Nvidia shows new research on using AI to improve chip designs
狈惫颈诲颈补听Corp., the world’s leading designer of computer听肠丑颈辫s used in creating artificial intelligence, on Monday showed听苍别飞听谤别蝉别补谤肠丑听that explains how听础滨听can be used to听颈尘辫谤辞惫别听肠丑颈辫听design.
The process of designing a听肠丑颈辫聽involves deciding where to place tens of billions of tiny on-off switches called transistors on a piece of silicon to create working听肠丑颈辫s. The exact placement of those transistors has a big impact on the听肠丑颈辫‘s cost, speed and power consumption.
颁丑颈辫听design engineers use complex design software from firms like Synopsys Inc.聽and Cadence Design Systems Inc聽.聽to help them optimize the placement of those transistors.
On Monday,聽狈惫颈诲颈补听聽showing that it could use a combination of artificial intelligence techniques to find better ways to place big groups of transistors. The paper听补颈med to听颈尘辫谤辞惫别听on a 2021 paper by Alphabet Inc’s.聽Google, whose findings later聽.
The聽狈惫颈诲颈补听谤别蝉别补谤肠丑听took an existing effort developed by University of Texas听谤别蝉别补谤肠丑ers听耻蝉颈苍驳听what is called reinforcement learning and added a second layer of artificial intelligence on top of it to get even better results.
狈惫颈诲颈补听chief scientist Bill Dally said the work is important because听肠丑颈辫聽manufacturing听颈尘辫谤辞惫别ments are slowing with per-transistor costs in听苍别飞听generations of听肠丑颈辫聽manufacturing technology now higher than previous generations.
That goes against the famous prediction by Intel Corp.聽co-founder聽.
“You’re no longer actually getting an economy from that scaling,” Mr. Dally said. “To continue to move forward and to deliver more value to customers, we can’t get it from cheaper transistors. We have to get it by being more clever on the design.” – Reuters
