SC10 Kicks Off Today With a Splash

If you are headed to New Orleans for SC10, be sure to bring your coat and be ready for a few splashes as the rain is coming down this morning. One big splash that kicked off the week yesterday was the unveiling of the new Top500 list. It is now official, according to the
latest Top500 rankings, Tokyo Tech’s TSUBAME2.0 system is the world’s 4th fastest supercomputer. Some highlights of the system include:

  • 2.4 PF multi-core Intel Westmere x86 CPU + Fermi GPGPU
  • 1432 HP SL390s nodes
  • 4224 NVIDIA Tesla (Fermi) M2050 GPUs
  • 1.9 million “CUDA cores”, 32K x 4K = 130 million CUDA threads(!)
  • 0.72 Petabyte/s aggregate mem BW,
  • Effective 0.3-0.5 Bytes/Flop, restrained memory capacity (100TB)
  • Voltaire Optical Dual-Rail IB-QDR BW, full bisection BW(Fat Tree)
    200Tbits/s, Likely fastest in the world, still scalable

  • Flash/node, ~200TB (1PB in future), 660GB/s I/O BW
  • 7 PB IB attached HDDs, 15PB Total HFS incl. LTO tape
  • Low power & efficient cooling, comparable to TSUBAME 1.0 (~1MW); PUE = 1.28 (60% better c.f. TSUBAME1)
  • Virtualization and Dynamic Provisioning of Windows HPC + Linux, job migration

    Even more impressive is that only a few weeks after being installed, TSUBAME2.0 is not only making a splash on the Top500 list, but making a splash with real research. Here is a simulation of a milk splash completed on TSUBAME2.0. Don’t drop your milk at breakfast!

    OK, that short video may not get across the real scientific value of a system a powerful as TSUBAME2.0, so here is one more, a real-time tsunami simulation (watch your pronunciation, the system is named TSUBAME after the Japanese swallow shown on the universities seal, a tsunami is what is shown in the video below). While supercomputers have been able to model tsunamis for many years, high fidelity real-time modeling has been difficult to achieve. As Tokyo Tech researchers are able to develop more accurate and real-time tsunami models, they will be able to help millions of people avoid the deadly splash of real tsunamis.

    That is all I have time for this morning, check back later for more SC10 news.

    Advertisements
  • About Marc Hamilton

    Marc Hamilton – Vice President, Solutions Architecture and Engineering, NVIDIA. At NVIDIA, the Visual Computing Company, Marc leads the worldwide Solutions Architecture and Engineering team, responsible for working with NVIDIA’s customers and partners to deliver the world’s best end to end solutions for professional visualization and design, high performance computing, and big data analytics. Prior to NVIDIA, Marc worked in the Hyperscale Business Unit within HP’s Enterprise Group where he led the HPC team for the Americas region. Marc spent 16 years at Sun Microsystems in HPC and other sales and marketing executive management roles. Marc also worked at TRW developing HPC applications for the US aerospace and defense industry. He has published a number of technical articles and is the author of the book, “Software Development, Building Reliable Systems”. Marc holds a BS degree in Math and Computer Science from UCLA, an MS degree in Electrical Engineering from USC, and is a graduate of the UCLA Executive Management program.
    This entry was posted in Uncategorized. Bookmark the permalink.