Elon Musk Assures Tesla's Full Self-Driving Improvements Will Soon Overcome Compute Constraints, Says NVIDIA and Tesla

Tesla Inc CEO Elon Musk Says Version 12 of Full Self-Driving Software Still Training, But Compute Training Will Not Slow It Down

Tesla Inc CEO Elon Musk Says Version 12 of Full Self-Driving Software Still Training, But Compute Training Will Not Slow It Down

Introduction

Tesla Inc CEO Elon Musk announced on Wednesday that version 12 of its full self-driving (FSD) software is still in the training phase. However, he assured that compute training will not slow down the process.

What Happened

Musk responded to a user who calculated that the training speed of version 12 should at least quadruple once chip-maker Nvidia Corp's H100 cluster is fully operational. Musk also mentioned that the training speed could increase even further if there is a lot of Dojo online.

Tesla Activates Nvidia H100 GPU Cluster

This week, Tesla activated the 10,000-unit Nvidia H100 GPU cluster. Musk stated that this cluster is three times faster than the A100 in the company's tests.

Difficulties in Bringing the H100 Online

Musk acknowledged the challenges faced in bringing the 10,000-unit H100 cluster online. He compared the experience to bringing their 16,000-unit A100 cluster online. Musk explained that uptime and performance are initially low but improve with extensive work by Tesla and Nvidia.

Why It Matters

Musk reiterated that Tesla will continue to rely on both Nvidia and Dojo to process a large amount of data from its growing fleet of vehicles and train its FSD software. Dojo is Tesla's custom-built supercomputer platform designed specifically for AI machine learning purposes, with a focus on video training using data obtained from its fleet of vehicles.

Conclusion

Version 12 of Tesla's full self-driving software is still undergoing training, but the compute training will not slow down the process. Tesla activated the Nvidia H100 GPU cluster, which is significantly faster than the A100. Despite the challenges faced in bringing the H100 cluster online, Musk remains confident in the continuous improvement of uptime and performance. Tesla's dependence on both Nvidia and Dojo highlights the company's commitment to processing a large amount of data and training its FSD software.

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

https://megazineusa.com/computing/elon-musk-assures-teslas-full-self-driving-improvements-will-soon-overcome-compute-constraints-says-nvidia-and-tesla/?feed_id=88339&_unique_id=64f047fcf154c

Comments

Popular posts from this blog

Revolutionize Your Command Line Experience with AI Assistant's Seamless Translation Capabilities

FTX Founder Bankman-Fried's Trial Sends Shockwaves Through Cryptocurrency Market: What's Next for the Industry?

Blockchain Expert Reveals Game-Changing Reasons for Holding XRP - Prepare to Be Amazed!