OpenAI published details on how its models are accelerating chip design. In partnership with AMD, the models identified improvements of up to 14% in PPA (power, performance and area) relative to already heavily optimized baselines on the company's flagship GPU IP.
Results with AMD
According to OpenAI, this represents roughly a year of progress achieved in less than five months. AMD has already included some of these changes in a recent design it sent for manufacturing and added others to production design files for future versions. The models rewrote SystemVerilog RTL code, using Synopsys tools to verify behavior and evaluate trade-offs.
AMD also verified single-digit percentage PPA gains in key blocks of its upcoming flagship CPU.
Work on the Jalapeño chip
OpenAI also detailed optimizations on its own Jalapeño inference chip. In one multiplier redesign, cell area was reduced by about 15% while latency dropped from two clock cycles to one. Overall estimates point to roughly 13% reduction in standard-cell area in the vector unit and 10% in the matrix unit. These optimizations were silicon-proven during Jalapeño's bring-up.
The company is also working with Synopsys on a GPT-Synopsys model to integrate frontier intelligence with the EDA tools engineers already use.
Sources
Content produced with the assistance of artificial intelligence. Facts verified from official public sources.
Por Grok