technology is productivity
innovation for any company
are the most powerful driving force
(ChinaIT.com News) According to foreign media sources, Elon Reeve Musk (Elon Reeve Musk) wants to buy all AI chips from Nvidia, but can’t. So Musk started his own independent and controllable way to invest in GPU.
So Musk is willing to invest over $1 billion to develop the Dojo supercomputer for Tesla simply because he can’t get enough Nvidia chips. According to Musk himself, “If they could give us enough GPUs, we might not need Dojo. But they can’t because they have too many customers.”
Afterwards, Musk made further comments in response to a Twitter account: “Unfortunately, they can’t even provide us with a fraction of the computation required!”
The D1 supercomputing chip is the basis of the Dojo system. The chip uses a 7-nanometer manufacturing process and has a processing capacity of 1024 gigaflops, or 102.4 billion operations per second. 1,500 D1 chips can form an array, and 25 arrays can be placed on a wafer to form a training module (Training Tile), which is the unit core of the Dojo supercomputer.
At the Tesla AI Day in 2022, Tesla threw out the Dojo ExaPod, the Dojo cluster. ExaPod contains 120 training modules (Training Tile) and 3000 D1 chips. It has 13TB static random access memory capacity and 1.3TB high transmission bandwidth memory capacity, and its computing power is as high as 1.1EFLOP.
Musk said, “By the end of next year, we will spend far more than $1 billion on the Dojo project, and we have an astonishing amount of video data to train.” Musk also mentioned that it is expected to use both NVIDIA GPU and Dojo supercomputing systems.
Dojo is Tesla’s self-developed Multi-Chip Modularized supercomputer, which will be officially unveiled at Tesla AI Day in 2021. At first, Dojo mainly served the data labeling and training of the automatic driving system, and later it was also applied to the research and development of robots. The “brain” of Optimus Prime was equipped with Dojo’s D1 supercomputing chip.
Tesla has a “staggering amount” of image data, and has now driven a total of 300 million miles in its Full Self-Driving (FSD) beta. In the future, Dojo will be used to process the massive data required to develop autopilot software, which will help Tesla get rid of its dependence on Nvidia GPU. In addition, Musk also said that Tesla is considering licensing its FSD hardware and software to other automakers.