Tesla Now Has the 7th-Largest GPU Supercomputer in the World

Image: @tim_zaman on Twitter

According to Tesla’s AI and Autopilot lead, Tim Zaman (@tim_zaman), the company now operates the seventh-largest graphics supercomputer in the world by GPU count.

Zaman said in a Friday tweet that Tesla recently upgraded its GPU supercomputer. It now comprises a whopping 7,360 of Nvidia’s A100 Tensor Core GPUs.

Each of these enterprise-level graphics processors from Nvidia features 80GB of graphics memory. The A100 also boasts the world’s fastest memory bandwidth at over 2 terabytes per second.

For comparison, Nvidia’s highest-end consumer graphics card at the moment — the RTX 3090 Ti — has 24GB of graphics memory running at a bandwidth of just less than 1 terabyte per second.

Tesla CEO Elon Musk said in a recent tweet that the company won’t have to buy nearly as many GPUs next year. Instead, the company will start phasing in its Dojo supercomputer. He also touted Tesla’s ability, as foremost a “car company,” to build supercomputers.

Zaman, who previously worked at Nvidia, announced in the same tweet that Tesla is sponsoring the upcoming Machine Learning & Systems Conference. He added that attendees can visit Tesla’s booth at the event to check out some of the company’s hardware and explore opportunities.

Tesla is operating at the cutting edge of AI. Musk has said on several occasions that the company’s work on Autopilot and Full Self-Driving essentially equates to “solving real-world AI.”

That said, Tesla is constantly looking to hire more AI and Machine Learning talent. Tesla’s previously delayed AI Day 2022 event is also slated for September 30. The company is expected to unveil a working prototype of its Optimus humanoid robot at the event.