD Ryzen AI 300 Series. It doesn't look all that different on the outside, with a 13.5-inch design that largely resembles the one from way back in 2021.

But there are new features. Beyond the processors, the Framework Laptop 3 is getting bumped up to Wi-Fi 7 and is getting a new thermal system, a "next-generation" keyboard, and new colorways for the Expansion Cards and bezels (though I still don't know why you would want a bezel in anything other than black). The new keyboard has a Copilot key (on Windows 11 pre-built models - there is alternate artwork for DIY models where you might want to use Linux), and Framework says that new structures under wide keys like Shift and Spacebar should keep them from buzzing when your speakers have the volume up.

The new mainboards will have Ryzen 5 and Ryzen 7 options, but for the first time will also have a Ryzen 9 (the HX 370) with 12 CPU cores and 16 GPU cores that also support Copilot+, with 50 NPU TOPS.

That may be why Framework moved onto a redesigned thermal system, with a 10mm heatpipe and a Honeywell's PTM7958 TIM. The mainboard has two slots for DDR5-5600 RAM (up to 96GB) and a PCIe 4 M.2 2280 SSD for up to 8TB of storage.

Because the laptop is so similar to the original, this new mainboard can be placed into any of the older Framework Laptops dating back to the original. Any updates since, including a 61Whr battery, better speakers, and an optional high-res matte display are still available.

The new mainboard still has four expansion card slots. The back two work with USB 4 while the front two can use USB 3.2 plus DisplayPort, and Framework says you can have up to four display outputs simultaneously if you choose the right Expansion Cards.

The Expansion Cards are going to get some nostalgic color options, with a new set of translucent colors (though the laptop still only comes in silver).

The new Framework Laptop 13 with AMD Ryzen AI 300 starts at $899 for a DIY Edition without storage or an OS, and $1,099 for a pre-built model. If you're buying the mainboard to put in an old system, that's $449. (Framework is keeping the Ryzen 7040 systems around starting at $749). No word for now on any new Intel models.

Outlook

Framework's Second Gen event has shown us a version of the company that is emboldened by its mission. When the first Framework laptops launched, I questioned whether you could trust that it would keep delivering. It's done that, and now their laptops are the best ultrabook if repairability is your number one concern.

Now we're seeing the company expand like never before, into new categories in computing. I suspect some, like the desktop, will be niche hobbyist devices. I'm excited to see the company take on lower-tier devices that could be used in schools and with budget-conscious customers, but we need to see pricing.

AMD's Strix Halo reviews are no longer guarded by an embargo. So we're seeing a wave of new die shots that underline the impressive engineering that went into the design of these chips. Thanks to Kurnal, Asus' Tony Yu (at Bilibili), and Decap at X (formerly Twitter), we're getting our first look at the internals of Strix Halo, formally known as Ryzen AI Max.

Let's get one thing straight: Strix Halo isn't your average run-of-the-mill APU. It wields 16 cores and 32 threads based on the latest Zen 5 architecture, alongside a massive 40 CU (Compute Unit) iGPU (integrated GPU). The package itself is composed of two Zen 5 CCDs, which are a tad different from the ones you see on Granite Ridge (Ryzen 9000 CPUs). Surrounding the chip are up to eight memory ICs, enabling support for 128GB of unified memory.

Thanks to high-res images from Tony Yu, we see that AMD has trimmed the D2D (Die to Die) interfaces between the CCD and I/O die by roughly 2mm. Although the CCDs have been tweaked slightly at 67.07mm², they still support 3D V-Cache, evidenced by the TSV connections in the die shots. The rest is similar to what you'd typically find on a desktop Zen 5 processor. Each of the eight cores carries 8MB of dedicated L2 cache, with a shared 32MB L3 pool in between.

Beneath the two CCDs lies a large I/O die that hosts the uncore part of the chip. The bulk of the 307.58mm² die size is occupied by the 40 Compute Unit integrated GPU (20 WGPs) developed using RDNA 3.5.

Slapping on more cores is only beneficial if they can be fed with data. We've seen firsthand how the performance of Strix Point (with 16 CUs) was hampered due to memory bottlenecks. AMD has taken a two-pronged solution to counteract this issue, with fast LPDDR5X memory across a wide 256-bit interface (eight 32-bit controllers) to the sides of the chip, and 32MB of MALL (Memory Access at Last Level) or LLC (Last Level Cache) split evenly across both sides of the iGPU.

To the left of the iGPU lies the XDNA 2 NPU (Neural Processing Unit), enabling 50 TOPS of AI performance, claimed to outclass even the RTX 4090. Some locations have been left unmarked. Either way, the bottom of the I/O die features a single PCIe 4.0 x16 interface, alongside USB 4 (40 Gbps), USB 3.2 Gen 2, and USB 2.0 controllers, followed by a Display Engine and two Media Engines with support for H.264, H.265, and AV1 codecs. Based on the provided statistics, the full chip equals roughly 441.72mm² or allegedly 475mm². That's including what appears to be structural silicon.

The first laptops powered by Strix Halo should hit shelves later this month, with Asus' ROG Flow Z13 tipped for a February 25 launch. While OEMs have initially been hesitant to jump on board with these chips, adoption is likely to improve, despite a slow initial uptake.

AMD's beastly Strix Halo APUs might be closer than you expect, with reviews reportedly going live tomorrow per hardware leaker HXL at X (formerly Twitter). This rumor follows teasers from various content creators. Even Asus has announced a release date for its high-end ROG Flow Z13 laptops; slated for a February 25 launch. We're yet to hear more from HP regarding its Strix Halo-equipped laptops and mini-PCs, but they should follow shortly afterward.

AMD pulled the veil off the Ryzen AI Max+ series, codenamed Strix Halo, at CES last month. Claimed to rival Nvidia's RTX 4070 mobile, AMD has equipped the top-end config with a beefy 40 Compute Unit integrated GPU based on RDNA 3.5, hosted by the massive I/O die in the center. To the north are two Zen 5 CCDs with up to 16 cores / 32 threads and 80MB of total L2 and L3 cache, capable of pushing 5.1 GHz. The entire package is fed by 128GB of unified memory, where 96GB can be dedicated to the iGPU alone. AMD allows OEMs to configure these APUs between 45W and 120W, depending on the device.

Strix Halo reviews are reportedly embargoed until tomorrow. On-shelf inventory will be limited to two laptops and a mini-PC from HP and Asus, but wider adoption is expected as demand grows. In addition, Asus confirmed its "Illusion X 2025" laptops (the Chinese variant of the ROG Flow Z13) featuring AMD's Ryzen AI Max + chips, will launch on February 25 in a post on Weibo.

