AMD Takes On Intel, Apple And Nvidia With Ryzen AI Max Chips

The forthcoming Ryzen AI Max chips for Microsoft’s Copilot+ PC program ramp up the CPU and GPU core counts while introducing a new unified, coherent memory architecture that allows the PC to allocate up to 96 GB of system memory for the GPU.

AMD said it’s taking AI PC capabilities to the next level with its new Ryzen AI Max chip lineup, calling it the “most advanced mobile x86 processor ever created” and promising to beat top chips from Intel, Apple and Nvidia in key areas.

Revealed at CES 2025 on Monday, the Ryzen AI Max series is meant to act as the “halo” product line in AMD’s family of processors for Microsoft’s Copilot+ PCs and is set to provide significant enhancements over the top Ryzen AI 300 chips from last year, ranging from increased core counts to a new unified, coherent memory architecture.

After debuting the Ryzen AI series brand last year to highlight chips that enable next-generation AI PC experiences, AMD expects the chip family to be used in more than 150 PC designs from Lenovo, HP Inc., Acer, Asus and other OEMs this year.

“The Ryzen AI Max adds a whole new class of PCs available for AMD next-gen AI platforms in the market,” said Rahul Tikoo, general manager of AMD’s client computing business, in a briefing. “This is something very, very special—a chip so unique and powerful it enables incredible performance, reshaping what customers can experience from the power of a workstation and thing-and-light laptop to incredibly small and powerful micro desktops.”

Set to debut in the first half of this year, the Ryzen AI Max series will appear in upcoming Copilot+ PCs such as the HP ZBook Ultra G1A mobile workstation, the HP Z2 Mini G1a mini desktop workstation and Asus ROG Flow Z13 gaming 2-in-1, according to AMD.

It’s not yet known if Copilot+ features will be available when these systems debut because they have only been made available to Windows Insider Community testers for now and Microsoft has yet to announce a general availability date.

AMD also revealed several other new chip lines, including a mid-range expansion of the Ryzen AI 300 series for consumer and commercial laptops, the Ryzen 9 9950X3D desktop CPU with 16 cores and a 5.7GHz boost frequency, the Ryzen 9955HX3D laptop CPU with 16 cores and a 5.4GHz boost frequency and the Radeon RX 9070 desktop GPUs.

AMD also announced the Ryzen 200 series processors for consumer and commercial laptops as well as the Ryzen Z2 series for handheld PC systems.

AMD Calls Ryzen AI Max Series ‘Unprecedented’

The Ryzen AI Max series consists of four models, the top three of which will have options for a consumer version. All four models, on the other hand, have a commercial version that comes with AMD’s Ryzen Pro management and security features.

With up to 16 Zen 5 cores, the Ryzen AI Max series increases the maximum core count by four while upping the number of XDNA 3.5 GPU cores by 16 to a maximum of 40 compute units compared to the flagship Ryzen AI 9 HX 375 that debuted last year.

The lowest-end chip in the Ryzen AI Max series comes with six cores, 12 threads, a maximum clock boost of 4.9GHz and 16 graphics cores.

Meanwhile, the lineup’s XDNA 2 NPU (neural processing unit) is capable of up to 50 trillion operations per second (TOPS) across all models. This is slightly under the 55 TOPS maximum that is possible with the Ryzen AI 9 HX 375.

The CPU, GPU and NPU are tied together with a new “unified, coherent memory architecture” that allows the PC to allocate up to 96GB of system memory for the GPU, according to Tikoo. This is boosted by a memory bandwidth of 256 GB/s, which the AMD executive called “unprecedented in any x86 mobile device.”

“It bridges the gap between awesome performance and efficient portability in a whole new way for Windows laptops,” Tikoo said.

How Ryzen AI Max Compares To Intel, Apple And Nvidia Chips

In handpicked comparisons to chips from Intel, Apple and Nvidia, the flagship Ryzen AI Max+ 395 was shown by AMD to perform better by varying degrees across benchmarks for 3-D rendering, graphics and running inference for large language models.

Compared to Intel’s recently launched eight-core Core Ultra 9 288V, the 16-core Ryzen AI Max+ 395 is, on average, 2.6 times faster across six 3-D rendering benchmarks, which included Cinebench 2024 nT, Corona and Blender Classroom, according to AMD.

As for graphics benchmarks such as Night Raid, Time Spy and Fire Strike, the Ryzen AI Max+ 395 is 40 percent faster on average, AMD claimed.

The chip designer also showed that the Ryzen chip is faster than Apple’s 12-core M4 Pro chip—which is available in new MacBook Pros and Mac minis—by 12 percent in Cinebench 2024 nT, 25 percent in Blender Classroom, 39 percent in Corona and 86 percent in Vray.

Compared to Apple’s 14-core M4 Pro, the Ryzen chip is faster by 10 percent in Blender Classroom, 24 percent in Corona and 59 percent in Vray, according to AMD. However, the company conceded that the Ryzen AI Max+ 350 is 3 percent slower than the 14-core M4 Pro in the Cinebench 2024 nT benchmark.

AMD also said the Ryzen AI Max+ 395 could outshine Nvidia’s 24-GB GeForce RTX 4090 desktop graphics card when it comes to AI workloads. When running a 70-billion-parameter Llama 3.1 large language model in LM Studio, the Ryzen chip was 2.2 times faster when measuring tokens per second, according to the chip designer.

The company noted that the Ryzen system used 128GB of system RAM while the GeForce system had 32 GB of RAM and 24 GB of GPU memory for this comparison.

AMD pointed out that the Ryzen system used significantly less energy than the GeForce system in the test, with the Ryzen AI Max+ 395 platform using 55 watts for its thermal design power, 87 percent less than the RTX 4090’s 450 watts.

“The Ryzen AI Max processor is really a game changer for creators, artists and developers,” said Tikoo.