OpenAI says its models boosted AMD GPU PPA by up to 14%, compressing a year of work into 5 months
mark_k · x · 2026-11-19
OpenAI revealed results from its chip-design work with AMD: its models achieved up to 14% improvements in power, performance, and area (PPA) on heavily optimized GPU designs — compressing roughly a year of engineering into under five months, with some AI-generated improvements already in AMD chips sent for manufacturing.
- OpenAI's models also cut the area of key compute blocks in its 'Jalapeño' inference chip by 10–13%;
- A shared chart shows GPT-6.2 Astra leading on PicoRV32 RISC-V RTL optimization, scoring about twice GPT-6 Astra at 112k token budgets;
- OpenAI is partnering with Synopsys to train a specialized chip-design model, and GPT-6.2 Astra can design full RISC-V cores from written specs.
The post highlights the feedback loop: better AI designs better chips, which enable even better AI.
Related event: OpenAI Models Boost AMD Flagship GPU Design by Up to 14% PPA(2 posts)→
More from Infra
- Bill Ackman slams Nvidia board for 'major governance failures' over Jensen Huang succession — Polymarket · 2026-11-19
- 5060 Ti 16GB vs 5070 12GB: Reddit debates VRAM vs speed for local AI use — BuckinBronco999 · 2026-11-19
- OpenAI says its models found up to 14% PPA gains on AMD flagship GPU IP — ryanmerket · 2026-11-19
- rakyll is bringing agentic compute and platforms to Kubernetes, asks what to build — rakyll · 2026-11-19
- Arm CEO: NVIDIA Partnership Unchanged; AI Boom Resembles 1880s Railroad Buildout — BenBajarin · 2026-11-19
- Blackstone's Jon Gray: AI Is a 'High Bubble Alert' but Negativity Has Gone Too Far — BenBajarin · 2026-11-19