electrek:

Tesla is reportedly ramping up its partnership with chip giant TSMC to produce the Dojo supercomputer chip.

Dojo is Tesla’s own custom supercomputer platform built from the ground up for AI machine learning and, more specifically, for video training using the video data coming from its fleet of vehicles.

The automaker already has several large NVIDIA GPU-based supercomputer clusters, which are some of the most powerful in the world, but the new Dojo custom-built computer uses chips and an entire infrastructure designed by Tesla.

The custom-built supercomputer is expected to elevate Tesla’s capacity to train neural nets using video data, which is critical to the computer vision technology powering its self-driving effort.

At Tesla’s AI Day in 2021, the company unveiled its Dojo supercomputer, but the company was still ramping up its effort at the time. It only had its first chip and training tiles, and it was still working on building a full Dojo cabinet and cluster, or “Exapod.”

A year later, at AI Day 2022, Tesla unveiled some progress on Dojo, including having a full system tray. At the time, the automaker was talking about having a full cluster by Q1 2023.

The first quarter of the year came and went without any news of Dojo being in operation.

But we finally learned that Dojo came online this summer with a plan to gradually ramp it up to a 100 Exa-flop capacity by the end of 2024.

Now a new report coming from China claims that Tesla is now ramping up its partnership with TSMC to produce the D1 Dojo chip (translated Chinese):

Tesla is sprinting towards the layout of its supercomputer “Dojo”. It is rumored that it will expand its cooperation with TSMC. Its supercomputer chip “D1” is produced using TSMC’s 7nm family process combined with advanced packaging. Next year, the volume of wafers produced by TSMC is expected to be higher. This year it has doubled to 10,000 pieces, and orders will continue to increase in 2025, fueling TSMC’s high-speed computing (HPC)-related order momentum.

The D1 chip is the basis of the Dojo supercomputer. It gets put into tiles, then trays, which are added together to create a Dojo supercomputer cluster.

Recently, there has been more attention brought to Dojo after a note from Morgan Stanley describing a potential future where the supercomputer is significantly contributing to the bottom line.

Source link

MuskWire TLDR:

Tesla is increasing its collaboration with chip giant TSMC to produce the Dojo supercomputer chip. Dojo is Tesla’s custom-built supercomputer platform designed for AI machine learning and video training using data from its vehicle fleet. While Tesla already has powerful NVIDIA GPU-based supercomputer clusters, the Dojo supercomputer utilizes chips and infrastructure developed by Tesla. The custom-built supercomputer will enhance Tesla’s ability to train neural networks with video data, which is crucial for its self-driving technology. After unveiling progress on Dojo at AI Day 2022, Tesla plans to have a full cluster by Q1 2023 and aims to reach a 100 Exa-flop capacity by the end of 2024. A recent report from China suggests that Tesla is expanding its partnership with TSMC to produce the D1 Dojo chip, which serves as the foundation for the Dojo supercomputer. The chip is integrated into tiles and trays to create a Dojo supercomputer cluster. The attention on Dojo has grown, with Morgan Stanley noting its potential to significantly contribute to Tesla’s bottom line.