Stanford University has been running the Folding@Home project for a while. This uses a highly complex mathematical system to see how proteins fold, proteins that fold in a different manner are thought to be part of the process that leads to certain diseases.
NVIDIA GPUs are really highly optimised engines that can work on mathematical data in a very parallel manner (that's really what a lot graphics programming is). They have a programming interface into the GPUs and the NVIDIA folding@home client uses that interface. It can speed up mathematical operations by up to about 140x compared to a standard CPU.
A team running the NVIDIA client has moved to the top 0.1% of the folding@home teams in under a month.
Other mathematical solutions may also speed up drastically using this technology, unfortunately NVIDIA have charged for access to their CUDA libraries.
Showing posts with label CUDA. Show all posts
Showing posts with label CUDA. Show all posts
2008/07/31
Subscribe to:
Posts (Atom)
