Other Published:
AMD and Intel agree on oneDNN as PyTorch's shared x86 CPU acceleration library
What happened
AMD and Intel are combining their x86 CPU optimization work for PyTorch in oneDNN, an open-source library governed by the UXL Foundation, and have aligned on it as the default open-source x86 acceleration library in PyTorch. AMD is integrating its ZenDNN optimizations into oneDNN and has joined the UXL Foundation as a steering member. The two companies are also working on AI Compute Extensions, a next-generation x86 instruction set for matrix multiplication, with PyTorch support planned through future oneDNN updates. No release dates, specific processors or performance figures were given.
Why it matters locally: Anyone running PyTorch inference on AMD or Intel CPUs should end up with one optimized code path instead of two vendor-specific ones.
Three ways to read it
For the market
Two long-time rivals sharing one software layer suggests they see more to gain from keeping x86 attractive for AI against Arm-based chips and GPUs than from competing on libraries. It could also make CPU-only servers an easier choice for light inference, though GPUs will likely stay the default for larger models.
In plain words
AMD and Intel, which make most computer processors, have agreed to put their AI speed-up tricks for a popular AI toolkit into one shared package. It is like two car makers agreeing on one type of charging plug: developers no longer have to build for each brand separately.
For practitioners
If you serve PyTorch models on EPYC or Xeon today, track which PyTorch and oneDNN releases pick up the ZenDNN work and re-benchmark against your current setup, since a shared library does not guarantee equal speed on both vendors. AI Compute Extensions need new hardware and future oneDNN updates, so nothing changes for existing servers yet.
Sprout's insight
Sprout compares this with other brands and with where real applications are heading. Free for members.
Sign in free to read it →Subscribe to Sprout's weekly
One email a week: the top stories and a taste of Sprout’s insights. Unsubscribe any time.
Source
Read the official source: AMD Newsroom: AMD and Intel collaborate to simplify AI development on x86 CPUs using PyTorch ↗Summaries are written by OnPrem World from the official source. We link to the original and never republish it.
More news
- vLLM v0.32 brings 4-bit KV cache to more GPUs and stops allowing every CORS origin by default
- DeepSeek-V4.1-Flash-1017: a same-size update with better agent scores and shorter answers
- Open WebUI 0.12.3 adds end-to-end encrypted chats and one search box for the whole workspace
- llama.cpp v0.7.0 adds a GPU cache for offloaded MoE experts and moves the server to port 9931
- Meta releases Muse Spark Open, an 816B open-weight model it runs on two Mac Studios