JAKARTA – Electric vehicle and clean energy maker Tesla has announced that it has designed a special chip to train artificial intelligence (AI) networks in data centers.

The D1 chip, announced at the company's live AI Day event, is part of the previously announced Dojo neural network training supercomputer system, which is currently under development. The chip is capable of "processing very large amounts of video data," such as video feeds from Tesla cars. Based on a 7-nanometer manufacturing process and optimized for machine learning workloads, the D1 chip has 362 teraflops of processing power.

The chip is said to be made up of more than 50 billion transistors and has a die size of 645 mm 2. The company says that 25 of these chips are housed on a single "training tile", 120 of which are then combined in multiple server cabinets, resulting in over exaflops of power.

“We will soon be assembling our first cabinet,” Ganesh Venkataramanan, senior director of Autopilot hardware at Tesla told attendees and viewers of the AI Day event. Tesla's technology, the company says, will be the fastest AI training computer.

As for when the Dojo supercomputer will be ready in the near future. "We should have the Dojo operational next year," said Tesla CEO Elon Musk.


The English, Chinese, Japanese, Arabic, and French versions are automatically generated by the AI. So there may still be inaccuracies in translating, please always see Indonesian as our main language. (system supported by DigitalSiber.id)