Innovation
21/09/2017

Its Own AI Chips For Self-Driving Cars Is Being Built By Tesla Together With AMD




For handling autonomous driving tasks in its cars, Tesla is getting closer to having its own chip.
 
Sources said that the carmaker is now running tests on the samples of the first implementation of its processor which it has received back.
 
Tesla's has been pushing to be vertically integrated and decrease reliance on other companies and the effort to build its own chip is in line with that effort.
 
But Tesla will build on top of AMD intellectual property and isn't completely going it alone in chip development, according to the media quoting sources.
 
Tesla was mentioned as an example of companies that were working with fabricators by Sanjay Jha, CEO of AMD spin-off and chip fabricator GlobalFoundries, said at the company's technology conference in Santa Clara, California, on Wednesday. A wafer supply agreement in place with AMD through 2020 is with GlobalFoundries, which fabricates chips.
 
GlobalFoundries denied Jha said that the company was working directly with Tesla and said that the company does not comment on customers or potential customers.
 
Getting closer to delivering totally autonomous driving could be possible for Tesla by a more power-efficient purpose-built chip. That capability will be available to consumers in 2019, Tesla CEO Elon Musk promised this year.
 
Since the departure of Apple veteran Chris Lattner in June, under the leadership of longtime chip architect Jim Keller, the head of Autopilot hardware and software,Tesla's silicon project is bounding ahead.
 
Sources said that under Keller, there are more than 50 people working on the initiative. Including director Ganesh Venkataramanan, principal hardware engineer Bill McGee and system circuit design lead Dan Bailey, Tesla has brought on several AMD veterans after hiring Keller.
 
As part of the Autopilot self-driving hardware, current Tesla vehicles use Nvidia graphics processing units. After an accident involving a Mobileye-powered Tesla, Mobileye and Teala parted ways. Mobileye was previously the company whose chips were used by Tesla. Intel acquired Mobileye for more than $15 billion earlier this year 
 
Intel enjoys a long working relationship with Alphabet’s Waymo car group and notably Tesla is not working with Intel.
 
Capable of doing many things but aren't designed with just a few narrow computing jobs in mind, are GPUs from the likes of Nvidia as a rule, which are widely used today for various types of AI workloads, and hence the custom chip development. Plus, if it switches to in-house hardware, Tesla would be less impacted by pricing dictated by Nvidia.
 
Two generations of an AI chip that can work as a GPU alternative have already been built by Alphabet. Work on its own AI chip for HoloLens is being conducted by Microsoft. And Apple has incorporated an AI chip into its top-of-the-line iPhone X.
 
(Source:www.cnbc.com) 

Christopher J. Mitchell
In the same section