AIMedium

OpenAI models improve AMD GPU design

OpenAI says its models found up to 14% PPA improvement on AMD flagship GPU IP in under five months, with some results in a recently taped-out design.

14%PPA improvement on AMD flagship GPU IP

What we know

OpenAI says working with AMD its models found up to 14% PPA improvement on flagship GPU IP that engineers had optimised for years. Greg Brockman says this equals roughly an extra year of progress in under five months and that some results are in a recently taped-out design. The approach also improved Jalapeño, OpenAI's first custom inference chip.

Who's who

Photos and descriptions: Wikipedia

What people are posting

Posts on X this story draws on. They are individual users' views and have not been verified by Global Eye 24.

@gdb

“roughly an additional year of progress on AMD's chip achieved in less than 5 months”

View post →

By the Global Eye 24 desk, with AI assistance. Updated as the story develops. Spot an error? Write to corrections@globaleye24.com.