NVIDIA GTC 2022 Keynote

This article is part of our GTC 2022 coverage.

NVIDIA's GPU Technology Conference (GTC) kicked off with CEO Jensen Huang's presentation. It runs over 90 minutes, but it shows off the latest technology and capabilities of NVIDIA GPU technology. You can skip around the video if you want to check out specific technologies. For example, you can jump ahead to learn about the latest Tensor core GPGPU, the H100, codenamed "Hopper."

00:17:10 Pretrained models, SDKs and tools

00:17:30 - Triton: hyperscale model inference server
00:18:30 - Riva - GPU accelerated speech support with pretrained models
00:20:11 - Maxine: SDK for building virtual collaboration and content
00:21:50 - Merlin: SDK for building recommender systems, Merlin 1.0 released at GTC 2022

00:20:11 - Nemo Megatron: open-source framework for developers to build and train state-of-the-art conversational AI models

00:27:17 - H100: latest Tensor core GPGPU, codenamed "Hopper," HBM3 memory

00:32:30 - SXM super chip module
00:34:00 - DGX H100 pod, built around eight SXM modules
00:39:10 - H100 CNX - H100 PCI Express card, direct from network to GPGPU

00:40:00 - Grace: 144 Arm v9 cores using two CPU chips, 500 W with 1 TB of memory
00:45:00 - RAPIDS - data science suite of SDKs
00:56:56 - CuOpt - logistics and route optimization SDK
00:46:48 - Morpheus - deep-learning cybersecurity framework
00:47:48 - cuQuantum - SDK for quantum-computing research
00:47:49 - Aerial: GPU-Accelerated 5G Virtual Radio Access Networks (vRAN)

00:48:37 - Sionna: 6G AI research framework

00:48:30 - Modulus: physics machine-learning models
00:50:59 - Omniverse: 3D simulation, enhanced support for third-party tools

Replicator - tool for developing large, complex virtual environments for simulation
01:00:02 - Avatar: framework for building digital humans, includes real-time Toy Jensen demo

01:04:29 - NVIDIA DRIVE - autonomous automotive platform, DeepMap
01:16:51 - NVIDIA Isaac: robot support