Home
Data
Computing
(Picture credit rating: Shutterstock / Timofeev Vladimir)
Tesla’s in-residence supercomputer has received an further 1,600 GPUs, a 28% originate bigger on the figure quoted a twelve months within the past.
Tesla Engineering Manager Tim Zaman claims this would place the machine seventh on the earth by GPU count.
The machine now capabilities a total of seven,360 Nvidia A100 GPUs, that are built namely for data middle servers, but employ the same architecture because the firm’s high-of-the-line GeForce RTX 30-sequence cards.
Telsa supercomputer upgradeIt’s seemingly Tesla needs your total processing vitality it is going to derive splendid now. The firm is currently working on ‘neural nets’, that are dilapidated to job the sizable portions of video data that the firm’s autos accumulate.
Essentially the most current upgrade is perchance precise the open of Tesla’s high-performance computing (HPC) ambitions.
In June 2020, Elon Musk mentioned “Tesla is increasing a neural web coaching laptop known as Dojo to job essentially sizable portions of video data”, explaining the deliberate machine would influence a performance of over 1 exaFLOPs, which represents one quintillion floating-point operations per 2nd, or 1,000 petaFLOPs.
Performance of over 1 exaFLOPs would place the machine among the many strongest supercomputers worldwide, as simplest a pair of fresh supercomputers fill formally exceeded the exascale barrier, including The Frontier supercomputer at the Oak Ridge National Laboratory in Tennessee, United States.
You furthermore could shall be ready to derive a job building the modern laptop. Musk requested his Twitter followers to “own in thoughts joining our AI or laptop/chip groups if this sounds spicy”.
Dojo would perchance furthermore fair now not be reliant on Nvidia hardware, nonetheless. The deliberate machine is place to be powered by Tesla’s modern D1 Dojo Chip, which the carmaker mentioned would perchance fill specs of as much as 362 TFLOPs at its AI Day occasion.
Must urge your luxuriate in AI compare within the cloud? Have faith a fill a look at our e-book to the correct cloud hostingVia Tom’s Hardware (opens in modern tab)
Will McCurdy has been writing about technology for over 5 years. He has a wide collection of specialities including cybersecurity, fintech, cryptocurrencies, blockchain, cloud computing, payments, man made intelligence, retail technology, and enterprise capital investment. He has beforehand written for AltFi, FStech, Retail Systems, and National Technology Data and is an experienced podcast and webinar host, as well to an avid long-create characteristic author.