OpenAI says its models helped AMD improve power, performance, and area (PPA) metrics by up to 14% on its flagship GPU design, compared with an optimized baseline design.
According to OpenAI, the models continued to find improvements by rewriting circuit-description code. The company says the results achieved in less than five months are equivalent to about a year of progress.
AMD incorporated some changes into recent designs that have moved to production. The remaining changes were recorded in the production design file for use in future versions.
AMD also confirmed that model-generated designs improve PPA by a single-digit percentage across critical modules of its next-generation flagship CPU.
The same approach was applied to Jalapeño, OpenAI’s first in-house inference chip. The related improvements were verified on the actual chip.
Source: https://tintucbitcoin.com/openai-mo-hinh-ai-cai-thien-toi-14-chi-so-thiet-ke-gpu-dau-bang-amd/
Thank you for reading this article!
Like, comment, and follow TinTucBitcoin to stay up to date with the latest news on the cryptocurrency market and never miss any important information!
$BTC $ETH $BNB $XRP $SOL
