Tesla Engineering Manager Tim Zaman revealed earlier that Tesla is a supercomputer built with NVIDIA A100 GPUs and incorporating its own self-developed artificial intelligence chip D1.Dojo (Dojo) has been upgraded to use 7360 A100 GPUs and 80GB HBM 2 high-density memory, which is a 28% increase compared to the version released last year. At the same time, in terms of GPU computing power performance, it has ranked 7th among supercomputers in the world.
However, Tim Zaman did not disclose the specific computing power performance of the upgraded supercomputer. But if we take the "Perlmutter" supercomputer used by the National Energy Research Scientific Computing Center (NERSC) in the United States as an example, the computing power performance can reach 6144 Petaflops when using 100 sets of A70.87 GPUs. Tesla's new supercomputer will obviously have higher computing power.
If combined with Tesla's self-developed artificial intelligence chip D1, it may be able to greatly assist Tesla vehicles in operating their autonomous driving assistance systems more accurately.
Tesla is sponsoring the @MLSysConf, come visit our booth for opportunities on the AI team and see our hardware.
We have recently upgraded our GPU supercomputer (photo) to 7360 A-100(80GB) GPUs, making it Top-7 by gpu-count. Reach out to build #1: https://t.co/mPiOOb8CJg pic.twitter.com/VN5orhbtLy— Tim Zaman (@tim_zaman) August 12, 2022
Tesla has already developed its Autopilot autonomous driving technology through an in-vehicle system equipped with an artificial intelligence computing chip. However, although it is advertised as fully autonomous driving and can automatically change lanes, drive on highways, and include automatic parking and automatic summoning functions, users still need to keep their hands on the steering wheel and continue to pay attention to the road conditions ahead. Therefore, it cannot actually be called a fully autonomous driving design, but rather an advanced driver assistance function.
In addition to the automatic driving assistance function applied to electric vehicles, Tesla has also decided to investhumanoid robotDevelopment requires more powerful supercomputers to support sufficient deep learning training and various artificial intelligence application needs.
