AMD enables PyTorch on Radeon RX 7000/9000 GPUs with Windows and Linux preview

AMD has released a public preview of ROCm 6.4.4 that enables PyTorch to run natively on both Windows and Linux for Radeon RX 7000 (RDNA 3) and RX 9000 (RDNA 4) GPUs , as well as select Ryzen AI 300 “Strix” and Ryzen AI MAX “Strix Halo” APUs. This follows through on the company’s Computex 2025 commitment to make ROCm a cross-platform stack with broader support for developers.

“At Computex this year, I shared our commitment to making ROCm a true cross-platform, developer-first stack. I said we’d bring ROCm to Radeon on Windows and Linux in the second half of 2025. I’m proud to say that today, we’re delivering on that promise.”

— Andrej Zdravkovic, AMD Senior Vice President and Chief Software Officer

The preview brings official PyTorch wheels, removing the need for dual-boot setups or workarounds when running AI workloads on Windows. Developers can now prototype, benchmark, and provide feedback directly on AMD consumer hardware. AMD stresses that this is only the starting point, with performance and feature coverage expected to improve in future releases.

Source: AMD

AMD describes this step as part of a wider strategy to expand ROCm beyond data center deployments into client platforms for creators and developers. At the same time, enterprise customers will continue receiving larger-scale optimizations through the upcoming ROCm 7.0 stack for Instinct and EPYC.

The PyTorch preview for Radeon and Ryzen AI hardware is available to download today. AMD says more frequent updates and expanded Windows support are planned for upcoming versions.

Source: AMD