It is used to train the neural networks that power the automated guides in Tesla cars and the upcoming self-driving artificial intelligence.
Over the past few years, Tesla has had a clear focus on computing power in and out of its cars.
The company needs computers powerful enough to run its self-driving software indoors. On the other hand, it needs supercomputers abroad to train its self-driving programs backed by neural networks fed by massive amounts of data from the fleet.
CEO Elon Musk has been teasing Tesla’s Dojo project, which apparently consists of a supercomputer capable of executing exaFLOP floating-point operations per second, making it one of the world’s most powerful computers.
Tesla has been working on the Dojo for the past few years, and Musk has been hinting that it should be ready by the end of this year.
But as a result, the company has developed other supercomputers on its way to the Dojo.
Tesla develops supercomputers:
Andrei Karpathi, head of artificial intelligence at Tesla, revealed its latest during a presentation at the 2021 conference on computer vision and pattern recognition.
During the presentation, Karpathi thanked the company’s supercomputing team and showcased their latest work, the third set of Tesla’s supercomputer.
The company presented some rather amazing specifications with this new set, which is supposed to make it the fifth most powerful computer in the world:
- 5760 GPUs.
- 1.8 exaFLOP.
- 10 petabytes of NVME storage.
- 640 terabytes per second of total transfer capacity.
Karpathi commented on the effort: “We have a neural network and a 1.5 petabyte data set that requires a huge amount of computing.” For us, computer vision is the foundation of what we do and what a robotic guide can do.
“We need to train huge neural networks and try a lot,” he added. As a result we have invested a lot in computing. We have a computer with 720 nodes 8x A100 of the 80 GB version. I think it ranks fifth in the world.
On the other hand, the Tesla engineer did not want to clarify the Dojo project. But he said it would be a better supercomputer optimized for neural network training than Tesla’s current group.
In addition, Musk previously said: The company plans to make its supercomputers available to other companies. This is so that it can train its neural networks on it.