|

INTEL: OpenAI says its models found up to 14% gains in AMD flagship GPU design

OpenAI says its AI models improved power, performance and area (PPA) by up to 14% on AMD’s flagship GPU IP, measured against heavily optimized baselines, in less than five months. The claim comes from an OpenAI blog post dated 18 November. It says AMD has already included some of the changes in a design sent for manufacturing.

According to OpenAI, the models wrote RTL, the code that describes a chip’s logic. OpenAI calls the result roughly a year of progress compressed into under five months. It says AMD added other changes to its production files for future versions of the design.

What OpenAI says about AMD’s CPU work

The post also says AMD verified single-digit PPA gains on key blocks of its next flagship CPU. OpenAI gives no figure for the CPU as a whole.

What is not specified

OpenAI does not name the AMD products involved. It gives no benchmark for a complete GPU, so the 14% figure applies to the optimized IP blocks, not to a finished chip. The numbers are OpenAI’s own and rest on its description of AMD’s verification.

The work adds a design role to an existing commercial relationship. OpenAI has a 6GW agreement with AMD for chips it plans to buy. AMD shares rose 3.3% on Wednesday after the news.

Share

Similar Posts