AMD Folds ZenDNN Optimizations Into oneDNN After x86 EAG Recommendation

AMD has folded its ZenDNN optimizations into oneDNN, a move intended to simplify AI development for x86 CPUs using PyTorch. Intel and AMD announced the step as an expansion of their collaboration through the x86 Ecosystem Advisory Group (EAG).

According to Wccftech, the change follows a recommendation by the x86 EAG. The source says the group has encouraged closer work between Intel and AMD on CPU optimizations and other improvements across the x86 ecosystem.

Bringing ZenDNN optimizations into oneDNN is presented as a way to offer developers a more unified software path for x86 CPU-based AI development with PyTorch. Intel’s Jeff McVeigh said developers should not need separate software paths for x86 CPUs, according to the report.

The announcement builds on the EAG’s stated goal of advancing the x86 segment through collaboration between the two companies. The source does not provide further technical details about the integration or specify when it will be available.