The 2-in-1 Flow Z13, functioning both as a laptop and a tablet, is configurable with the Ryzen AI MAX 395+, 128GB of LPDDR5X memory, a 13.4-inch 2.5K IPS display at 180 Hz with HDR support, and a 70Wh battery. All this power comes at a steep price of $2,699 which is comparable to high-end gaming laptops. Still, you get the advantage of a better battery life and more memory for inference. We'd love to see Strix Halo-powered mini-PCs but only HP's Z2 Mini G1a has been announced thus far, with no set release date.

Strix Halo might easily stand as one of AMD's most innovative and radical designs in recent memory. These APUs are designed to compete against Apple's M-series SoCs and mobile workstations outfitted with dedicated Nvidia graphics. Given that AMD's discrete GPUs never really took off in the laptop space, Strix Halo could signal a more APU-focused future, though it could be a while until we see a potential Strix Halo successor based on RDNA 4.

A leaked benchmark of the Radeon 8060S, powering the graphics engine of AMD's Ryzen AI 300 "Strix Halo" flagship is shown to match Nvidia's dedicated RTX 4060 mobile GPU in a leaked 3DMark Time Spy result. A Chinese user at Baidu (via HXL), shared a couple of screenshots with what appears to be the Ryzen AI Max+ 395 flexing its muscles, beating AMD's latest Radeon 890M iGPU by almost 3x. Since the tested sample is based on early engineering silicon, there is surely still some room for improvement. However, this leak should be viewed cautiously as the CPU OPN code and the integrated GPU don't align.

AMD extended its Ryzen AI 300 lineup with mainstream Krackan Point and flagship Strix Halo APUs last month at CES. Strix Halo, or Ryzen AI Max+ is a one-of-a-kind processor delivering (up to) 16 Zen 5 CPU cores bundled with 40 RDNA 3.5 Compute Units for workstation-grade laptops and high-end mini-PCs. Bear in mind, all this power is packaged on a single chip, featuring two CCDs and a massive I/O die beneath, bordered by (up to) 128GB of fast unified memory. For context, AMD's marketing material positions the Radeon 8060S (the subject of this article) as an equivalent to Nvidia's RTX 4070 laptop dGPU.

It's kind of pointless to compare laptops with different TGPs and thermal designs so it's best not to read too much into these results. For the sake of comparison, we'll look over the average Time Spy score of several relevant GPUs, obtained via 3DMark's score explorer feature. Another screenshot shows that the laptop or mini-PC in question features 128GB (16GBx8) of LPDDR5-8532 memory, with 96GB allocated to the iGPU. Both screenshots inaccurately label the iGPU as the Radeon 8050S, however, the OPN code reveals it's actually the Radeon 8060S with a 40 CU configuration. That's probably due to the silicon's premature nature.

In 3DMark's Time Spy benchmark, the Radeon 8060S scores 10,106 points, almost matching Nvidia's RTX 4060 laptop and AMD's own RX 7700S. Against the Radeon 890M seen on Strix Point, the 8060S lands ahead by a gigantic 2.7x but that was kind of expected given the large difference in shader counts. Still, it loses to the RTX 4070 by almost 20% which is disappointing but you should wait for independent reviews to see how these Strix Halo APUs perform in real-world scenarios.

You should see laptops and workstations equipped with these processors from partners across Q1 and Q2 this year, which is a rather vague timeframe. HP is readying the ZBook Ultra G1a workstation laptop and the HP Z2 Mini G1a mini-PC, while Asus has announced the ROG Flow Z13, with no definite release date provided for any system.

Late last year, our extensive testing showed that Intel's Lunar Lake Arc 140V integrated graphics outperformed AMD's top-end 890M integrated silicon (driver concerns and software aside) by an average of 7% over 27 tested games. Now that Intel has launched its Arrow Lake mobile parts, we've spent some time testing an Intel-provided MSI Prestige 16 AI Evo laptop running an Intel Core Ultra 9 285H, to see if Intel's integrated performance edge still holds. And in a broad sense, the answer is yes. The Arc 140T iGPU in the 285H (despite being based on the older Xe+ architecture rather than the Xe2 of Lunar Lake's 140V) maintains a slight edge over AMD's 880M, which we also recently re-tested in MSI's Prestige A16 AI+ laptop for comparison. But things aren't cut and dry once you drop below 1080p.

Also, AMD still leads on drivers and software for gaming, so you might have a better overall gaming experience if you opt for AMD. And really, while most games should broadly be playable at reduced settings and sub-1080p resolutions, those at all serious about on-the-go gaming should still strongly consider stepping up to a laptop with dedicated graphics, or perhaps something built around AMD's flagship Strix Halo, which the company says offers up to 1.4x the performance of the 140V iGPU in Intel's Core Ultra 9 288V.

Below, we'll look at the specs of our two MSI Prestige test laptops. Then we'll delve into our limited gaming test results across five titles and see how the two similarly configured systems stack up in gaming, as well as in CPU performance and battery life.

Note that we also managed to re-test Lunar Lake's 140V graphics in an Asus Zenbook S14 and AMD's flagship Strix Point iGPU, the 890M, inside a Zenbook S16. These tests were run right before the deadline for this article and are in different laptop chassis, with specs that vary more than the two MSI Prestige laptops. So we aren't going to comment on these results. They're just in our charts below for context.

Gaming Tests: Intel Core Ultra 285H / Arc 140T versus AMD Ryzen AI 9 365 / Radeon 880M

Given time constraints, we couldn't test a wide swathe of games so we focused on some titles that we use for laptop testing and ordered them in the charts below based on when the games were released. As you'll see, some less-demanding titles that are several years old run pretty well even at high settings. But newer games will force you to drop below 1080p and dial down the detail. We aren't going to bother reporting performance at the laptops' native 2560x1600 resolution, because neither integrated chip got anywhere close to playability while pushing that many pixels.

This 2018 franchise title gives Intel's GPU its first win, though the results are close. The 3 FPS lead the 140T holds over AMD's 880M at 1080p is at least reasonably definitive, but stepping down to 720p for more playable frame rates, there's a single-frame difference between the two competitors. These results are an average of three runs, but that's still within general benchmark variance.

