Tesla Inc (NASDAQ:TSLA) CEO Elon Musk on Wednesday said that version 12 of its full self-driving (FSD) software is still training but compute training will not slow it down.
What Happened: “Training compute should soon not be so much of a limiting factor,” Musk wrote. He was responding to a user who calculated that the training speed of V12 should at least quadruple once chip-maker Nvidia Corp‘s (NASDAQ:NVDA) H100 cluster is fully operational or even further if there’s a lot of Dojo online.
Tesla activated the 10,000-unit Nvidia H100 GPU cluster this week, which Musk said is three times faster than the A100 in the company’s tests.
However, talking about the difficulties of bringing the H100 online, Musk said, “Very difficult bringing the 10k H100 cluster online btw. Similar experience to bringing our now 16k A100 cluster online.”
“Uptime & performance are low at first, then improve with lots of work by Tesla & Nvidia,” he explained.
Why It Matters: Earlier this week, Musk reiterated that Tesla is going to use both NVIDIA and Dojo to process a large amount of data from Tesla’s growing fleet of vehicles and train its FSD software.
Dojo is Tesla’s custom-built supercomputer platform, specifically designed from scratch for AI machine learning purposes, with a particular focus on video training using data obtained from its fleet of vehicles.
Photo courtesy: Thomas Hawk on Flickr
Check out more of Benzinga’s Future Of Mobility coverage by following this link.
Related Link: Tesla To Mark 10 Years Of Superchargers In Europe With Electric Summer Events
