Software enables massively parallel computation

Software enables massively parallel computation

Complex computational problems in consumer, business, and technical applications can take an eternity on a standard PC, but a software tool called CUDA enables a tremendous performance improvement by operating on large quantities of data in parallel. The technology gives computationally intensive applications access to the huge processing power of the graphics processing unit using a new programming interface coupled with the standard C language.

CUDA programmers use “C” to create programs called kernels that use many threads to operate on large quantities of data in parallel. Multiple GPU cards can be easily used, offering incredible performance boosts. Not all mathematical problems can benefit from this type of capability, but those that can see a tremendous speed up.

One important innovation offered by the software/hardware is the ability for threads to communicate when solving a problem. NVIDIA GPUs with CUDA technology have an on-chip parallel data cache, allowing computing threads to quickly share information.

The CUDA Toolkit includes standard FFT and BLAS libraries, a C-compiler for the NVIDIA GPU, a runtime driver, and many examples with source code.

NVIDIA , Santa Clara , CA
Sales 408-486-2000
http://www.nvidia.com