AMD and Intel plan a shared oneDNN path for CPU AI optimization
Announced October 20 and reviewed in ITmedia's October 25 coverage. Integration and future updates are not presented as completed.
IllustrationTheir collaboration aims to combine PyTorch optimizations for x86 CPUs in oneDNN. October 25 coverage highlighted the potential to reduce separate development paths.
Read for 8 seconds to inform recommendations.Reading history & counting settings
Announced October 20 and reviewed in ITmedia's October 25 coverage. Integration and future updates are not presented as completed.
A common software route
AMD says the companies will bring their x86 CPU optimization work for PyTorch into the open-source oneDNN library. AMD is integrating ZenDNN optimizations and has joined oneDNN's governing UXL Foundation as a steering member.
Development and maintenance matter too
The companies aim to simplify building, maintaining and deploying AI applications. Support for the ACE instruction set is planned through future oneDNN updates. This does not promise identical performance on every CPU or unify all GPU software; actual supported environments and progress still matter.
Sources & further reading
Our own summary based on the public information below.
This is a record of information available at publication. Later reports may change the details. Check the source for current weather, safety and market information.
