AMD and Intel Partner to Unify PyTorch CPU Optimization for x86

On October 20, AMD and Intel announced a joint effort to optimize PyTorch on x86 CPUs, consolidating their respective optimization work into the open-source deep learning library oneDNN as the default x86 acceleration path within PyTorch.

AMD will integrate its ZenDNN optimizations into oneDNN and join the UXL Foundation as a steering committee member. The collaboration was facilitated by the x86 Ecosystem Advisory Group, marking a rare partnership between the two CPU rivals on software.

Beyond oneDNN, the companies are advancing AI Compute Extensions (ACE), a next-generation x86 instruction set architecture focused on accelerating matrix multiplication, a core operation in AI model training and inference. ACE capabilities are planned to be delivered to PyTorch through future oneDNN updates.

Developers are expected to benefit from a unified software path for PyTorch workloads across both AMD and Intel CPUs, reducing fragmentation in the x86 AI ecosystem and lowering duplicated development and maintenance efforts.