Tesla Unveils New Dojo Supercomputer So Powerful It Tripped Power Grid Tesla has unveiled the latest version of its Dojo supercomputer, and it’s apparently so powerful it’s tripped Palo Alto’s power grid. The dojo is Tesla’s own dedicated supercomputing platform, built specifically for AI machine learning and video training using video data from its fleet of vehicles.
The automaker already has a large NVIDIA GPU-based supercomputer that is one of the most powerful in the world but uses Dojo’s new custom computer chips and all infrastructure designed by Tesla. The purpose-built supercomputer is expected to boost Tesla’s ability to train neural networks using video data, which is critical for its computer vision technology to enable its autonomous driving.
Last year at Tesla AI Day, the company unveiled its Dojo supercomputer, but the company was still working on its efforts. He only had the first chips and practice tiles and was still working on building a complete dojo cabinet and cluster or “Exapod”.
Now, Tesla unveiled the progress made on the Dojo program over the past year during AI Day 2022 last night. The company confirmed that it has moved from a chip and tile to a system tray and a full cabinet. Tesla claims it can replace 6 GPU units with a single Dojo tile, which costs less than the company claims a single GPU unit. A tray contains 6 such tiles.
Tesla says that one tray is equivalent to “3-4 fully loaded supercomputer racks”. The company integrates its host interface directly into the system tray to create a large full host build. Tesla has released the key specifications of the Dojo Exapod: 1.1 EFLOP, 1.3TB SRAM, and 13TB high-bandwidth DRAM.
The company used the event to try to recruit more talent, but also revealed that it is on schedule to have a full first cluster or Exapod in the first quarter of 2023. There are currently plans to build 7 Dojo Exapods in Palo Alto.
Why does Tesla need a Dojo supercomputer?
It’s a fair question. Why is an automaker building the world’s most powerful supercomputer? Well, Tesla will tell you that it’s not just an automaker, but a technology company developing products to accelerate the transition to a sustainable economy.
Musk said it made sense to offer Dojo as a service, perhaps to take on friend Jeff Bezos’ Amazon AWS and called it “a service that you can use available on the web, where you can train your models quickly and in less time.” “Less money.”
But specifically, Tesla needs Dojo to automatically label training videos from its fleet and train its neural networks to create its own autonomous driving system.
Tesla realized that developing an autonomous driving system using neural network training on millions of videos from its customer fleet required a lot of computing power. And to provide this power it was decided to build their own supercomputer.
This is a short-term goal, but Tesla will use many of the supercomputers in the future as the company has big ambitions to develop other AI programs.