Brockman: OpenAI Gave AMD a Year of Chip Gains in 5 Months
Reported by Blockchain News, Crypto Briefing, FourWeekMBA and 3 more outlets6 sources ↓
OpenAI says its models improved power, performance and area by up to 14% on AMD flagship GPU designs in under five months, which it equates to about a year of further optimization. AMD has incorporated some changes into a design sent for manufacturing, according to OpenAI.
Key points
- OpenAI says its models delivered single-digit PPA gains on key blocks of an upcoming AMD flagship CPU
- Some model-written GPU changes were included in an AMD design sent for manufacturing, OpenAI says
- OpenAI reports that model-assisted changes to its Jalapeño inference chip were tested on silicon
- OpenAI and Synopsys are developing GPT-Synopsys to use electronic design automation tools
Sources
- Blockchain NewsRead ↗OpenAI reports 14% PPA gains with AMD, Synopsys collaboration
- Crypto BriefingRead ↗OpenAI says its models helped AMD compress a year of chip design into five months
- FourWeekMBARead ↗Brockman: OpenAI Gave AMD a Year of Chip Gains in 5 Months
- Unite.AIRead ↗OpenAI Reports Model-Driven Gains in AMD Flagship Chip Designs
- The DecoderRead ↗OpenAI says its models delivered a year's worth of chip optimization on top of AMD's own engineering
- Tom's HardwareRead ↗OpenAI helped AMD improve its upcoming flagship GPU design by 14% PPA — GPT models also helped Lisa Su's outfit boost its new CPUs
Details
OpenAI says its models rewrote SystemVerilog code describing AMD GPU circuits, with Synopsys tools used to check that the changes preserved behavior and assess power, performance and area trade-offs. The company says the GPU designs had already been optimized by AMD engineers for years before the additional work, which took less than five months.
OpenAI also reports single-digit percentage PPA gains across key blocks of an upcoming AMD flagship CPU. It says AMD included some GPU changes in a design sent for manufacturing; the articles do not specify which changes or provide an independent measurement of the claimed gains.
The same approach was applied to Jalapeño, OpenAI’s custom inference chip. Unite.AI reports that model-proposed changes reduced standard-cell area by about 13% in its warp engine and 10% in its matrix unit, with the optimizations tested on silicon.
OpenAI is also working with Synopsys on GPT-Synopsys, which is intended to use electronic design automation tools in chip-design tasks.
Written from coverage as of
Images
- Tom's Hardware
- Unite.AI
- Blockchain News
- Crypto Briefing
Stocks in this story
Change since open: 20 Nov to 21 Nov 2026
Last refresh