NVIDIA GTC 2019 Keynote
1:09:55 “It turned out we have a ton of announcements and I just wanted to put all of the announcements into one slide. So the first announcement is that we now automatically, automatically do the mixed precision necessary for our Tensor Core Architecture, which is Nvidia’s new AI architecture for GPUs.” Share
Jensen Huang presents NVIDIA’s GTC 2019 keynote, moving from RTX ray tracing, Omniverse collaboration and GeForce Now to AI inference, Rapids data science, Mellanox networking and AWS. He introduces Jetson Nano and Isaac for robotics, then details NVIDIA Drive, Constellation simulation and a new Toyota partnership for autonomous vehicles.
Passages
6:11 “But accelerated computing is a collaboration, it’s a co design, it’s a continuous optimization between the architecture, the chips, the systems, the algorithms and the applications. The Nvidia stack looks basically like this.” Share
44:13 “The Omniverse basically connects up all of the designers and studios. It works with every tool.” Share
1:15:54 “And for conversational type models, it is very likely that latency will ultimately define the quality of service and therefore we have to process it incredibly fast. Here’s the challenge.” Share
Search what Huang said from Claude, ChatGPT or your own code. See plans