Well, it’s finally here. We’ve been crossing our fingers for the AMD Strix Halo for a long, long time, and now the wait is over. As one of the many rabbit hats for CES 2025AMD has just announced the top end of its Strix Point mobile processors, combining Zen 5 processor cores, integrated RDNA 3.5 graphics, and an NPU.
These new “Halo” processors are Ryzen AI Max 385, AI Max 390 and AI Max+ 395. (Yes, and AI Max 380, but this is a processor only for the Pro version, that is, for business.) They will be available in the first or second quarter of 2025.
We’ve long suspected a Strix Halo launch, and we’d heard about the name AI Max and possible characteristics since September last year. As it turned out, those characteristics that were previously rumored were almost absolutely. The only difference is that we now know that the Ryzen AI Max 390 will only have 32 CUs, not 40 as stated in previous rumors. This number of CUs is the privilege of only the top Max+ 395.
Here is the full set of characteristics:
Header Cell – Column 0 | Zen 5 cores/threads | RDNA 3.5 computing units | Frequency increase | Shared cache | Peak TOPS | Custom TDP |
---|---|---|---|---|---|---|
Ryzen AI Max+ 395 (and Pro) | 16 / 32 | 40 | 5.1 GHz | 80 MB | 50 | 45–120 W |
Ryzen AI Max 390 (and Pro) | 12 / 24 | 32 | 5 GHz | 76 MB | 50 | 45–120 W |
Ryzen AI Max 385 (and Pro) | 8 / 16 | 32 | 5 GHz | 40 MB | 50 | 45–120 W |
Ryzen AI Max 380 (Pro only) | 6/12 | 16 | 4.9 GHz | 22 MB | 50 | 45–120 W |
If you’re wondering how powerful those 40 RDNA 3.5 CUs on the AI Max+ 395 will be, keep in mind the Z1 Extreme found in Asus ROG Ally X and another couple of PDAs have only 12 RDNA 3 CUs, very similar in architecture. Even with AI Max 385 and 390, the graphics capabilities of such laptops are almost three times higher.
But these AI Max chips probably won’t be designed for laptops – what’s the use of so many CPU cores plus NPUs? Instead, AMD is clearly targeting these chips at the laptop market, especially for creative and artificial intelligence tasks.
This is thanks in part to a “unified coherent memory architecture” that can allocate up to 96GB of memory for graphics with 256GB/s bandwidth. This throughput, according to AMD, is “unprecedented in any x86 mobile device.” Ultimately, all of this combined with the NPU means AI Max chips can handle large AI workloads with “performance better than high-end desktop graphics cards.”
AMD backs up these claims with its own charts, which of course should be taken with a grain of salt. Comparing the top-end AI Max+ 395 processor with the Intel Core Ultra 9 288V processor, AMD claims 1.4x faster graphics performance and 2.6x faster rendering (including a whopping 402% performance boost in Blender Classroom). And it even appears to rival or beat Apple’s M4 rendering chips.
Obviously, we can expect good results in games as well compared to other systems with integrated graphics. For context, we found that the 16-CU 890M has modern games that can be played in 1080p resolutionand these new AI Max processors will have at least twice as many processors as this mobile GPU.
That’s why I certainly take AMD seriously when they say this is “simply the most advanced x86 mobile processor ever made.” The only question is price and target market.
Unfortunately, I don’t think these AI Max chips are the best options for gaming. Sure, they’ll be good at gaming, but they’re sure to be expensive considering their overall feature set, which will benefit creative professionals and graphics/AI developers.
In this case, I suspect we’ll get more gaming performance out of gaming laptops with dedicated mobile GPUs. On the other hand, the Strix Point uses less power than a full-fledged gaming chip and GPU, so perhaps the battery life will justify the likely premium price. We’ll have to wait and see.