Modern graphics processors contain tens of millions transistors, while high-end graphics cards cost up to $500 nowadays. The only thing such expensive products can do is render 3D games providing entertainment and relaxation to end-users. Some think that the muscles of today’s GPUs can be used more wisely and they seem to be right – theoretical power of visual processing units is far beyond the fastest CPUs these days. Seems like there is a way to adopt GPU for general purpose task!

A group of developers from Stanford University released their BrookGPU – a compiler and runtime system that provides an easy, C-like programming environment for today’s GPUs. According to this report from Slashdot, a shader program running on the NVIDIA GeForce FX 5900 Ultra GPU achieves over 20GFLOPS, roughly equivalent to a 10GHz Intel Pentium 4. A real powerhouse that is available now, has some issues with programming, as GPU programming is not a trivial task. BrookGPU adds simple data parallel language additions to C which allow programmers to specify certain parts of their code to run on the GPU. The compiler and runtime takes care of the rest. To obtain more information on the matter, turn to this or this page.

This is probably not the first experience with trying to utilize graphics processors for general-purpose tasks. In the past there were quite a lot of initiatives of this kind, but none of them appeared to be popular or useful.

In general, for a few years now there has been a confrontation underway between GPU and CPU developers. The former bring more and more brute computing power and may serve as CPU for certain tasks, the latter provide us some more techniques for creating high-quality 3D graphics in real-time. Moreover, some CPU makers, such as Intel Corporation, attack GPU companies by offering inexpensive chipsets with integrated graphics core. More than a half of graphics solutions in the market are integrated into chipsets at this time, according to market researchers.