GPGPU.org points us to this research paper from the University of Melbourne in Australia. Researchers Alex D. Stivala, Peter J. Stuckey, and Anthony I. Wirth first developed “an improved heuristic for tableau-based protein structure and substructure searching using simulated annealing,” and then went on to create a parallel implementation for GPUs.
The GPU implementation achieves up to 34 times speedup over the CPU implementation of tableau-based structure search with simulated annealing, making it one of the fastest available methods. To the best of our knowledge, this is the first application of a GPU to the protein structural search problem.
I think papers like this show us that real discoveries are happening from real applications on GPUs. And thanks to gamers, GPUs are everywhere. The scientific applications that map well to this architecture may be hard to find at first, but that problem is no match for the parallel processing of millions of brilliant minds.