Tesla’s In-House Supercomputer Uses NVIDIA A100 GPUs For Mind Boggling Performance

Tesla never ceases to amaze us. Initially, it was their electric cars, and now Musk has opened up about Tesla’s one of the most powerful supercomputers.

The computer offers 1.8 exaFLOPs of power, equivalent to 1,800 petaFLOps, or 1.8 quintillion floating-point operations per second. Mind you, this is not the Dojo that Elon Musk teased last year while it was being developed. He also called it a beast and stated that it can process genuinely vast amounts of video data to train its autopilot technology.

small nvidia a100

DOJo will take its sweet time to arrive. On the other hand, Tesla’s senior director of artificial intelligence, Andrej Karpathy, shared an interesting fact about the computer at the 4th  CCVRP 2021; the supercomputing cluster is powered by hundreds of nodes and thousands of A100 machine learning GPUs based on Ampere.

Below are the details of Tesla’s Supercomputer:

The director also called it an Insane supercomputer which Tesla uses to manage the massive databases it requires for refining its autonomous driving technology. The supercomputer has not been benchmarked, but Karpathy believes it might land among the top dogs.

download.jpeg 1 1

The question arises as, why Tesla requires such a tremendous amount of processing power. The answer is that Tesla’s cars record footage from 8 cameras operating at 36 fps. Therefore, the total sums up to about 1 million 10-second video clips bringing the total to a mind-blowing 1.5 petabytes of data. This data is used to train Tesla’s deep neural networks for its autopilot and self-driving technologies. This is where the powerful supercomputer plays its part.

This data also allows Tesla’s engineers to experiment and come up with new solutions.