AMD’s open source software development platform ROCm (against Nvidia CUDA) recently released version 6.0, introducing improved hardware and software support.
ROCm 6.0 announced compatibility with ONNX Runtime. As an open standard for machine learning algorithms and software tools, ONNX provides developers with a seamless way to transform AI models between different frameworks.
Therefore, after upgrading to ROCm 6.0, users can reason about a wider range of source data on local AMD hardware.
On the hardware side, the ROCm 6.0 update brings support for Radeon PRO W7800 and Radeon RX 7900 GRE graphics cards. The update means that ROCm 6.0 now fully supports the Navi 31 GPU family’s desktop games and workstation portfolio.
As previously reported by Igeekphone, in the release of ROCm 5.7 last November, the platform introduced support for Radeon RX 7900 XT and PyTorch.
Notably, a third-party open source project called “ZLUDA” allows AMD Gpus to run Nvidia CUDA applications, and in Blender 4.0 rendering tests, AMD Gpus performed approximately 10-20% faster than native ROCm/HIP support.