Tesla has unveiled a new supercomputer called the Dojo, which is used to train artificial intelligence.
According to News i and quoting IyaAndre Carpathi, Tesla’s senior director of artificial intelligence, unveiled the electric carmaker’s new supercomputer at the 2021 CVPR Conference on Cognitive Insight and Patterns.
Last year, Ilan Musk announced Tesla’s plans to build a neural network training supercomputer called the Dojo.
For years, the company has been talking about the Dojo supercomputer, which Mask claims will be the fastest supercomputer in the world, and one of the fastest current supercomputers in the world, Fugaku in Japan, with a speed of 415 petaflops. Float per second (FLOPS) works, will work better.
The new supercomputer appears to be a prelude to the Dojo project, which Carpathi claims is currently the world’s fifth largest supercomputer in terms of floating point performance per second.
This supercomputer is not lacking in processing. As Carpathian pointed out in his speech, the supercomputer has 720 nodes of the 80GB “x A1008” (a total of 5760 GPUs). It also has a switching capacity of 640 terabytes per second.
A node is a single computer in a system that has a number of processor cores.
Tesla’s new car is Tesla’s third supercomputer, and it will no doubt teach valuable lessons to the Tesla team before the “Dojo” project is fully realized.
During his presentation, Carpathian said he wanted to test the crazy supercomputer that is currently being built and used. He also said that computer vision is the basis of what Tesla is doing, and that it is what powers the “autopilot” car system of Tesla cars.
The supercomputer will be used to train Tesla’s autopilot system and Tesla’s fully automated artificial intelligence (FSD) on a huge amount of data at extremely high speeds, which Carpathi says requires a lot of computing.
Carpathian did not disclose any information about the Dojo project itself.
Musk also said that Tesla would allow other companies to use the Dojo supercomputer to train their neural networks, which is good news for self-driving technology.