Uploaded August 2022 | Updated September 2026, 3 weeks ago
Keynote 2, Hot Chips 34 (2022), Tuesday, August 23, 2022
Ganesh Venkataramanan, Tesla Motors
This keynote discusses important trends in machine learning (AI) performance needs and power requirements, as driven by applications such as self-driving cars and their need to understand the surrounding environment in real-time. Conventional architectures just cannot keep up with these requirements, so new, highly-integrated architectures are required. The Tesla DOJO system is used as a particular example of the kind of innovative architectures that are being designed to meet these requirements. It packs 25 chips containing a large 2D array of cores into a single "tile" that can pack massive amounts of computation into a small volume.
Keynote 2, Hot Chips 34 (2022), Tuesday, August 23, 2022
Ganesh Venkataramanan, Tesla Motors
This keynote discusses important trends in machine learning (AI) performance needs and power requirements, as driven by applications such as self-driving cars and their need to understand the surrounding environment in real-time. Conventional architectures just cannot keep up with these requirements, so new, highly-integrated architectures are required. The Tesla DOJO system is used as a particular example of the kind of innovative architectures that are being designed to meet these requirements. It packs 25 chips containing a large 2D array of cores into a single "tile" that can pack massive amounts of computation into a small volume.
