GPUs Could Speed Scientific Discovery

Scientific Computing reports that the advent of GPU parallel processing and heterogeneous systems like Tianhe-1A could speed the advance of science.

Such hybrids will probably take over in the future. “If we want bigger or faster machines, this is what the future looks like: undoubtedly multicore and almost undoubtedly heterogeneous multicore,” says Bronson Messer, acting director of science at the Oak Ridge Leadership Computing Facility at Oak Ridge National Laboratory.

According to Messer, the types of problems that could benefit most involve high-energy physics. “It takes every sort of physics you can think of and all at once,” Messer says. “To get a precise element synthesis, we knew we’d need to ramp up the computational intensity.”