To further its Full Self-Driving (FSD) technology, Tesla has now activated a large new system. Sawyer Merritt, an investor and Tesla fanatic, walked over some of the system’s finer points on X.com for the benefit of readers.
Tesla’s neural networks, which are designed to address AI-based challenges, are trained using Graphics Processing Units (GPUs).
In order to educate Tesla’s AI on how to drive a car safely and autonomously, the company uses massive volumes of film for training purposes. NVIDIA, a market leader in GPUs, just announced a new chip—the H100—that’s a big improvement in terms of performance over its predecessor, the A100.
The H100, however, is not inexpensive. The latest GPU from NVIDIA aimed at training costs around $30,000, and Tesla needs many more than that.
Ten thousand H100 NVIDIA GPUs will fuel Tesla’s new training cluster, driving up the price by tens of millions of dollars.
NVIDIA is unable to supply Tesla and the industry with enough H100 units due to the enormous demand for this cutting-edge processor.
Tesla’s Commitment to FSD: Investing $2 Billion in Computing Power
This is why Tesla is spending over $1 billion on its own supercomputer project, codenamed Dojo. To prepare Tesla’s neural networks for use in FSD and Optimus, Dojo has developed a specialized processor.
Elon Musk, CEO of Tesla, has said that the company may not have needed to develop Dojo if NVIDIA had been able to provide sufficient CPUs. Tesla, though, is in a position where it has to take matters into its own hands.
Merritt emphasized the difficulty posed by the high amount of computing power needed to train Tesla’s FSD technology. But Tesla is working hard to fix this problem.
In 2023 and 2024, the corporation will invest over $2 billion to expand its computing capacity, as Musk recently revealed. This expenditure demonstrates Tesla’s dedication to being a leader in automotive industry innovation and the development of FSD technology.
You can also read about the technology facts by clicking the links given below:
- Tesla Truck Release Date: Here is Everything You Need to Know About the Cybertruck
- Google Enhances Workflow Efficiency with Integrated eSignature Support for Google Docs and Drive
Tesla’s First-of-a-Kind Data Centers
Besides working on its own supercomputer, Tesla is also creating what it calls “1st of its kind Data Centers.” The difficulty of this endeavor is illustrated by a recent Austin, Texas, job posting for a Senior Engineering Program Manager at Tesla’s Giga Factory.
The construction of these data centers is crucial to Tesla’s overarching goals. When processing and analyzing massive volumes of video data for its self-driving software, the company’s need for powerful computing resources is critical. Their dedication to this field is demonstrated by the existence of the Dojo supercomputer.
If you have the theimportantenews bookmarked on your browser, you will always have prompt access to all of the most recent news and other material.