This is our first online industrial talk.

Registration: Free, Deadline 8 March 2021; get the Zoom link by registering here.

The recording is available on our YouTube channel - part 1 is here and part 2 is here. The slides are available: part 1 and part 2.

Title: GPUs for biomolecular researchers: an overview of accelerated computing, plus a deep dive into GROMACS on GPU

Speakers: Paul Graham, Senior Solutions Architect; Alan Gray, Senior Developer Technology Engineer

Abstract: Graphics processing units (GPUs) have now become an essential tool in the researcher’s toolbox, reducing time to insight for traditional simulations, but also enabling new approaches to science via the application of artificial intelligence.

In the first part of this talk will cover the basics of GPU architecture, in particular some of the features of modern GPUs which make them suited for HPC simulation and deep learning. Some of the accelerated applications and frameworks for researchers to easily exploit the hardware will be introduced. For those wishing to develop their own applications, some of the different ways for programming for GPUs will be covered, including recent advancements using standard ISO C++ and ISO Fortran directly. An overview of resources available to academics to assist in making the best use of GPUs will conclude this section.

The second part will focus on GROMACS, describing the large performance improvements available in recent versions, for both single and multi-GPU runs, through a long term collaboration between NVIDIA and the core GROMACS developers. Future directions will also be discussed. More information on this is available in the blog article Creating Faster Molecular Dynamics Simulations with GROMACS 2020.