In addition to
Following the recent
ROCm.AI provides skills for popular agents like Claude, Codex, Cursor, and Gemini to help them become "ROCm superusers" to provide AI-assisted optimizations for adapting or improving codebases around ROCm and is capable of end-to-end AI workload optimizations with its new Hyperloom component. Hyperloom aims to automate the optimization of end-to-end inference workloads and helps with analyzing, kernel optimizations, and validation in a much shorter time than would take with the effort manually.
ROCm.AI leverages AI-powered optimizations for compute kernels, memory management, parallelization, and scheduling for enhancing both AI inference and training performance.
ROCm.AI in theory sounds fantastic: using AI to help adapt your software for AMD's AI stack. In practice it will be interesting to see how well it works across diverse codebases. AMD's claims are that ROCm.ui delivers an average of 3.3x inference improvement and 2.4x training improvement -- though that's compared to ROCm 7.0 as a baseline.
Those are the light details for now but we look forward to learning more about ROCm.AI in its next release due out in August.
Instinct MI455X and Helioslaunching along with the newAMD EPYC 9006 "Venice" processors, AMD used their annual Advancing AI day to announce ROCm.AI as an AI-driven platform for developers.Following the recent
ROCm 7.14release, ROCm.AI will be part of the next release in August for leveraging AI to help developers adapt their code for in turn running on ROCm.ROCm.AI provides skills for popular agents like Claude, Codex, Cursor, and Gemini to help them become "ROCm superusers" to provide AI-assisted optimizations for adapting or improving codebases around ROCm and is capable of end-to-end AI workload optimizations with its new Hyperloom component. Hyperloom aims to automate the optimization of end-to-end inference workloads and helps with analyzing, kernel optimizations, and validation in a much shorter time than would take with the effort manually.
ROCm.AI leverages AI-powered optimizations for compute kernels, memory management, parallelization, and scheduling for enhancing both AI inference and training performance.
ROCm.AI in theory sounds fantastic: using AI to help adapt your software for AMD's AI stack. In practice it will be interesting to see how well it works across diverse codebases. AMD's claims are that ROCm.ui delivers an average of 3.3x inference improvement and 2.4x training improvement -- though that's compared to ROCm 7.0 as a baseline.
Those are the light details for now but we look forward to learning more about ROCm.AI in its next release due out in August.