Dev.to AI 🤖 Ai 👁 0 📖 2 min read

OpenAI's Models Rewrote AMD's GPU Circuits — and Got 14% More Out of Them

AI models writing code is old news. AI models redesigning the chips that run them is the new frontier — and OpenAI just published receipts. In a technical report released November 18, "Advancing chip design with OpenAI m

AI models writing code is old news. AI models redesigning the chips that run them is the new frontier — and OpenAI just published receipts. In a technical report released November 18, "Advancing chip design with OpenAI models," the company says its frontier models rewrote the circuit code of AMD's flagship GPU and improved power, performance, and area by up to 14%.

That is against a baseline AMD's own engineers had already tuned hard. OpenAI describes it as roughly a year of design progress in less than five months. And the changes are not sitting in a slide deck: AMD has already put some of them into a chip design sent for manufacturing, with the rest written into production design files for future products.

The models worked directly on SystemVerilog RTL, the code that describes a chip's actual logic. Each rewrite was checked with Synopsys electronic design automation tools to confirm the block still behaved exactly as specified, then measured on the three-way trade-off engineers call PPA: power, performance, and area. AMD also verified single-digit percentage PPA gains across key blocks of its next-generation flagship CPU using model-written designs.

OpenAI's own chip project tells the same story. The company began using GPT-4o to help design Jalapeño, its first custom inference chip, in early 2025. Model-proposed changes cut standard-cell area by about 13% in the vector unit and 10% in the matrix unit — measured in real silicon. One multiplier redesign used about 15% less cell area while cutting latency from two clock cycles to one. The chip went from initial design to tape-out in nine months.

Keep the caveats in mind. Fourteen percent is an "up to" figure on specific GPU IP, not a whole-chip gain. OpenAI named no AMD product, gave no breakdown between power, speed, and area, and reported no measurements from finished AMD silicon. All of it is vendor-reported, not independently reproduced. Still, a double-digit gain on an already optimized design is the kind of result that normally takes months of expert work.

OpenAI is now going further. It has partnered with Synopsys to train GPT-Synopsys, a model that can use Synopsys's tools the way expert chip designers do. If the AMD results hold up in finished silicon, chip design may be one of the first fields where AI does expert-level engineering at scale — and the chips it designs will run the next generation of AI models.

Read the full story, with all the numbers and caveats, on Rohan Tech Official: https://rohantechofficial.com/tech-news/openai-ai-redesigned-amd-flagship-gpu/

📰 Read the original article on Dev.to AI

Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.