AMD Copilots Tesla AI Chip, Says Report

AMD Copilots Tesla AI Chip, Says Report

SAN JOSE, Calif. — Tesla is testing samples of a machine-learning chip that it developed in collaboration with Advanced Micro Devices, according to a report from CNBC. AMD and Tesla both declined to comment on the story.

The chip was developed by Tesla’s Autopilot group, a team of about 50 engineers under Jim Keller, a veteran microprocessor designer who led work on AMD’s Zen x86 processor. The chip is expected to replace an Nvidia GPU that Tesla currently uses, which itself replaced a Mobileye chip, said the CNBC report.

Keller joined Tesla in January 2016, the Electrek online automotive news site reported at the time. Within four months, Tesla hired several of Keller’s former colleagues, including processor veteran Peter Bannon and several senior AMD engineers.

A quick check of LinkedIn shows that at least a dozen senior AMD employees joined Tesla in the first months of 2016. Among the most senior was engineer-turned-strategist Keith Witek, Tesla’s director of business development who left AMD after 14 years, ending as its corporate business development strategist.

The rest of the former AMD group are engineers now generally working on silicon technology for Tesla’s Autopilot group. One notable exception is Junli Gu, who mainly worked on deep-learning software for four years at AMD before leaving to build “and lead the machine-learning team at Tesla Autopilot,” according to her online bio.

AMD’s GPU team, acquired in 2006 with ATI Technologies, has a long history of developing ASICs, manly for video game consoles such as the Xbox One X. A follow-up report from CNBC quoted Wall Street analysts saying that the Tesla chip marks a potentially disruptive entry for AMD into the market for silicon for self-driving cars.

Given Tesla’s relatively small volume of car sales, the deal, if true, is not expected to be material to AMD revenues in the near future. However, it could provide a powerful calling card in a market for machine-learning chips that is already becoming crowded with offerings from established and startup companies.

— Rick Merritt, Silicon Valley Bureau Chief, EE Times Circle me on Google+

Related posts:

  • Big Data Reshapes Silicon
  • Is ADAS Ready for a Little 'Disruption?'
  • Microsoft Describes Next Xbox SoC
  • Fatal Tesla Crash: That�s Not All, Folks

PreviousMediaTek, Huawei Partner in 5G Development
Next    EDICON: Views from the Exhibits