AMD's Ryzen AI 400 series includes the first Copilot+ desktop CPU — Team Red refreshes Zen 5 APUs and Strix Halo

AMD’s upcoming Ryzen AI 400 series of laptop APUs, dubbed ‘Gorgon Point,’ will arrive in the first quarter of 2026, the company announced at CES. The refreshed lineup still sports up to 12 Zen 5 CPU cores and up to 16 RDNA 3.5 GPU cores, but comes with increased frequency and memory support, as well as a souped-up XDNA 2 NPU. Alongside the laptop lineup, AMD announced the first Copilot+ desktop processor, as well as two new entries in its Strix Halo lineup.

There are seven chips in the laptop range, with the highest-end SKU being the Ryzen AI 9 HX 475. That chip comes with the full Gorgon Point die, including 12 Zen 5 cores, 24 threads, 36MB of combined L2 and L3 cache, 60 NPU TOPS, 16 RDNA 3.5 GPU cores that can boost up to 3.1 GHz, and support for memory speeds up to 8,533 MT/s. Ryzen AI 300 chips, dubbed ‘Strix Point,’ topped out at LPDDR5X-8000.

You can see the full mobile lineup from AMD below (we don't have much information on that desktop chip). Although AMD has bumped memory speeds on some of the range, the bottom three SKUs still top out at 8,000 MT/s.

AMD uses the same mixture of Zen 5 and Zen 5c cores on its Gorgan Point SKUs as it did on Strix Point SKUs across most of the lineup. The Zen 5c cores are roughly 25% smaller than a standard Zen 5 core, and they deliver the same performance with like-for-like clock speeds. However, it’s a space-optimized design. They’re clocked lower than standard Zen 5 cores, and peak frequency is designed around the few full-fat Zen 5 cores available on Gorgan Point.

All of the Ryzen AI 9 models, as well as the Ryzen AI 7 450, come with four Zen 5 cores, with the remaining core count comprised of Zen 5c. The Ryzen AI 7 445 and Ryzen AI 4 435, both six-core/12-thread chips, have two Zen 5 and four Zen 5c cores, while the four-core Ryzen AI 5 430 comes with a single full Zen 5 core. Read our technical deep-dive on the Zen 5 architecture to learn more about the differences between Zen 5 and Zen 5c.

AMD uses the same process as it did with Strix Point, which is TSMC’s N4X node. It’s an enhanced version of TSMC’s 5nm process with higher density.

Although AMD is light on gen-on-gen benchmarks, Gorgon Point looks like a minor refresh. The flagship Ryzen AI 9 HX 475 features just a minor 100MHz bump in boost clock speed alongside increased memory speed support. We’ve seen similar lateral spec moves from AMD with its past mobile releases, such as when the company moved from the Ryzen 9 7940HS to the Ryzen 9 8945HS. We’ll have to wait and see if there are any major performance improvements, however.

Based on the spec sheet, Gorgon Point chips look mostly like a frequency bump. The CPU and GPU cores are clocked higher, and they’ll be paired with faster memory. In instances where there isn’t a frequency bump, AMD has pushed up other specs. For instance, the Ryzen AI 5 430 matches the Ryzen AI 4 330 across the board, but the Gorgon Point offering comes with a more capable Radeon 840M iGPU.

Compared to the Intel Core Ultra 9 288V running at 30W, AMD says the Ryzen AI 9 HX 470 running at 28W is 1.3x faster at multitasking, 1.7x faster at content creation, and 1.1x faster in gaming. AMD said its multitasking results are based on an average Procyon Office Suite while running Microsoft Teams. The content creation results from Blender, Cinebench, Handbrake, PugetBench for Photoshop and DaVinci Resolve, and 7-Zip, while the gaming results were gathered from Black Myth: Wukong, Borderlands 3, Counter-Strike 2, Cyberpunk 2077, F1 25, Monster Hunter Wilds, and Civilization V at 1080p with Low settings.

Image 1 of 4

AMD thankfully broke down these results individually, providing a bit more context. In content creation apps, you can see that applications that favor a heavy thread count see the highest boost, such as Blender and Handbrake, while editing apps like Photoshop and DaVinci Resolve see a smaller bump. Multitasking performance is up across the board, but that’s hardly surprising. In the previous generation, Strix Point chips handily beat Lunar Lake in raw performance.

Intel’s winning corner with Lunar Lake was efficiency and battery life. AMD is claiming up to 24 hours of battery life, particularly with the Ryzen AI 7 445. But the company notably didn’t include any competitive comparisons in its battery life benchmarks.

In addition to clock-speed bumps, AMD also updated the NPU on its Gorgon Point range, which is now capable of up to 60 TOPS. Compared to Intel in Procyon’s AI Vision benchmark, AMD would have you believe its NPU is far ahead of Intel. The graph shared, which you can see above, is deceptive, however. The NPU performance is just 5% ahead of the Core Ultra 9 288V.

Unfortunately for AMD, it’s comparing its Gorgon Point range to Lunar Lake, Intel’s last-gen architecture as of a few hours ago at the time of writing. Intel finally pulled the curtain back on its long-awaited Panther Lake range at CES, so we’ll have to see where the performance gap between Intel and AMD lands once laptops sporting these chips finally arrive.

In addition to the main Gorgon Point range, AMD will release Pro variants for businesses. The lineup is identical, minus the Ryzen AI 5 430. This lineup will come late in the first quarter of 2026.

As for the first Copilot+ desktop processor, that’s the only name we know of the chip by now. AMD tells us that it isn’t announcing any specific products or release dates, and it didn’t provide any specs. The chip technically falls in the Gorgon Point range, however, so it will likely sport Zen 5 CPU cores, RDNA 3.5 GPU cores, and, of course, a built-in XDNA 2 NPU to earn its Copilot+ certification.

AMD isn’t adding a new generation for its Ryzen AI Max series, otherwise known as Strix Halo. AMD admits that “Ryzen AI Max chips are not for the vast majority of use cases,” so we likely won’t see the annual generational bump that AMD uses for its main mobile lineup.

The Ryzen AI Max+ 395 remains AMD’s flagship Strix Halo offering, but it’s adding two SKUs lower down the stack. The Ryzen AI Max+ 392 and Ryzen AI Max+ 388 are identical to two lower offerings in the original Strix Halo lineup (Ryzen AI Max 390 and Ryzen AI Max 385, respectively), but they come with 40 RDNA 3.5 CUs, matching the flagship chip in the range.

In addition to rehashed benchmarks for the Ryzen AI Max+ 395, AMD shared some competitive comparisons to Intel with the Ryzen AI Max+ 392 and Ryzen AI Max+ 388. AMD unsurprisingly wins across the board, particularly in games. One of the strongest elements of Strix Halo is the iGPU performance, and Intel doesn’t currently have a competent counterpoint.

The bigger question for Strix Halo is availability. Even with three chips already available, we haven’t seen a major influx of products sporting these chips. They’re niche, even AMD admits it, so you shouldn’t expect to see Strix Halo everywhere. Although AMD says the five SKUs will live together, we’ll likely see the 392 and 388 replace the 390 and 385 eventually.

Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.