NVIDIA GPUS SPEED VIRGINIA TECH RESEARCH

by Linda Barney

Virginia Tech is conducting research studying biomolecular functions to determine biological properties of macromolecules. The research is designed to understand the mechanisms behind molecular biological functions such as catalytic activity or ligand binding.

According to Wu Feng, Virginia Tech Assistant Professor and researcher, “Computation on large biomolecules can take days on a typical desktop computer. And biologists have little desire to learn how to program a job submission script and submit it to a supercomputing resource.”

Virginia Tech began using PCs equipped with NVIDIA GPUs based on the CUDA architecture in their research and found a number of benefits.  Feng states, “Research that used to take half a day on a supercomputer now takes us mere seconds on PCs running NVIDIA.”

In a demo at SC09 on a Supermicro 1U server using Tesla GPUs, Feng demonstrated an astonishing speedup (over 4 orders of magnitude) on a GPU, compared to a serial computation on a CPU.  Feng states, “A change is needed in how research is done. The world is now using parallel computing and we need the tools that allow us to do our research in a parallel, real-time mode.”