Resource for Macromolecular Modeling and Visualization

Summary This administrative supplement application seeks to secure funds to acquire a multi-node, multi-GPU compute system to be used for code development and optimization of NAMD, one of the two major programs developed by the Resource for Macromolecular and Visualization at the University of Illinois Urbana-Champaign. Optimizing the architecture of NAMD for these modern platforms is a major aim of our parent grant (R24-GM145965), primarily since this feature is heavily requested by our users. We note that NAMD is currently used by more than 35,000 users. A large fraction of these users run NAMD using computational facilities at the National supercomputing centers, many of which have/are moving towards GPU-dense machines that take advantage of multi-node, multi-GPU architectures. Therefore, making NAMD compatible with these platforms is of great need and demand and continues to be a major objective of our Resource. In order to achieve this goal, we are in need of a multi-node system, with multiple GPUs or multiple superchips (combined CPU-GPU-memory systems-on-chip) per node, with modern high-speed connectivity at each level of hardware organization. The requested equipment meets these architecture criteria and provides a fast communication model to allow us to optimize NAMD for supercomputing platforms that are already in place at some centers, e.g., TACC.