AMD has launched what one executive called “the most advanced mobile X86 processor ever created” at CES 2025: The Ryzen AI Max and AI Max+, with absolutely massive capabilities to run graphics and AI workloads.

AMD is positioning this “Strix Halo” chip as a sort of hybrid for graphics and AI workstations, comparing it to Nvidia’s existing GeForce 4090 GPU in terms of running AI LLMs at up to 70 billion parameters. But the Ryzen AI Max offers more than just that.

It’s an APU with graphics capabilities that push into discrete GPU territory — in the 3DMark Steel Nomad benchmark, for example, the chip offers 258 percent the graphics performance of Intel’s Core Ultra 9 288V (Arrow Lake) CPU. It also offers excellent AI performance, both with or without the NPU.

The subtext behind this announcement hearkens back to our report early last year that the NPU doesn’t matter for AI capabilities as much as AMD, Intel, or Qualcomm originally hyped.

The NPU is the most efficient AI core on modern CPUs. But for raw horsepower, the GPU, especially a discrete GPU, far outperforms either the NPU or the GPU by its lonesome. What the AI Max appears to do is combine some of the NPU and GPU’s best features, alongside an already powerful AMD Zen 5 CPU.

AMD will offer the Ryzen AI Max in both consumer and “Pro” versions, which the company will sell to businesses. At CES, AMD is touting three design wins, including the HP ZBook Ultra G1a and Z2 Mini G1a mini workstations, plus the Asus ROG Flow Z13 tablet.

“This is something very, very special,” said Rahul Tikoo, the senior vice president and general manager of AMD’s client computing business said in a recorded briefing for reporters.

“It’s just so unique and powerful. It enables incredible performance and reshaping what customers can experience from the power of workstation and thin-and-light laptops to incredibly small and powerful micro desktops. This is simply the most advanced mobile x86 processor ever created.”

AMD’s Ryzen AI Max and its features

Interestingly, AMD is offering four of the AI Max chips as commercial products — the Ryzen AI Max+ Pro 395, the Ryzen AI Max Pro 390, the Ryzen AI Max Pro 385, and the Ryzen AI Max Pro 380. However, the first three are being offered in consumer-friendly non-Pro versions.

• Ryzen AI Max+ 395: 16 cores/32 threads, 5.1GHz turbo, 80MB cache; 40 graphics cores; 50 peak TOPS, 45-120W cTDP
• Ryzen AI Max 390: 12 cores/24 threads, 5.0GHz turbo, 76MB cache; 32 graphics cores; 50 peak TOPS; 45-120W cTDP
• Ryzen AI Max 385: 8 cores/16 threads, 5.0GHz, 40MB cache; 32 graphics cores; 50 peak TOPS; 45-120W cTDP

The fourth chip, the Ryzen AI Max Pro, is a 6 core/12-thread part with 22MB of cache and running at 4.9GHz. All of these will be available in either the first or second quarter of 2025, AMD said.

What’s interesting about the AI Max parts is they’re really not differentiated by clock speed. Instead, CPU cores count, the number of graphics CUs and especially the cache size separate the three chips. (It’s not quite clear what AMD means by ay the AI Max+ versus the more generic AI Max.)

What this chip doesn’t appear to have is the stacked V-Cache that differentiates AMD’s Ryzen X3D chips — just a lot of what appears to be ordinary level 3 cache.

In graphics, AMD has only launched two RDNA 3.5 GPUs, the Radeon 880M and 890M, which are part of the “Strix Point” CPU architecture that makes up the Radeon AI 300. Those chips boast 12 CUs and 16 CUs, respectively, far less than what the new Ryzen AI Max chips include.

Alternatively. the Radeon 7000 series includes both the Radeon 7600XT as well as the Radeon 7700XT, which boast 32 CUs and 54 CUs. Both of those GPUs use the older RDNA 3.0 architecture. (Read our Radeon 7600 review as well as our Radeon 7700XT review for more details.)

AMD executives said that the Ryzen AI Max series is designed with an 256GB/s coherent memory interface of unknown width that can address up to 96GB of graphics memory. Nevertheless, that’s close to the RX7600’s memory bandwidth of 288GB/s — not so critical for graphics, but for the Large Language Models (LLMs) that depend on high memory bandwidth to run.

AMD is claiming that the Ryzen AI Max is the world’s first Copilot+ processor to run an undisclosed 70 billion parameter LLM about 2.2 times as fast as the Nvidia RTX 4090 24GB, at 87 percent less power.

AMD is currently positioning the Ryzen AI Max at artists, developers, and creators. But with such a powerful integrated GPU, can gamers be that far behind?

Author: Mark Hachman, Senior Editor, PCWorld