Over the last decade, GPU computing has had a profound impact on scientific computing. A number of scientists who are using GPUs to accelerate their research have been profiled on NVIDIA’s website. Meet the three that caught our attention:
#1 Unraveling Membrane Proteins with GPUs
Erik Lindahl, professor of Biophysics at Stockholm University, studies proteins in the body, the small molecules that carry blood and comprise the hair and skin. Using the GPU-accelerated GROMACS application, he and his colleagues are learning about how molecules move on an atomic level and a micro-second scale. “GPUs have been a revolution for what we do,” said Lindahl. “Since we started playing around with CUDA roughly four or five years ago, at the time it was a small side effect. Now 90 percent of our computational resources are GPUs.” The research has important implications for understanding and treating diseases including alcoholism.
#2 Accelerating Digital Rock Physics with GPUs
James McClure, a computational scientist with Advanced Research Computing at Virginia Tech, explains how his team uses the DOE’s Titan supercomputer with the K20 GPU to model and visualize how fluids move below the surface of the earth. As McClure describes, first the team obtains images of geologic materials using synchrotron light sources. The result is a three-dimensional image showing the microstructure of rock or granular media. These images are used to simulate flow processes and other transport processes. Th research benefits any application that relies on a resolving these processes, most notably the extraction of oil and gas and ground water.
#3 Simulating Reionization of the Universe
Dr. Paul Shapiro, professor at University of Texas at Austin, studies cosmology, the origin and the evolution of the universe. “We are living in a very exciting time where we are pushing the boundaries back in time to what we can directly observe in order to guide the observes and predict what they will see and interpret it to learn about the origin and evolution of the universe as we can see clues at that period.” To do this requires gigantic simulations on the world’s largest supercomputers. In this video, Professor Shapiro details how his team modeled the way the first stars and galaxies were born to affect the future generation of galaxies and stars. The new code they created to do this was run on the Tesla-accelerated Titan Supercomputer at Oak Ridge National Laboratory. Watch to find out more about how his team used GPUs to balance mismatched time scales between the way they track radiation and ionization versus the way they follow matter.