OpenAI published a technical report showing its models have been used in real chip design work: on its in-house Jalapeño inference chip, standard-cell area fell about 13% in vector units and 10% in matrix units, while AMD's flagship GPU IP saw up to 14% PPA improvement, compressing roughly a year of optimization into under five months. OpenAI also announced a partnership with Synopsys to train GPT-Synopsys to operate EDA tools for design and verification.
PPA
功耗性能面积
Performance, power and area, the core metrics for evaluating chip design optimization.
Curated Intelligence
OpenAI and AMD jointly achieved up to 14% PPA (power, performance, area) improvement on AMD's flagship GPU IP in under five months, equivalent to roughly a year of additional progress beyond AMD's own multi-year optimization, with some changes already folded into a tape-out design. The same method was applied to OpenAI's first custom inference chip, Jalapeño, delivering best-in-class results on open RISC-V cores.
OpenAI detailed how its frontier models delivered expert-level PPA improvements in its custom inference chip Jalapeño, and it is extending the same approach to AMD hardware. OpenAI is also partnering with Synopsys on a GPT-Synopsys model that brings frontier intelligence into the EDA tools chip designers already use.
OpenAI detailed how its frontier models delivered expert-level PPA improvements in its custom inference chip Jalapeño, spanning vector-unit floating-point adders, multipliers, and reduction functions, and is now applying the same approach to third-party AMD hardware. The company is also partnering with Synopsys on a dedicated GPT-Synopsys model that embeds frontier intelligence directly into the EDA tools chip engineers already use for design and verification.