Cancel Preloader

AMD Rebrands Next-Gen Mobile CPUs ‘Ryzen AI’, Claims 50 TOPS NPUs

 AMD Rebrands Next-Gen Mobile CPUs ‘Ryzen AI’, Claims 50 TOPS NPUs

Only six months ago, AMD (and Intel) was touting its “AI PC” laptop processors with the company’s second-gen neural processing unit for low-power AI acceleration — and they’re already so last year. Now it’s announcing new mobile CPUs with a theoretically much better third-gen NPU (XDNA 2, with up to 50 TOPS), incorporated into an updated CPU using the new generation Zen 5 process and an upgraded integrated GPU (RDNA 3.5). Rebranded Ryzen AI, AMD’s launching with a couple of Copilot Plus-ready 300 series parts.

Formerly code-named Strix Point, the Ryzen AI 300 series are the successors to the Ryzen 8000 generation; since the older chips don’t have the NPU bandwidth of the 300s — they’re pretty limited — they will retain their current naming. 

Laptop launch partners for the new processor are the same Copilot Plus PCs as the ones Nvidia had for its RTX AI PCs earlier in the day, predominantly from Asus.

AI Atlas art badge tag AI Atlas art badge tag

What’s in a name?

Arriving in laptops starting in July, the Ryzen AI chips keep the same essential naming conventions, just with AI shoehorned in and modifiers like HX now used to indicate the market tier rather than the power-consumption class. 

AMD’s launching with two flagship parts, the Ryzen AI 9 HX 370 and Ryzen AI 9 365, which differ by the number of CPU and GPU cores.

The new Zen 5 CPUs actually seem to be split into 4nm Zen 5 cores and 3nm Zen 5c cores;  AMD wouldn’t provide any details on the architecture, so it’s not clear why the CPU mixes the two, though it’s not unprecedented for AMD (it mixes on its graphics cards too). 

It could be a way to match the performance/efficiency cores of Intel’s and Apple’s chips for extra power savings or to manage thermals. Since the new chips are intended to serve a wide range of laptop types, with a power draw of anywhere from 15 watts to 54 watts (small ultralights to larger thin-and-lights), it could be structured to allow manufacturers more granularity in defining the overall power draw of the system. As you’d expect, there are bumps in the number of cores, frequencies and cache size, which should result at least in the typical generation-over-generation performance improvements.

Ryzen AI 300 series chip specs

Ryzen AI 9 HX 370 Ryzen AI 9 365
Zen 5c CPU cores (3nm) 8 6
Zen 5 CPU cores (4nm) 4 4
Total threads 24 20
Single core burst frequency 5.1 5
CPU cache 36MB 34MB
iGPU Radeon 890M Radeon 880M
GPU compute units 16 12
NPU XDNA 2 up to 50 TOPS XDNA 2 up to 50 TOPS
Power envelope 15-54w 15-54w

The Radeon 890M and 880M integrated GPUs aren’t a huge leap over the last gen, but the 890M does introduce a new maximum for the number of compute units. The previous top-of-the-class was the 780M with 12 CPUs, and now there’s an 890M that increases the maximum number of GPU CUs to 16.

Leapfrogging the NPU

If you were wondering why Qualcomm and Microsoft made their Copilot Plus launch two weeks before Computex, my theory is it would otherwise have been overshadowed by AMD (and maybe Intel?) — because, among other reductionist talking points, 50 TOPS is more than 45 TOPS. TOPS, a measure of integer math performance in NPUs, only really matters to Microsoft, because that’s its primary criteria to be considered a Copilot Plus PC. Copilot uses only integer math to perform its tasks, like summarizing, drafting, quick and dirty background removal in videoconferences, tracking everything you do and more.

But integer math is pretty limited if it needs to perform calculations with very big or very small numbers, like those needed for image and video-output generative AI. Those use floating point calculations  (which refers to how the numbers are manipulated in computer registers).  Most existing applications that have implemented AI tools or features — creative, gaming and more — already use 16-bit floating point (called “single precision”), and to a lesser extent, 32-bit FP (double precision).

That’s all to say there’s another aspect of the consumer CPU’s NPU where Qualcomm has just been lapped by AMD: Block floating point. It’s essentially a way to compress 16-bit floating point values so that they can be manipulated potentially as fast as the 8-bit integers NPUs use, theoretically without losing the accuracy of 16-bit.

Ryzen 9000 series

Ryzen 9 9950X Ryzen 9 9900X Ryzen 7 9700X Ryzen 5 9600X
Zen 5 CPU cores (4nm) 16 12 8 6
Total threads 32 24 16 12
Single core max burst frequency 5.7 5.6 5.5 5.4
CPU cache 80MB 76MB 40MB 38MB
Power 170w 120w 65w 65w
Socket AM5 AM5 AM5 AM5
Available July 2024 July 2024 July 2024 July 2024

AMD also debuted new Ryzen 9000 X series desktop processors using Zen 5 architecture. They still use the AM5 socket, so you can upgrade old systems with them, but if you replace your motherboard with one of the new models incorporating the X870/X870 chipsets, you’ll gain USB 4.0, PCIe 5 and higher speeds for AMD’s EXPO (profiles used for its one-click overclocking).

Related post

Leave a Reply

Your email address will not be published. Required fields are marked *