AMD Has Enabled PyTorch Support on Windows for Radeon RX 9000, RX 7000 GPUs & Ryzen AI APUs with ROCm 6.4.4

AMD has finally enabled PyTorch support on Windows for Radeon RX 9000 and RX 7000 series GPUs, along with Ryzen AI APUs such as the 300 “Strix” and MAX “Strix Halo” models. This is achieved through ROCm 6.4.4.

AMD Radeon RX 9000, RX 7000 GPUs, & Ryzen AI APUs Now Support PyTorch on Windows Thanks to ROCm 6.4.4

At Computex 2025, AMD promised to bring PyTorch support for Windows through ROCm to its consumer GPU/APU lineup. Today, with the release of ROCm 6.4.4, this promise is fulfilled, adding PyTorch support on both Windows and Linux for a range of consumer-centric GPUs and APUs.

PyTorch support on Windows/Linux via ROCm 6.4.4 will be available on the latest Radeon RX 9000 (RDNA 4), Radeon RX 7000 (RDNA 3) series GPUs, along with Ryzen AI 300 “Strix”, and Ryzen AI MAX “Strix Halo” APUs.

At Computex 2025 we committed to making ROCm a true cross-platform, developer-first stack. Today we take an important step on that journey by enabling native PyTorch support on Windows and Linux for users with AMD Radeon 7000 & 9000 GPUs and Ryzen AI 300 & Ryzen AI Max APUs.

This preview release means developers can now run AI models directly on AMD hardware in the Windows environment. The preview is designed to be a foundation for developers, helping shape what comes next as performance and feature coverage continue to improve and evolve.

AMD ROCm 7 Software Stack | Image Credits: AMD

ROCm 6.4.4 is being released as a preview, enabling developers to run AI models on supported hardware in the Windows environment. Future upgrades will focus on enhancing performance and adding features to ROCm on Windows. On the enterprise side, AMD has already released ROCm 7, which further accelerates AI and productivity innovation on Instinct/EPYC hardware.

To get started with ROCm, you can visit this link. AMD also provides full documentation for PyTorch on Windows/Linux here.