AMD and Intel announce collaboration on x86 CPU optimization for PyTorch

On October 21, 2026, AMD and Intel announced a collaboration on x86 CPU optimization for PyTorch, unifying their respective PyTorch x86 CPU optimization efforts into the open-source deep learning acceleration library oneDNN. According to a report by Semiconductor Industry Observer, AMD merged its in-house ZenDNN into it and joined the UXL Foundation under the Linux Foundation as a guiding member; after the integration, developers need only interface with a single set of APIs to obtain low-level optimizations for both AMD EPYC and Intel Xeon. The two parties are also jointly developing ACE, a next-generation x86 instruction set extension for matrix multiplication, which is planned to be gradually integrated into PyTorch through subsequent oneDNN versions. A report by Electronic Technology Application on October 22 likewise mentioned that oneDNN will become the default open-source library for x86 acceleration within PyTorch, and that AMD joined the UXL Foundation as a steering committee member. No further progress has been disclosed at present.

AI-generated from public reports

1. Report date Oct 21· 5 reports

   AMD and Intel announce collaboration on PyTorch x86 CPU optimization

   Source:English text is being prepared.