Using a 2019-era title that wasn't particularly GPU-demanding when it was new, it's not exactly a surprise that both GPUs here flirt with 30fps at 1080p in Borderlands 3, and the Intel laptop has a nearly 2 FPS edge there. But interestingly, when dropping down to 720p, the AMD silicon pulls ahead by about 6.5 FPS.

Moving forward in terms of time and difficulty, we get to Cyberpunk 2077, where we had to turn off ray tracing entirely and drop the settings down to low to stay well out of single-digit territory. This test is a mixed bag for Intel's chip, where it pulls three FPS ahead at 1080p but again falls behind by 8 FPS at 720p – which is what you'll probably want to stick to with this title if you want anything approaching smooth results on modern integrated graphics.

The 2022 strategy title, Total War: Warhammer III is one title where Intel's 140T has a clear advantage. It beats the AMD 880M in our MSI test laptop by 2 FPS at 720p and an impressive 13 FPS at 1080p.

Last up is last year's hit, Black Myth: Wukong, which wasn't exactly smooth on either iGPU, even when we dropped our typical test settings down from Cinematic to the Medium preset. Intel's silicon turned in a 7 FPS edge over AMD at 1080p, but once again, dropping the resolution down a more playable 720p saw AMD's 880M pull to just a few FPS behind Intel's result. These results show that, while Intel's integrated graphics tend to perform better overall, things are going to vary between titles, especially when gaming below 1080p.

It also shows that modern demanding games are already pushing these integrated chips to their gaming limits. You'll likely be able to play most games without dedicated graphics for now at reduced settings and resolutions. But before long, future titles may leave you wishing you'd invested in a true gaming laptop with dedicated graphics.

A quick look at CPU performance and battery life

I was curious to see whether AMD's advantage in some of our gaming tests at 720p could be an indicator that the Ryzen AI 9 365 might have an edge over the Core Ultra 9 285H on the CPU side. But that doesn't seem to be the case, as you can see below, where we compared the two MSI Prestige laptops in a few of our productivity benchmarks.

Starting with the Geekbench 6 synthetic CPU benchmark, the AMD Ryzen 9 365-based laptop is competitive in the single-core test, falling just 1.59% behind the Core Ultra 9 285H in the MSI Prestige 16 AI Evo. But on the multi-core test, Intel's 16 cores and 16 threads managed to best AMD's 10 cores and 20 threads by 20.62%.

Shifting to our real-world Handbrake test, where we transcode a 4K video file to 1080p, Intel's lead is 24 seconds or just under 10%. That's a win for Intel and Arrow Lake. But keep in mind that we're comparing mid-2024 AMD silicon to an Intel chip that's arriving on the scene in February of 2025.

Battery Life

Our battery test is a mix of web browsing, video streaming, and graphical tests with the screen brightness set to 150 nits. So it's not an indication of gaming performance, but it's still a key consideration for any laptop.

Once again, the Intel-based machine is the clear leader here, offering up four extra hours of unplugged power in our test. But don't read much into those numbers as an indication of power efficiency. According to MSI's specs, the battery in the Intel-based Prestige (99.9 Whr) is 21.9% larger than the (82 Whr) battery in the AMD model. Still, the Intel laptop did get 36% longer run time, so it seems likely Intel does have some advantage in power efficiency, but we'll have to test more Arrow Lake laptops to say for sure. At this point, we're just happy to see that both options deliver over ten hours with their large 2560 x 1600 screens.

Final thoughts

After putting Intel's Core Ultra 9 285H chip and its 140T integrated graphics through our typical laptop tests compared to a similarly equipped MSI Prestige system running AMD's Ryzen AI 9 365 and its 880M graphics, we are left with more or less the same conclusions we had late last year in our much more extensive laptop graphics testing.

After many years of dominance by AMD, Intel's integrated graphics are now the best in terms of raw performance overall. And thanks to its work over the past few years on dedicated cards like the Arc B580, its driver and software stack have improved greatly. But we'd still give the edge to AMD on the software side, which isn't surprising given its GPU focus since acquiring ATI way back in 2006.

And regardless of how much integrated graphics have improved in the last few years, just because you can play most modern games these days on an iGPU doesn't necessarily mean everyone will want to. As our testing shows, you'll often have to drop below 1080p resolution, even on older games at high settings. And the more demanding games that have come out in the last few years will force you to stick to low settings and 720p to keep things running above even 30 FPS. So unless you're really counting on game developers downgrading the visuals due to rising costs, you should probably still choose a laptop with dedicated graphics if you want to play the best games of 2025 and beyond.

AMD has provided instructions on how to run DeepSeek’s R1 AI model on AI-accelerated Ryzen AI and Radeon products, making it easy for users to run the new chain-of-thought model on their PCs locally. Several LLMs using R1 are compatible with RX 7000 series desktop GPUs and select Ryzen CPUs with XDNA NPUs. However, they require the optional driver Adrenalin 25.1.1 to run.

The guide has everything AMD users need to get DeepSeek R1 running on their local (supported) machine. LM Studio has a one-click installer tailor-made for Ryzen AI, which is the method AMD users will use to install R1. AMD shows how the application needs to be tuned for its hardware, including a list of the maximum supported LLM parameters.

DeepSeek R1 allegedly has only recently been distilled into "highly capable" smaller models, small enough to run on consumer-based hardware. The DeepSeek-V3 model was initially trained on a cluster of 2,048 Nvidia H800 GPUs for context.

The maximum supported LLM parameters are based on memory capacity. The RX 7600 XT, 7700 XT, 7800 XT, 7900 GRE, and 7900 XT all support up to “DeepSeek-R1-Distill-Qwen-14B”. The flagship RX 7900 XTX supports up to “DeepSeek-R1-Distill-Qwen-32B”. The RX 7600, with its 8GB of VRAM, supports up to “DeepSeek-R1-Distill-Llama-8B”.

Similarly, Ryzen 8040 and 7040 series mobile APUs are equipped with 32GB of RAM, and the Ryzen AI HX 370 and 365 with 24GB and 32GB of RAM can support up to “DeepSeek-R1-Distill-Llama-14B”. The Ryzen AI Max+ 395 can support up to “DeepSeek-R1-Distill-Llama-70B”, but only in 128GB and 64GB memory capacities. The 32GB supports up to “DeepSeek-R1-Distill-Qwen-32B”.

DeepSeek’s new AI model has taken the world by storm, with its 11 times lower computing cost than leading-edge models. Two days ago, it was solely responsible for Nvidia’s record-breaking $589 billion market cap loss. The DeepSeek R1 model relies on extreme optimization levels to provide its 11X efficiency uplift, relying on Nvidia’s assembly-like Parallel Thread Execution (PTX) programming for most of the performance uplift.

