News
RapidMind releases “Performance Evaluation of GPUs Using the RapidMind Multi-core Development Platform”
Waterloo, Ontario - October 27, 2006 - RapidMind today posted the results of a collaboration with HP High Performance Computing Division to evaluate the performance of GPUs using the RapidMind Multi-core Development Platform.
In conjunction with an updated release of the v2.0 RapidMind Multi-core Development Platform Beta, RapidMind has released a new paper demonstrating the performance potential of GPUs and showing up to 30 times increase in performance over CPUs. The high-performance processors in video accelerators, GPUs, can be used as numerical co-processors in a variety of applications. The RapidMind Multi-core Development Platform is a software development platform that allows the developer to use standard C++ programming to easily create applications that run on the GPU. Using the RapidMind platform, this study compares the performance of implementations of the FFT, BLAS dense linear algebra operations, and quasi-Monte Carlo option pricing on the GPU against highly tuned CPU implementations on the fastest available CPUs. The advantages and limitations of GPU acceleration are discussed as well as techniques for optimizing performance.
The paper is a preview of what HP and RapidMind will be presenting together as part of the Poster Session at SC06. The Poster reception will take place on Tuesday, Nov. 14, from 5:15-7:15PM and researchers from both HP and RapidMind will be on hand to discuss the results. An executive summary and the full paper are available for download from the RapidMind web site.