AMD unveiled ROCm 6.0: Supports Radeon PRO W7800 and RX 7900 GRE GPUs

At the close of the previous year, AMD announced that ROCm 5.7 and PyTorch now support the Radeon RX 7900 XTX, Radeon RX 7900 XT, and Radeon PRO W7900 graphics cards, encompassing the pinnacle products utilizing the RDNA 3 architecture, thereby enabling developers to select the solutions most suited to their requirements.

Recently, AMD unveiled ROCm 6.0, further broadening the array of supported devices to include the Radeon PRO W7800 and RX 7900 GRE graphics cards, signifying that all desktop and workstation graphics cards equipped with the Navi 31 GPU have received comprehensive support. AMD’s intention, by extending support across a wide product range, is to assist the artificial intelligence (AI) community in accessing a diverse array of desktop graphics cards at various price points and performance levels, to enhance their development efforts.

Additionally, AMD has augmented its solution stack through support for ONNX Runtime. ONNX, an acronym for Open Neural Network Exchange, serves as an intermediary machine learning framework, establishing an open standard for the conversion of AI models across different ML frameworks.

PyTorch, developed by Facebook AI Research, is a deep learning library that, as an open-source tool, has garnered acclaim within the industry and broad support from CPU and GPU manufacturers. AMD’s initial response to integrating support for PyTorch, particularly for consumer-grade Radeon and Radeon PRO graphics cards, was notably slow. However, this scenario has seen significant evolution over the past year. Moreover, prioritizing support for flagship graphics cards is also influenced by the needs of large language model (LLM) applications for AI, which require extensive memory capacity.