Nvidia and AMD GPUs aren’t the only GPUs that can run R1; Huawei has already implemented DeepSeek support into its Ascend AI GPUs, enabling performant AI execution on homegrown Chinese hardware.

Intel processors have always powered NUCs. However, things changed when Asus took over the NUC line from the chipmaker in 2023. The manufacturer is seemingly experimenting with AMD chips after customs filing discovered by ComputerBase (machine translated) showed an entry for a “Personal Computer mini-PC (NUC14LNS) Asus AMD Ryzen AI Max-395/W 8060S, 64 BIT”.

If this is true, this will be the first NUC to feature an AMD chip, and it will come soon after Asus launches a family of five NUCs powered by an Intel processor. It won’t be its first device to use a Ryzen AI Max+ (Strix Halo) CPU, though, as it’s rumored to power the upcoming ROG Flow Z13 laptop (also listed in the customs document).

AMD also claims that the Ryzen AI Max 395+’s integrated graphics, the Radeon 8060S, can beat the RTX 4070 laptop GPU by up to 68%—a massive claim by the company but is helped by the fact that it has 40 RDNA 3.5 graphics cores. If true, this will disadvantage Asus NUCs if it only sticks with Intel processors, especially as some smaller manufacturers like GMK have already hinted at releasing a mini-PC powered by this processor.

The mini-PC customs description says it’s “for BIS testing,” which likely means the Bureau of Indian Standards. If that’s the case, then Asus likely already built and tested the AMD-powered NUC within the company. Indian authorities are now testing it to ensure it conforms to local standards. This might also mean that this device is destined for launch in India, although there is no official word from Asus yet about whether this is the case or if it will have a global launch.

Given that Asus is investing in product certification, there’s a good chance this is a genuine product. The company is just holding off on its announcement until the authorities give it the green light. We also hope that it will make its way to our shores. That way, mini-PC fans will have more options, and we could enjoy the massive performance that its integrated GPU delivers.

At CES, AMD unveiled its Ryzen AI 300 flagship lineup codenamed Strix Halo with a beastly configuration of 16 cores, 40 Compute Units, and 128GB of unified memory. New first-party benchmarks assert that the top-end Ryzen AI Max+ 395 - that's a mouthful - is up to 68% faster than Nvidia's RTX 4070 mobile graphics in gaming. The slides also showcase how Strix Halo holds up versus Apple's M4-series chips in productivity and even dare to mention the desktop RTX 4090.

Strix Halo is split into three dies with two CCDs offering eight cores each and a large I/O die housing the iGPU (Integrated GPU) and the NPU. The highest-end Ryzen AI Max+ 395 packs 16 cores and 32 central processor threads based on the Zen 5 architecture, accompanied by the Radeon 8060S iGPU with 40 RDNA 3.5 Compute Units. All the AI Max chips have a base TDP of 55W, and can be configured from 45W to 120W. Surrounding the chip is up to 128GB of unified memory, which can be allocated in varying pools (up to 96GB) to the iGPU alone.

With the specs out of the way, let's talk performance. The Nvidia comparison system is the Asus ROG Flow Z13 with the i9-13900H and the RTX 4070 mobile (65W TGP). Across a suite of 17 games at 1080p high settings, Strix Halo handily bests its Nvidia equivalent, with an up to 68% lead in Borderlands 3, per AMD. Being first-party benchmarks, we shouldn't take them at face value. It is also prudent to mention that the RTX 4070 mobile can be configured up to 115W, but a 65W variant features in this comparison, due to the Asus ROG Flow Z13 design.

In productivity tests, Blender sees the 395+ blaze past the Core Ultra 9 288V (Lunar Lake) by 302% but that isn't an apples-to-apples comparison since the latter is a 17/30W chip. Versus Apple's M4 Pro, AMD touts an 86% lead in V-Ray, dropping to 12% in Cinebench R24. Moving over to AI performance, the XDNA 2 NPU enables the Ryzen AI MAX series to dish out 50 TOPS of AI performance, exceeding the Lunar Lake and Snapdragon X Elite series by two TOPS. AMD also claims its Strix Halo APUs can deliver 2.2x more tokens per second than the RTX 4090 when running the Llama 70B LLM (Large Language Model) at 1/6th the TDP (75W).

We highly suggest you wait for independent testing before jumping the gun, like perhaps pre-ordering a Strix Halo device. These laptops are likely to target enthusiasts as suitable alternatives to dedicated GPU options. With Strix Point laptops still costing over $1,000, pricing will play a key role in determining if Strix Halo can displace dedicated mid-ranged laptop GPUs like the RTX 4070 can be found in.

An engineer from AMD's Nod.ai acquisition published a feedback post on AMD's GitHub ROCm support page, seeking guidance on additional GPU support the ROCm community would like to be added to AMD's list of supported ROCm devices.

The engineer put together a poll consisting of various GPU lineups and their associated architectures: the RX 7000 series (RDNA 3), the RX 6000 series (RDNA 2), Phoenix mobile APUs (RDNA 3), Strix / Strix Halo mobile APUs (RDNA 3.5), and RDNA 1 (i.e., the RX 5000 series). The last category was labeled as GCN and prior architectures.

So far, the RX 6000 series RDNA 2 GPUs have won the leading category, followed by Strix / Strix Halo and the RX 7000 series. RDNA 1 has received the fewest votes, while "GCN and prior architectures" take fourth place.

The comments section also focused on what GPUs AMD should support with ROCm. The highest-upvoted comment recommended supporting any GPU with 16GB or more VRAM, including integrated GPUs that can occupy 16GB or more memory through system DRAM. The second-most highly upvoted comment recommended supporting ROCm for all relatively recent GPUs from the last five to six years.

The AMD ROCm open-source software platform currently supports all modern AMD Instinct GPUs, particularly those based on iterations of AMD's CDNA architecture. Likewise, all of AMD's modern workstation graphics cards support ROCm on Linux.

Where support gets touchy is Windows support and support for Radeon-based products. The only Radeon GPUs supported on Linux are the RX 7900 series graphics cards and the Radeon VII. On Windows, most of the RX 6000 series and all of the RX 7000 series desktop GPUs support ROCm, but the RX 6700 - RX 6750 XT don't support the HIP SDK, only Runtime. Similarly, the Radeon Pro W7900, down to the W6600, is supported, but the W6600 lacks HIP SDK support, and the Pro VII and W5500 aren't supported on Windows. No mobile GPUs support ROCm, according to AMD's device support list.

