AMD and Intel align on oneDNN for PyTorch on x86 CPUs
The move gives PyTorch developers a common open-source path for CPU acceleration across both chipmakers’ systems.
AMD and Intel said they will unify their x86 CPU optimization work for PyTorch through oneDNN, giving developers a single open-source path for CPU-accelerated AI workloads on their processors. AMD is folding ZenDNN into oneDNN and has become a steering member of the UXL Foundation to help move the effort forward in the open. The companies say this should reduce fragmentation and make it easier to build, maintain, and deploy PyTorch applications across x86 systems. They also plan to work on AI Compute Extensions, a future x86 instruction set aimed at speeding up matrix multiplication in PyTorch via later oneDNN releases.
Why it matters
For teams building PyTorch apps on x86, this cuts down on having to handle separate optimization paths for AMD and Intel processors. AMD folding ZenDNN into oneDNN, plus the shared work on AI Compute Extensions, points to a more consistent base for current and future CPU features inside the framework. That matters most for developers trying to keep PyTorch builds and deployments aligned across different x86 machines.
Matters or hype?
Does this story matter, or is it hype? Vote before you see what everyone else said.
Sources
- AMD Newsroom