Tesla Reveals Specs of its New AI-Powered Full Self-Driving Computer

resource from: anandtech.com   

In April, at a special event at Tesla's Palo Alto, California headquarters called Tesla Autonomy Investor Day, Tesla CEO Elon Musk announced that Tesla vehicles are using a new custom-designed processor to power its Autopilot full self-driving (FSD) system.

At the time Musk said that no chip was available that had the processing power and power constraints that Tesla required, so the automaker built its own from scratch. Now the technical details of the new chip have been revealed for the first time.

At the Hot Chips conference in San Francisco on Tuesday, Tesla's VP of hardware engineering Pete Bannon revealed of the details of the chipset that will power Tesla's future Autopilot, full self-driving (FSD) system. Bannon said that the new AI-powered chip is 21 times faster that the Nvidia chip it's replacing and only 80% of the cost.

The details were first reported by Dr. Ian Cutress, Senior Editor of AnadTech, who was live blogging from the event.

Tesla said it needed more robust hardware to achieve its 2019 goal of "full self-driving" which includes not only driving on highways, but extends to driving on local roads with traffic lights and road signs to obey.

The new AI processor has 6 billion transistors.

"It was clear to us, in order to meet our performance levels at the power constraints and the form factor constraints we had, we had to design something of our own," said Ganesh Venkataramanan, Senior Director Autopilot Hardware at Tesla, who also spoke at Tuesday's event. Venkataramanan previously worked for AMD and is one of the chip's designers.

The new chips is already being installed in Tesla models. All Tesla model produced after March 2019 have been shipping the new FSD chip, including the Model 3.

Tesla Had Specific Performance Targets for the New Chip

Tesla has specific performance targets for its new chip. First, it had to consume less than 100 watts of power, it needed to be retrofitted into existing Tesla models, have built-in security, and it had to have full redundancy, so if any part of the chip failed the chip would still be able to power the Autopilot FSD system. In addition, it also needed a redundant power supply and overlapping camera field with redundant paths.

Each Tesla computer has two independant AI chips which allow for redundancy for an extra layer of safety. This redundancy even extends to the power supply further guarding against a power interruption. For the vehicle's forward facing cameras each on has its own power supply to guard against failures.

Tesla designed the new chip to offer more than 50 TeraOPS (tera operations per second) of performance for its AI-powered deep neural nets used for autonomous driving.

"There are a lot of redundancy features, which makes sure nothing untoward happens to the system" said Venkataramanan.

Each of the AI chips can perform 36 trillion operations per second, giving the entire processor 72 TFLOPS of processing capability at 2GHz. The chip also has its own dual neural network accelerators (NNA). Each one is 96x96 MACs and can perform 36.8 TOPs per NNA.

The hardware is designed for redundancy, so that each chip makes its own assessment of what the car should do next. The computer compares the two independent assessments and the car only takes the action and if the chips agree. If the chips disagree, the car just discards the frame of video data and tries again, Venkataramanan explained. 

Venkataramanan said that even if the two chips do not agree, the framerate is so high that if one frame is discarded, the overall system performance is not affected. 

Tesla vehicles do not rely on lidar, so the cameras are a necessary component for self-driving. That's one of the reasons Tesla wanted powerful AI chips that could handle such a high frame rate for video. 

Tesla said the chip took 14 months to design, and the processor being manufactured by Samsung.

resource from: anandtech.com

Eric Walz

Originally hailing from New Jersey, Eric is a automotive & technology reporter covering the high-tech industry here in Silicon Valley. He has over 15 years of automotive experience and a bachelors degree in computer science. These skills, combined with technical writing and news reporting, allows him to fully understand and identify new and innovative technologies in the auto industry and beyond. He has worked at Uber on self-driving cars and as a technical writer, helping people to understand and work with technology.