ROCm is an open-source software platform developed by AMD to aid programmers in running HPC and AI applications on AMD GPUs. It is essentially AMD's counterpart to Nvidia's CUDA architecture, but unlike Nvidia, it does not support all AMD GPUs dating back to the mid-2000s.

AMD launched a slew of new CPU products at CES, all of which were covered extensively by Tom's Hardware. Here's your first look at all the CPUs intact and delidded, showcasing the beautiful internals and allowing us to appreciate the meticulous engineering behind each chip.

Fire Range

Fire Range, which falls under AMD's umbrella of Ryzen 9000HX processors, employs desktop-grade silicon geared at workstations, with V-Cache-equipped variants for gamers as well. The CPU boasts two Zen 5-based CCDs (Core Complex Dies) with an IOD (I/O Die) beneath, mirroring desktop Ryzen 9000 processors.

Strix Halo

Next is the Ryzen AI Max 300 series, dubbed "Strix Halo." This massive APU carries a unique design philosophy, featuring two CCD-esque chiplets alongside a massive I/O Die that houses an up-to-40 Compute Unit iGPU (Integrated GPU) and the NPU.

Krackan Point and Hawk Point Refresh

The Ryzen AI 300 (Krackan Point) and Ryzen 200 (Hawk Point Refresh) models are new to AMD's APU portfolio and set to power affordable yet capable laptops. Krackan Point is a dialed-down version of Strix with upwards of eight Zen 5 cores and up to eight Compute Units based on RDNA 3.5. Hawk Point Refresh follows along, coupling Zen 4 with RDNA 3. Both are monolithic, making it a bit difficult to distinguish one from the other. However, based on visual cues, the wider chip is likely Krackan, while the smaller chip is likely Hawk Point Refresh.

Ryzen 9 9950X3D and 9900X3D

Following in the footsteps of the world's fastest gaming processor, the Ryzen 7 9800X3D, AMD's new Ryzen 9 9900X3D and Ryzen 9 9950X3D are optimized for games and creator workloads. Boasting a total of 16 cores, AMD touts its Ryzen 9 9950X3D as 20% faster in gaming and 10% faster in productivity than Intel's Core Ultra 9 285K. The Ryzen 9 9900X3D and 9950X3D share similar internals with two CCDs; one featuring the extra SRAM, adjoined with an IOD beneath.

Going from top to bottom, Fire Range is slated for launch in H1 2025, with Strix Halo debuting across Q1 and Q2 this year, followed by Krackan Point in Q1 2025, and Hawk Point Refresh in Q2 2025. Lastly, the 16-core Ryzen 9 9900X3D and 12-core Ryzen 9 9950X3D are planned for March 2025.

AMD announced its ‘Strix Halo’ Ryzen AI Max series laptop processors here at CES 2025 in Las Vegas, and by any definition, the new APUs look to be absolute monsters for the enthusiast thin-and-light gaming and AI workstation laptop markets, featuring what AMD bills as the fastest integrated graphics available in the Windows ecosystem courtesy of a new disruptive integrated memory architecture that unlocks new capabilities and a beastly 40-core RDNA 3.5 integrated GPU. 

In fact, AMD says the AI Max chips deliver up to 1.4X faster gaming performance than Intel’s flagship ‘Lunar Lake’ Core Ultra 9 288V flagship, and up to 84% faster rendering performance than the Apple MacBook M4 Pro. AMD also says the chip can deliver up to an incredible 2.2X more performance in AI workloads than the discreet desktop Nvidia RTX 4090 GPU, but at an 87% lower TDP.

If AMD keeps with tradition, which we fully expect, we will see these monstrous APU chips come to desktop PCs in the future.

Ryzen AI Max Specifications

The fire-breathing 120W Zen 5-powered flagship Ryzen AI Max+ 395 comes packing 16 CPU cores and 32 threads paired with 40 RDNA 3.5 (Radeon 8060S) integrated graphics cores (CUs), but perhaps more importantly, it supports up to 128GB of memory that is shared among the CPU, GPU, and XDNA 2 NPU AI engines. The memory can also be carved up to a distinct pool dedicated to the GPU only, thus delivering an astounding 256 GB/s of memory throughput that unlocks incredible performance in memory capacity-constrained AI workloads (details below). AMD says this delivers groundbreaking capabilities for thin-and-light laptops and mini workstations, particularly in AI workloads. The company also shared plenty of gaming and content creation benchmarks.

As designated by the ‘+,’ the Ryzen AI Max+ 395 serves as the flagship of the new series, but it’s accompanied by three other processors in varying configurations ranging from a 6-core 12-thread / 16 GPU core Ryzen AI Max Pro 380 to the 12-core 24-thread / 32 GPU core Ryzen AI Max 390. The Ryzen AI Max 390 and 385 are armed with the Radeon 8050S graphics engine. AMD also has an unlisted 16-core 16-CU model for OEMs.

All of the AI Max chips have a 55W base TDP, but also a configurable TDP that ranges from 45 to 120W to unleash more horsepower in designs that can handle the thermal output. The top three models will come in both standard and Pro series models, with the latter

including a RAS feature set for professional users. AMD plans to release these processors across Q1 and Q2 of this year.

As you can see above, the chip has a large central I/O die that houses the GPU and NPU, while the two smaller dies contain the CPU cores. The chip connects to standard memory, but you can use the Radeon Adrenaline software to make a custom memory allocation dedicated to the GPU only. 

For instance, if you have 128GB of total system memory, up to 96GB can be allocated to the GPU alone, with the remaining 32GB dedicated to the CPU. However, the GPU can still read from the entire 128 GB memory, thus eliminating costly memory copies via its unified coherent memory architecture. However, it can only write to its directly allocated 96GB pool. The combination of a shared memory pool and a dedicated memory pool enables faster performance in a range of workloads by leveraging the memory as a slower VRAM of sorts, including disruptive performance in AI workloads that we’ll outline below. But first, gaming and content creation performance.

Ryzen AI Max+ gaming and content creation benchmarks

The Strix Halo chips address both the enthusiast gaming and workstation markets, so performance in gaming is key. AMD claims the Ryzen AI Max+ is 1.4X faster than Intel’s flagship Lunar Lake Core 9 288V, but the company didn’t share real-world gaming benchmarks. Instead, AMD shared synthetic 3DMark tests that often don’t correlate well to real-world gaming performance. We won’t know more about real-world gaming performance until Ryzen AI Max hits the market, and as always, you should approach vendor provided benchmarks with the appropriate skepticism (we’ve included AMD’s test notes at the end of the article).

