Hybrid Computing for Scientific Simulations
From biology and chemistry, to high-energy and nuclear physics, scientists have come to rely on accurate models of physical phenomena to understand the world we live in. When time and costs are a factor, scientific simulation is often used to supplement real world experiments. But this is 2017 and the A.I. renaissance of massively parallel processing, utilizing GPUs, combined with Machine Learning algorithms is altering the scientific landscape
The accuracy of these simulations is dependent on the number of calculations performed. We will demonstrate the power of modern Compute clusters which leverage emerging technologies like NVLink and High Bandwidth Memory to enable new capabilities for A.I. assisted HPC (High Performance Computing.) In this presentation, Rob Farber and Greg Scantlen will discuss the key advantages of hybrid computing and breakthrough methodologies in science and machine learning with the Simul8 HPC Environment.
We will learn about how hybrid computing is optimizing materials science (chemoinformatics), high energy and nuclear physics, biology, and other scientific disciplines.
Rob Farber was a staff scientist in the theoretical division at Los Alamos where he did basic research that established the machine and deep learning technology now used in the Internet and self-driving cars. He has been on staff at Berkeley, and other organizations around the world. Mr. Farber co-founded two startups with successful exits, one of which was a computational drug discovery company that utilized machine learning. Rob travels the world and is globally published. His latest book, Parallel Programming with OpenACC is available in English and will soon be available in Chinese as well. He works closely with major semiconductor companies including: NVIDIA, Intel, IBM, and ARM.