ATI's Stream Computing uses GPUs to solve problems
updated 01:45 pm EDT, Fri September 29, 2006
ATI Stream Computing
Many computers have fast graphics processors, but when these chips aren't being used for 3D images, much of their potential power goes unused. ATI sees these free processor cycles as an opportunity to accomplish more; the graphics card maker today announced that it has developed a concept called Stream Computing to use these spare resources for better purposes. Since most recent graphics cards are generalized and can do more than just render images, Stream Computing can dynamically assign calculations to recent ATI cards, such as the Radeon X1950, that would normally be handled by the often busy main system processor. The result is that some operations can be dramatically accelerated, especially when they are well-optimized to the massively parallel nature of video hardware. An upcoming version of Stanford University's Folding@Home distributed computing software will calculate as much as 40 times faster when Stream Computing is being used, according to ATI. Physics in games and other simulations will also stand to benefit. ATI ultimately has plans to allow systems with multiple graphics cards to devote even more resources to these tasks and could optionally split processing work equally between graphics and general problem-solving.