AMD also included numerous rendering benchmarks of its 16-core flagship against the 12-core Apple MacBook M4 Pro, claiming an up to 86% advantage in a v-ray workload. Naturally, the 14-core M4 Pro, also included in the benchmarks, is more competitive, but AMD still holds a stout lead in the Blender, Corona, and v-ray selection of benchmarks. However, the Ryzen AI Max+ isn’t as performant in the multi-threaded Cinebench 2024 test, beating the 12-core M4 Pro by a scant 2%, and trailing the 14-core M4 Pro by 3%.

AMD also took another swipe at Intel’s Lunar Lake Core Ultra 288V flagship, showing massive 3D rendering performance advantages of 2.6X faster on average, with advantages ranging from 340% to 402% in this selection of benchmarks. The 288V is geared for lower-power laptops, so comparisons against Intel’s newly-launched Core Ultra 200H and 200HX series, the latter with a dGPU, will be interesting.

AMD used a reference motherboard for its tests (a test board that typically has unlimited cooling capacity), so these results won’t reflect the performance you can expect in thermally constrained laptop. In contrast, the Apple and Intel benchmarks were generated with shipping laptops. AMD lists the tested TDP at a base of 55W, but power consumption can stretch much higher during heavy loads, especially with beefy cooling on a reference motherboard. This can lead to lopsided comparisons – for instance, the 288V has a 37W max TDP. All of the Ryzen AI Max configs were tested with only 32GB of RAM for these benchmarks.

Ryzen AI performance and systems

AMD also shared some rather impressive results showing a Llama 70B Nemotron LLM AI model running on both the Ryzen AI Max+ 395 with 128GB of total system RAM (32GB for the CPU, 96GB allocated to the GPU) and a desktop Nvidia GeForce RTX 4090 with 24GB of VRAM (details of the setups in the slide below). AMD says the AI Max+ 395 delivers up to 2.2X the tokens/second performance of the desktop RTX 4090 card, but the company didn’t share time-to-first-token benchmarks.

Perhaps more importantly, AMD claims to do this at an 87% lower TDP than the 450W RTX 4090, with the AI Max+ running at a mere 55W. That implies that systems built on this platform will have exceptional power efficiency metrics in AI workloads. The Ryzen AI Max models all have an integrated XDNA 2 NPU AI engine capable of delivering up to 50 TOPS of performance, but AMD didn’t share benchmarks for the NPU.

Speaking of systems, AMD’s partners will have systems built around the AI Max chips coming to market this year. HP has a ZBook Ultra G1 compact workstation and an HP Z2 mini G1a workstation, but the release date for those systems is undefined. Additionally, Asus has its gaming tablet with a detachable keyboard, the ROG Flow Z13, coming to market in Q1 of this year. That implies a hefty amount of gaming performance packed into a very slim form factor. Here’s hoping we can find these at retail, as AMD’s flagship gaming chips have often suffered from poor availability.

The upcoming Ryzen AI MAX 395+ from AMD's Ryzen 300 or Zen 5 family of APUs has emerged at Geekbench with performance that's largely identical to the desktop Ryzen 9 7950X processor — at least in short bursts. The 395+ is seen powering the upcoming ROG Flow Z13 laptop from Asus, expected to debut at CES next month. Against the last-gen Ryzen 9 7945HX, AMD has managed to extract a massive 22% performance uplift with Strix Halo.

The enthusiast laptop market will soon be greeted by AMD's flagship APUs codenamed "Strix Halo" under the Ryzen AI MAX Pro lineup. Likely to arrive at CES, these APUs reportedly serve as a one-stop solution for workstation laptops uniting Zen 5 cores with a powerful RDNA 3.5 iGPU (Integrated GPU) on a single chip.

Rumor has it that the top-end Radeon 8060S may be configured with up to 40 CUs (Compute Units) with support for fast LPDDR5X-8000 memory. Strix Halo is a one-of-a-kind product and should not be confused with the flagship-grade Fire Range series, succeeding Dragon Range (Ryzen 7045HX).

The test bench features the Asus ROG Flow Z13 "GZ302EA," listings for which surfaced a few days ago at a Belarusian retailer. Geekbench reports the Ryzen AI MAX 395+ with 16 cores and 32 threads based on AMD's Zen 5 architecture. The CPU offers a base frequency of 3 GHz, going as high as 5.1 GHz. Much like 16-core desktop chips from AMD, the 395+ is equipped with 64MB of L3 cache and 16MB of L2 cache.

Impressive scores of 2928 points and 19484 points in the single-core and multi-core categories, respectively, land the Ryzen AI MAX 395+ largely on par with the Ryzen 9 7950X. The limited clock speeds prevent it from breaching the 3,000 barrier when it comes to single-core performance. However, multi-core sees it pass with flying colors, beating the Ryzen 9 7945HX by 22%. Notably, the relatively short-duration Geekbench benchmark leads to ideal scores for the chip, as the chip will heat up and likely slow as it runs longer-duration benchmarks.

Realistically, Strix Halo has no direct equivalent from Intel's side and is more targeted as a competitor to Apple's M-series chips. It wouldn't make sense to compare Arrow Lake versus Strix Halo as the latter employs a much larger integrated GPU. Judging from the leaks, the Radeon 8060S powering the graphical end of the 395+ offers 40 CUs, which is 25% more than the Radeon RX 7600. If priced accordingly, this configuration could give upcoming RTX 5060 laptops a run for their money.

AMD is expected to reveal Strix Halo alongside Krackan Point and the RDNA 4 family at CES next month.

A listing for an Asus ROG Flow Z13 GZ302EA-RU003W has appeared on the Belarusian eCommerce site algo.by, and its description says that it’s powered by an AMD Strix Halo (2024) processor. This makes it one of the first laptops to arrive equipped with an AMD Strix Halo chip—a processor expected to debut in a few weeks’ time at CES 2025—with hardware detective @Olrak29_ sharing the link to the listing on X. Unfortunately, the listing is marked as ‘Out of stock’ and the retailer did not give an indication of its price.

The Strix Halo processors are expected to be AMD’s top-of-the-line mobile chips and will be armed with an RDNA 3.5 integrated GPU under the Radeon 8000S series. In fact, we’ve already seen a leaked Geekbench result for an AMD Ryzen AI Max+ Pro 395 with Radeon 8060S; while the initial numbers are a bit underwhelming, these results are expected to improve as the manufacturer makes further optimizations.

