Key Info

OpenAI reports that its models, working with AMD, identified up to 14% PPA (power, performance, area) improvement on flagship GPU IP in less than five months, building on designs AMD engineers had already spent years optimizing.

Highlights

  • The same approach improved Jalapeño, OpenAI's first custom inference chip, with gains verified on real silicon.
  • New evaluations show the models getting better at certain chip design tasks.
  • OpenAI is working with Synopsys to bring its model capabilities together with Synopsys' trusted EDA design and validation tools to more chip designers.
  • The results suggest AI models can contribute optimization gains on hardware already heavily tuned by human engineers.