Pages

November 09, 2006

GPU supercomputers

Graphic chips like the Nvidia G8800 have about 520GFLOPS of performance. Some have been adapted for more general processing using the C language and are being banded together for supercomputers with 2-40 times the performance of regular CPUs.

Wired magazine talks about an array of 536 GPUs significantly outperformed some 17,485 CPUs from Linux boxes. The 536 GPUs producing 35 trillion calculations per second compared to 21 trillion calculations per second for the CPUs.

The UNC Chapel Hill Gamma Research Team under laboratory-type conditions put an Nvidia 7900 GTX GPU up against two different leading-edge optimized CPU-based implementations running on high-end, dual-3.6-GHz Intel Xeon processors or dual AMD Opteron 280 processors. The research team, which included Manocha, Naga K. Govindaraju and Scott Larsen from UNC and Jim Gray from Microsoft Research, put these systems through three fairly standard numeric-based computational algorithms, including sorting, FFT (fast Fourier transform) and matrix multiplications.

The results they recorded show that the GPU performed at anywhere from two to five times the speed of the CPU-based systems on these specific applications.

The co-lead of the Gamma group, Ming C. Lin, is leading the development of many new GPU-based technologies for physics simulation -- including collision detection, motion planning and deformable simulations -- with speeds in many cases increasing 10 to 20 times beyond previous methods.

SiCortex, a startup is also producing 5.8 teraflops for about 1-1.5 million

2 comments:

Dezakin said...

Application specific! You don't get that kind of performance doing general application work.

bw said...

Supercomputers often need to have the programs that run on them to be highly optimized to take advantage of the speed. $1 million to get 100 teraflops of hardware for running special jobs and then $1 million for the 1 year for a team of Phds to tune your application. That would still be worthwhile. You will probably be running some really big simulation or grand challenge problem. It would bring the technology to reasonably well funded universities.

Plus like the graphic applications, one could program them to solve a useful common class of problem.