Nvidia teams up with Tokyo Tech for more supercomputery

THE WORLD SEEMS to have gone supercomputer mad and now the Tokyo Institute of Technology (Tokyo Tech) has also jumped on the bandwagon along with Nvidia. A collaboration between the two sees the Green Goblin’s Tesla GPUs put to use bumping up the computational horsepower of Tokyo Tech’s Tsubame supercomputer.

170 Tesla S1070 1U systems were apparently added to Tsubame, giving the machine a major performance boost to the tune of almost 170 TFLOPS of theoretical peak performance and 77.48 TFLOPS of measured Linpack performance. All that extra horsepower has ensured the Tsubame stays well within the top ranks of the Top 500 list.

Nvidia’s Tesla S1070 1U GPU system is based on the firm’s much touted Cuda architecture based on an industry standard C language programming environment which, the firm claims, lets developers and researchers tap into the parallel architecture of the GPU quickly and easily.

Satoshi Matsuoka, division director of the Global Scientific Information and Computing Center at Tokyo Tech noted “it had become clear to us that to make the next major leap in performance, Tsubame had to adopt GPU computing technologies”. He added that “Tesla GPUs delivered speed-ups that we had never seen before, sometimes even orders of magnitude”.

Apparently upgrading the whale of a computer took only one week, and the Tsubame stayed live and operational throughout, something quite unprecedented in top-level supercomputing.

Tokyo Tech is also not alone in starting to look to GPUs for supercomputer, with universities and supercomputing centres like the National Center of Supercomputing Applications (NCSA) at the University of Illinois, Rice University, University of Heidelberg, University of Maryland, Max Planck Institute and University of North Carolina all signing up too. µ