Stanford EE Computer Systems Colloquium
Wednesday, June 7, 2023 430-6 PM Pacific
EE380 Summer3 Quarter Ends. This is the last Colloquium of the 2022-2023 Academic Year as EE380 will not be offered this Summer. I am trying to arrange a regular weekly Open Forum Colloquium throughout the Summer. I’ll let everyone know what’s happening and when.
Curator’s note The focus for this Wednesday’s talk will be the hardware that enables the amazing things we have been discussing in the Large Language Model generative AI systems. In this last Spring meeting we will focus on the hardware that is needed to enable generativeAI. A primary supplier is Nvidia Corporation. We’ll focus on their A100 platform which is “affordable”. Perhaps we’ll also explore the follow-on procssor, the Hopper. Thanks to Bill Dlly and the Nvidia team.
NVIDIA’s H100 GPU
Jack Choquette, Nvidia
Overview of key features of the H100 GPU, and how they help accelerate HPC, AI and data analytics datacenter workloads. Notable new features include a Fourth-Generation Tensor Core, new Tensor Memory Accelerator Unit, a new CUDA Cluster capability, and HBM3 DRAM.
Jack Choquette is a Senior Distinguished Engineer at NVIDIA, where he has led the architecture development of NVIDIA's GPGPU Streaming Microprocessors for multiple generations. Jack has been leading CPU and system designs for over 25 years.
This Colloquium will be conducted with an Open Q&A Forum and the result posted to YouTube. If you want to participate in the question session, please sign a release form so we can incorporate your image and voice in the colloquium. Specify EE380 Spring 2023 as the date and you won’t need to do it again. We prefer to include both sound and video. If you do not want to appear on video, please turn off your camera. Please keep your microphone muted if you are not speaking. https://stanforduniversity.qualtrics.com/jfe/form/SV_6tWexlcSD39HFCB
To receive an invitation to join the presentation, CLICK HERE.