AMD has announced significant updates to its Radeon Open Compute (ROCm) platform, bringing hardware acceleration support to the recently unveiled Ryzen AI Max 300 ´Strix Halo´ client processors and the Radeon RX 9000 series gaming GPUs. With this update, the Ryzen AI Max 300 can fully leverage the performance of its 40 RDNA 3.5 compute units, which house 80 dedicated Artificial Intelligence accelerators and 2,560 stream processors. In addition, up to 16 Zen 5 CPU cores with an advanced 512-bit floating-point unit now offer enhanced computational capabilities, including support for AVX512 instructions tailored for complex, AI-driven workloads.
The Radeon RX 9000 series stands to benefit notably from ROCm´s expanded support. These GPUs, based on the RDNA 4 architecture, feature up to 64 compute units, up to 128 AI accelerators, and as many as 4,096 stream processors. By enabling ROCm hardware acceleration, AMD is positioning these gaming graphics cards as viable solutions for high-performance computing tasks, machine learning, and Artificial Intelligence applications far beyond traditional gaming scenarios.
On the software front, AMD´s ROCm platform will now better accommodate diverse operating environments. Immediately available is support for OpenSuSE Linux, while Ubuntu and Red Hat EPEL distributions are scheduled to receive official ROCm support in the second half of 2025. In a notable move, AMD confirmed full Windows compatibility for ROCm, including integration with leading frameworks such as Pytorch and ONNX-EP. Previews for Pytorch support are anticipated in the third quarter of 2025, while ONNX-EP integration previews are slated for July 2025, ensuring that developers and users across both Linux and Windows ecosystems can tap into the expanded artificial intelligence capabilities of AMD´s latest hardware.
