NVIDIA SC20 Special Address

TL;DR

Nvidia discusses their advancements in supercomputing and AI, including the launch of the DGX A100 system with 80GB memory and the seventh generation of Nvidia Mellanox InfiniBand with NDR 400Gbps speed.

Transcript

do do hello everyone i'm mark hamilton and i'm excited to be here today as the host for nvidia's special address at sc20 this is a different sc20 with it being fully virtual but it's nevertheless very exciting we have three from individually today that are going to talk and explain how nvidia is powering the change in scientific computing it really... Read More

Key Insights

• 👨‍🔬 Nvidia is leveraging supercomputing and AI to tackle significant challenges in drug discovery and Covid-19 research.
• ❓ The DGX A100 system with 80GB memory provides improved performance for HPC, AI training, and inference workloads.
• 💨 The seventh generation of Mellanox InfiniBand, NDR 400Gbps, offers faster data throughput, in-network computing, and improved scalability for supercomputing.
• 🪛 Nvidia's advancements in supercomputing and AI are driving innovation in various industries, including healthcare, climate modeling, and astrophysics.

Questions & Answers

Q: How is Nvidia using AI to support drug discovery and Covid-19 research?

Nvidia is applying computational chemistry and AI algorithms to search for potential drug compounds, narrow down candidates, and study genomics and molecular structures for drug development. They are also using AI to analyze medical data, accelerate simulations, and improve the discovery and treatment of diseases like Covid-19.

Q: What are the key features of the DGX A100 system?

The DGX A100 system has a memory capacity of 80GB, offering enhanced performance for various applications, including high-performance computing, AI training, and inference. It also incorporates the new tensor core technology, providing high teraflops of performance in AI training, inference, and HPC workloads.

Q: How does the seventh generation of Mellanox InfiniBand, NDR 400Gbps, improve supercomputing?

NDR 400Gbps delivers twice the data throughput of previous generations, enabling faster and more efficient data delivery. It also includes advanced in-network computing engines and scalability, offering up to 1.6 petabits per second bi-directional bandwidth on 2000 network ports. These advancements enhance overall supercomputing performance and enable new possibilities in AI processing.

Q: Can the DGX A100 system be upgraded with the new 80GB GPUs?

No, the DGX A100 system is not compatible with the previous generation GPUs. The DGX A100 system has a redesigned architecture and refrigeration system, which allows for the integration of the new 80GB GPUs. However, the previous generation DGX systems are still valuable and offer high-performance capabilities with their V100 GPUs.

Summary & Key Takeaways

• Nvidia showcases their efforts in using supercomputing and AI to address significant challenges, such as drug discovery and Covid-19 research.

• Computational chemistry and drug discovery benefit greatly from computation and AI, helping narrow down potential drug candidates and understand their genomics and molecular structure.

• Nvidia highlights their work in Covid-19 research, including simulating the Covid molecule to understand its effects and the application of AI in drug discovery and treatment.

• Nvidia introduces the DGX A100 system with 80GB memory, offering improved performance for high-performance computing, AI training, and inference.

• The seventh generation of Nvidia Mellanox InfiniBand, NDR 400Gbps, is announced, delivering faster data throughput, in-network computing, and scalability for supercomputing.