The video card isn't just for gaming anymore. Distributed computing pioneer Folding@Home announced today that it will release a client early next week that can take advantage of Radeon X1900 and X1950 class ATI GPUs to perform protein folding calculations on PCs. Folding@Home recently announced a client that will work on Sony’s PlayStation 3 console designed specifically to take advantage of the system’s multicore Cell processor.
Adding GPU processing power is expected to increase Folding@Home performance anywhere from 10 to 100 times compared to the current client that only uses the CPU, according to Folding@Home director and Stanford University associate professor of Chemistry Vijay Pande.
The Folding@Home scientific research project aims to increase our understanding of how proteins fold, with the hopes of applying the resulting work to cure diseases such as Alzheimer's, Parkinson's, and Huntington's. A protein consists of a long chain of amino acids that literally folds upon itself to perform vital functions. Properly folded proteins function within the body to enable many different kinds of processes. Genetic mutations can cause amino acid substitutions within a protein, which then cause it to fold improperly, rendering the protein either useless or actively harmful to the body.
Predicting how a protein folds requires immense amounts of computing power. Distributed computing gives scientists studying protein folding access to an enormous amount of idle computational power available within the general population. Each computer receives a tiny piece of the project to work on via the Internet. The PC sends the completed work back to Folding@Home servers and receives a new assignment.
According to Pande, Folding@Home currently has access to roughly 1.8 million CPUs, of which 200,000 are active, giving the group the equivalent of approximately 200 Teraflops of sustained computational power. In comparison, the largest supercomputer in the world, the US Department of Energy's Blue Gene supercomputer, has 131,000 processors and 280 teraflops (367 teraflops peak) of computing power. Now that Folding@Home has figured out how to use GPUs, Pande estimates that the power of its distributed network could rise to 1 to 10 petaflops.
While not as hardware intensive as common 3D benchmarking programs, Pande admits that running both your CPUs and GPUs at full throttle for extended periods can still generate a significant amount of heat. Folding@Home advises home client users to keep the computer case free of dust and to make sure all the fans are in working order.
The project expects to expand GPU support to the older Radeon X1800 series, but Nvidia GPU support is less certain due to "technical" difficulties.