Armed with the graphics card in your laptop and a few programming tools, you can traverse the leading edge of computational science.
Graphics processing units were developed to render computer graphics at high speed to meet the demands of display technologies and consumer applications such as realistic video games and movies. But it turns out that the speed and power of GPUs as parallel processors can be harnessed for general-purpose computing for science and other data-intensive applications, with significant potential savings in cost and time.
This potential is behind the formation of the SciGPU community. In 2008, the National Science Foundation funded a three-way collaborative project proposed by scientists associated with the Harvard Initiative in Innovative Computing. That project, part of NSF’s program on Cyber-Enabled Discovery and Innovation (CDI), focuses on algorithms and architectures suited to applying GPU programming to three different challenges, drawn from radioastronomy, chemistry and neuroscience. This site and the SciGPU seminar series are managed by the CDI postdoctoral fellows to help disseminate knowledge gained through this research.
A major partner is NVIDIA, a leading manufacturer of GPUs, which has designated Harvard as a CUDA Center of Excellence. (CUDA is NVIDIA’s environment for general-purpose GPU programming.) Boston University is a founding partner in the CCoE.
