SGER: Evaluation of the Sony PS3 for projects in computational physics

University Of Massachusetts, Dartmouth, North Dartmouth MA

Investigators

Abstract

This award funds exploring the use of the Sony Playstation 3 (PS3) gaming console for scientific computing in gravitational wave physics. If successful, this approach would yield a significantly higher cost effectiveness (measured as performance per dollar) compared with traditional Beowulf compute clusters. The project will focus on investigating the performance of the PS3 for projects in numerical relativity and LIGO data analysis. Part of the project is dedicated to identifying the specific codes and libraries that will need tuning for the PS3/Cell architecture in order to perform well. Any Cell processor specific optimizations that are implemented in publicly available source code as a result of this study, will be made freely available via a website. This project will explore the feasibility of a novel extension of the cost-effective Beowulf cluster approach towards supercomputing, using very low-cost special purpose hardware, such as the Sony PS3. If successful, this approach suggests that it may be possible to reach a performance per dollar that is over an order of magnitude higher than what can be achieved through traditional means. Thus, this project will clearly have serious implications for cluster computing and scientific computing, in general. In particular, if PS3 clusters yield these benefits, searches of LIGO data and simulations of gravitational wave sources could move into computer intensive regimes not now possible.