Appendix 5 – GTC 2023 Keynote with NVIDIA

GTC 2023 Keynote with NVIDIA CEO Jensen Huang - 22 Mar 2023

NVIDIA cuLitho library running on GPUs enables ASML, TSMC and Synopsys to perform computational lithography and create new chip photo reticles 40x faster with wafer features at 2nm and smaller. The H100 GPU requires 89 reticles that previously each took 2 weeks to create. TSMC currently uses 40,000 CPU servers to do this work, and can now do this work with 500 DGX H100 systems and reduce the power consumption from 35MW to 5MW.

Over the past 10 years, cloud computing has grown into a $1T global industry relying on around 30 million CPU servers to do this work. The power consumption of this infrastructure is enormous. New chip architectures are needed that provide higher performance while also consuming less power in order to reduce carbon emissions while also increasing global datacenter infrastructure. Datacenter industry requires accelerating computing architectures such as those designed and built by NVIDIA.

NVIDIA Grace CPU Superchip is 1.3x faster than the latest Intel x86 processors with 60% of the power.

“DGX has become the essential instrument of AI” – Jensen Huang – GTC 2023 Keynote

The GPU of DGX is eight H100 modules.

H100 has a Transformer Engine designed to process models like the amazing ChatGPT – Generative Pretrained Transformers

“Generative AI is a new kind of computer – one that we program in human language. This ability has profound implications. Everyone can direct a computer to solve problems. This was a domain only for computer programmers. Now everyone is a programmer” – Jensen Huang – GTC 2023 Keynote

“AI is at an inflection point as Generative AI has started a new wave of opportunities, driving a step-function increase in inference workloads. AI can now generate diverse data, spanning voice, text, image video, and 3D graphics to proteins and chemicals.” – Jensen Huang – GTC 2023 Keynote