Early Experiences with Algorithm Optimizations on Clusters of Playstation 3's

A 50 TeraFLOPS cluster of Playstation 3s^reg (PS3) is being built at the Air Force Research Laboratory, Information Directorate (AFRL/RI) in Rome, NY as an High Performance Computing Modernization Program (HPCMP) effort to provide early access to the IBM Cell Broadband engine chip technology included in the low priced commodity gaming consoles. A heterogeneous cluster with powerful subcluster headnodes is used to balance the architecture in light of PS3 memory and input/output constraints. Several cases studies have shown that 30-70% of the 153 GFLOPS peak performance of these nodes can be sustained on problems of Department of Defense (DoD) interest. Notable power efficiency approaching 1 GFLOPS/Watt has also been observed.