The results show that this Strix Point processor will have 16 cores and 32 threads, which is four cores and eight threads more than the Ryzen AI 9 HX 375 chip. It has a higher base clock of 3.00 GHz, although its maximum frequency is slightly lower at just 4.4 GHz. Aside from that, it has a much larger L3 cache at 32MB x 2 (versus the latter’s 24MB).

Nevertheless, what makes these upcoming chips more interesting is the integrated GPU. Although it doesn’t quite have an RDNA 4 integrated GPU yet (it still uses RDNA 3.5), leaks say it will have 40 compute units, which is eight more than the desktop RX 7600 and RX 7600 XT GPUs can offer. This could potentially give it great gaming performance at 1080p and 1440p gaming, meaning laptop gamers could possibly get good graphics performance without having to rely on a more power-hungry discrete GPU.

We expect Strix Halo-powered laptops to arrive in the early part of 2025, so we can’t wait to get our hands on one so we can test it against our picks for the best gaming laptops in 2024. Hopefully, AMD has taken steps to rectify its issues with laptop OEMs complaining about poor support, chip supply, and communications from the company.

Preliminary listings at Australian and Greek retailers (credit: Huang514613 at X) reveal that Asus's next wave of laptop offerings—powered by AMD's Krackan Point and Intel's Arrow Lake-H processors—is on the horizon.

Many rumors suggest that AMD and Intel plan to unveil their next-generation mobile offerings next month at CES 2025. AMD is expected to showcase its flagship Strix Halo (Ryzen AI Max Pro 300) and budget Krackan Point (Ryzen AI 300) APUs. At the same time, Intel will introduce Arrow Lake in HX, H, and U configurations, each geared toward different power envelopes and price points.

The laptops range from Asus' budget Vivobook to its premier Zenbook lineup, featuring the Core Ultra 9 285H, Core Ultra 5 255H, Core 7 250H, AMD Hawk Point (Ryzen 7 250), and the Krackan Point-based Ryzen AI 7 350, which we covered a few days ago.

To distinguish among the different models, on Intel's side of the ring, each SKU with the "Ultra" moniker uses Arrow Lake silicon - the Core Ultra 5 255H and Core Ultra 9 285H, while the others are just rebadged Alder Lake/Raptor Lake chips - the Core 7 250H. Adding to the mix, not all Core Ultra 200 chips are based on Arrow Lake; leaks suggest that Intel's Core Ultra 200U offerings will be a slightly enhanced version of Meteor Lake on Intel 3.

Arrow Lake uses a modified form of Xe-LP, termed Xe-LP+ or Alchemist+, and features XMX cores. This will allow these processors to use Intel's upcoming XeSS Frame Generation technology, which is not supported on Meteor Lake due to the lack of XMX cores. It is no Battlemage, but most Arrow Lake-H/HX laptops will be paired with a discrete GPU—likely from Nvidia's upcoming RTX 50 family.

On the other hand, Hawk Point (Ryzen 8040) will be refreshed next year as Ryzen AI 200 to better align with AMD's new naming scheme. It is still based on Zen 4 and RDNA 3. Krackan Point is a budget alternative to Strix Point, maintaining all the bells and whistles of Zen 5 and RDNA 3.5. However, the GPU features eight CUs (Compute Units)—half that of Strix Point.

Krackan is said to compete directly with Intel's Lunar Lake chips but could fall behind in efficiency and graphics performance. At the same time, AMD is also readying its Fire Range chips —using desktop silicon —to rival Intel's Arrow Lake-HX processors.

As we inch closer to CES, laptop manufacturers are firing all cylinders to get their offerings ready beforehand - as indicated by this leak. Expect these laptops to retail a few weeks after the announcement, possibly in late January or early February.

AMD (per VideoCardz) is expected to reveal its budget Krackan Point and flagship Strix Halo offerings next month at CES 2025. In addition to a previous leak, at least in the CPU department, another Krackan Point APU has surfaced at Geekbench, competing against Intel's Lunar Lake Core Ultra 7 258V.

The leaked benchmark features a Krackan Point Engineering Sample with the OPN Code "100-000000713-40_Y" - likely the Ryzen AI 7 350. The APU was spotted on Acer's upcoming Swift Go 16 laptop variant, which is set to arrive next year. The CPU packs eight cores, split across two clusters of four Zen 5 and four Zen 5c cores alongside a base and max frequency of 2 GHz and 5.05 GHz. Like all octa-core APUs from AMD, the Ryzen AI 7 350 features 16MB of L3 cache and eight MB of L2 cache.

The Ryzen AI 7 350's single-core and multi-core scores of 2677 and 11742 place it in Lunar Lake territory, edging out the Core Ultra 7 258V. It does lose to Strix Point by some margin, but the latter has 50% more cores and threads and a larger power envelope. Compared to Hawk Point, the Ryzen AI 7 350 is 5% faster in single-core performance but surprisingly loses in multi-core, possibly due to early silicon/software.

Krackan Point (shorthand: KRK) is a cost-effective alternative to AMD's Strix Point APUs under the Ryzen AI 300 moniker. These APUs are rumored to employ up to eight cores in a hybrid configuration (four Zen 5 + four Zen 5c) with up to 16MB of L3 cache. The iGPU (Integrated GPU) solution drops to just eight RDNA 3.5 Compute Units (Radeon 880M) - down from the 16 CUs on Strix Point. There has been some discussion regarding a possible dual-CCX layout with separate ring buses and L3 caches for each core type, but that has yet to be seen.

Competition against Lunar Lake will be stiff, but Krackan Point will generally target the mainstream segment, with rumors that Copilot+ laptops will start at $799. Per our exhaustive testing, the Arc 140V (Lunar Lake) lands faster than the Radeon 890M (Strix Point); expect this delta to widen with Krackan Point.

With less than one month left to CES 2025, leaked benchmarks have started to surface starring AMD's flagship Strix Halo APUs. The Ryzen AI MAX+ Pro 395 - now that's a mouthful - has emerged on Geekbench packing 16 CPU cores, 64GB of RAM, and a massive 40 CU (Compute Units) Radeon RX 8600S iGPU (Integrated GPU). Despite the apparent low score - set to improve with further optimizations - the real juicy details lie in the specifications.

