Say Goodbye To Nvidia, Tesla's Chip Has Finally Come Out.

Abstract: at present, Tesla is developing a hardware called "Hardware3". Compared with the ready-made technology, the design and manufacture of its own AI hardware can bring a higher competitive advantage to Tesla and help Tesla stand out in the carmaker.

Micronet news has long been a secret for Tesla to build its own customized chip for autopilot.

It can be said that Tesla is one of the few automakers that develop autopilot technology, but the difference is that the company has been developing its own chips for Artificial Intelligence Computing, rather than relying on the system of microchip manufacturers such as the NVIDIA.

Developing autopilot chip from AMD angle

Musk made the first public mention of the self-developed autopilot chip project at a company event in December 2017.
At the time, Tesla's autopilot hardware used the video card of the NVDA, and musk called chip architect Jim Keller from AMD as Tesla's vice president for hardware development, and said, "we think that Jim Keller's team developed autopilot specialized AI hardware is the best in the world."

However, Keller left Tesla in April this year, after a car accident in Tesla was referred to as an autopilot system. Even so, Keller and many other AMD former architects and senior executives have formed a team within Tesla.
Today, the team is led by Pete Bannon, a chip architect who also moved to Tesla over two years ago. He worked in Apple Corp for nearly 8 years and joined Tesla in 2016.

"We have been preparing this affair for two or three years in private," Musk said at the company's second quarter financial conference call conference. "I think it's time to get the news to the public." Tesla is developing its own hardware called Hardware 3, which will be used in Models S, ModelX, and Model3 to automate driving.

Bannon confirmed at the Q2 earnings call that his team has now produced chips and carried them on a test vehicle for actual road testing.

Why Tesla gave up the NVIDIA

For a long time, Tesla relies on the Drive platform, but the reason for its self research is that it can be more personalized and efficient through self research.

When it joined Tesla, Bannon said, the staff considered a number of options, including the Vidia card, and said they had a conversation with ARM, and ARM provided technical support for chips manufacturers such as Qualcomm. "No one started bottom-up design from scratch, but we chose to go this way."