For the uninitiated, Strix Halo or the Ryzen AI MAX 300 APUs are said to be AMD's top-of-the-line offering for workstations and laptops next year. The test bench utilizes the "AMD MAPLE-STXH" reference board, designed for the FP11 socket. Likewise, Ryzen AI MAX+ Pro 395—which we'll call Ryzen AI 395 for simplicity's sake—is paired with 64GB of RAM, going as high as 128GB per shipping manifests. The performance isn't anything to write home about at this stage, with the Vulkan score on par with an RTX 2060 resulting from early silicon.

In line with previous rumors, the Ryzen AI 395 features 16 cores based on the Zen 5 microarchitecture and 32 threads. The suspected dual-CCD design lands it 32MBx2 (64MB) of L3 cache alongside 16MB of L2 cache (1MB per core). Geekbench lists the maximum clock speeds at 4.4 GHz, which will likely improve with time.

Strix Halo's graphics solution will fall under AMD's Radeon 8000S umbrella of iGPUs: the Radeon 8060S and Radeon 8050S are based on RDNA 3.5. On an architectural level, the iGPU likely resides in the large SoC die with upwards of 32MB of MALL or Infinity Cache to avoid memory bottlenecks. Geekbench doesn't explicitly list the CU count, but previous leaks hint towards a 40 CU configuration for all Radeon 8060S models—25% more than the RX 7600.

Strix Halo is designed to compete against Apple's M-series silicon and Nvidia's dedicated GPUs in the laptop segment. It may not rival the RTX 4090 laptop GPU, but it could give mid-ranged offerings like the RTX 4070 laptop a run for its money.

Pricing is a valid concern since Strix Point - the Ryzen AI 300 family - is limited in availability and costs a pretty penny if you want the best. Nonetheless, we expect AMD to unveil its Strix Halo lineup of APUs at CES 2025 in addition to Krackan Point and the Radeon RX 8000 series.

The Surface and Surface Pro lineups will get Lunar Lake models in 2025, a report from Windows Central claims.

Citing his sources, Windows Central senior editor Zac Bowden claims Microsoft is set to launch a slew of new Surface devices, including Lunar Lake-powered Surface and Surface Pro laptops, a new Surface Laptop Studio model sporting a Snapdraon X Plus, and a new Surface device with an 11-inch screen.

The Lunar Lake variant of the latest Surface Laptop 7 has been anticipated since October when a model sporting the Core Ultra 7 268V was spotted on Chinese e-tailer Goofish. Intel's Lunar Lake will also make its way to the Surface Pro 11, the report alleges. These Lunar Lake models will be pretty much the same as their Snapdragon counterparts, except for sporting an anti-reflective display, a feature that the previous Surface Laptop 6 and Surface Pro 10 had. Bowden believes that the Lunar Lake Surface Laptop could even sport 5G connectivity.

These new Surface devices would be Microsoft’s first Copilot+ PCs powered by Intel CPUs and will arrive in the first quarter of 2025, the report says. If true, then that may mean Microsoft plans to reveal these Lunar Lake laptops at CES in January.

Bowden says Microsoft is also working on a Copilot+ version of the Surface Laptop Studio. Although the Surface Laptop Studio 2 only came out last year, it’s in need of an upgrade because it uses one of Intel’s 13th Gen CPUs based on Raptor Lake, and those were already out of date by then. Raptor Lake has already been succeeded thrice over, by Raptor Lake Refresh, Meteor Lake, and now Lunar Lake.

It’s unclear to Bowden what processor Microsoft will choose for the new Surface Laptop Studio, but since it’s a Copilot+ PC, that means there are not a ton of options. It may use a Lunar Lake-based Core Ultra 200 CPU, but AMD’s Ryzen AI 300 APUs powered by Strix Point are also on the table, as are Qualcomm’s Snapdragon X processors. Depending on how far out this Surface Laptop Studio is, Microsoft may even choose processors set to launch in 2025 like Panther Lake.

There’s also a new Surface device in the works, with a screen size of 11 inches; it’s apparently similar to the Surface Laptop Go but more premium. According to the report, Qualcomm has scored this one, as Microsoft is using the Snapdragon X Plus for the upcoming Surface device. The cut-down version of Qualcomm’s high-end Snapdragon X Elite chip, the X Plus ranges from 8 to 10 cores rather than 12 and is equipped with a much weaker integrated GPU, but still has the same NPU capable of 45 TOPS and is more optimal for devices with reduced thermal headroom.

The new Surface Laptop Studio and the upcoming 11-inch Surface are expected to land in 2025 as well, but Bowden says they could arrive as late as fall.

A new hexa-core processor under AMD's Krackan Point (KRK) lineup of APUs has emerged at Geekbench - thanks to Olrak at X. The CPU wields a rather unconventional 3+3 hybrid configuration - taking a different approach than AMD did with its Phoenix 2 APUs.

Krackan Point is AMD's cheaper alternative to Strix Point - packing upwards of eight cores and an eight Compute Unit (CU) RDNA 3.5-based iGPU (Integrated GPU). Krackan Point utilizes a hybrid-core design to save valuable die space, featuring up to four Zen 5 and four Zen 5c cores. The six-core offering has an OPN code of "100-000001600-40_Y," corresponding to a Ryzen 5 APU per previously leaked shipping manifests. The benchmark, being an AI test, isn't that exciting once you consider that Krackan employs the same 50 TOPS XDNA 2 NPU as Strix Point.

Going over the listed specs, the budget Ryzen 5 300 APU packs three Zen 5 and three Zen 5c cores alongside 12 threads - quite different from the Phoenix 2-based Ryzen 5 7545U - which had two Zen 4 and four Zen 4c cores. Moreover, the APU was being tested in an Asus validation platform and had a relatively low base frequency of 2 GHz - but that's likely a result of early silicon.

AMD reintroduced a dual-CCX design with Strix Point - splitting up the Zen 5 and Zen 5c cores across different rings - with separate L3 caches for both core types. Interestingly, for the Krackan Point APU in question, Geekbench divides the core count into two clusters. However, Olrak suggests that AMD might package all six cores together - connected through a unified ring bus - sharing a coherent L3 cache. To chip in, the Ryzen 5 7545U (Phoenix 2) also employed a single CCX, but the final decision rests on AMD.

Rest assured, laptops powered by AMD's Krackan Point offerings are set to arrive in early 2025 - starting at $799 if we go by leaks. Overall, CES 2025 will be jampacked with next-gen launches from all major players. Expect to see AMD unveil Krackan Point alongside Strix Halo APUs, Fire Range CPUs, and Radeon RX 8000 "RDNA 4" GPUs.