Electronic Design
What Kind Of Super Computer Can You Build With 26 Amps?

What Kind Of Super Computer Can You Build With 26 Amps?

This year's Super Computing Conference, in New Orleans, LA, is host to the SC10 Student Cluster Competition where eight teams of undergraduates are seeing how well their supercomputers can take advantage of only 26A. Teams actually get two 110V 13A PDUs (power distribution units).

The teams have received some heavy duty support from a range of companies well known in the computer and high performance computing (HPC) environment. The teams and their sponsors include:

  • National Tsing Hua University (Hsinchu, Taiwan) + Acer Incorporated, Tatung Company, and National Center for High-Performance Computing (NCHC)
  • Nizhni Novgorod State University (Nizhni Novgorod, Russia) + IBM, Microsoft, nVidia, and PGI
  • Florida A&M University (Florida, USA) + Atlantic Computer, LLC (HP Partner for Higher Education)
  • Louisiana State University (Louisiana, USA) + HP and LATG, Mellanox, PGI, Adaptive Computing
  • University of Colorado (Colorado, USA) + Dell, AMD, Mellanox, and FusionIO through the HPC Advisory Council
  • The University of Texas at Austin (Texas, USA) + Dell
  • Purdue University (Indiana, USA) + HP and AMD
  • Stony Brook University (New York, USA) + Cray, Inc.

The top eight teams were chosen from a larger field in a preliminary competition. Each team will be running the following applications over a period of about two days.

  • FLASH: adaptive mesh hydrodynamics code for modeling astrophysical thermonuclear flash
  • NAMD: a parallel molecular dynamics code designed for high-performance simulation of large biomolecular systems
  • Weather Research and Forecasting (WRF) Model: a next-generation mesocale numerical weather prediction system designed to serve both operational forecasting and atmospheric research
  • Password Recovery

All of the applications lend themselves to parallelization. The challenge is to process the most data accurately within the given constraints. Laptops and graphical front ends for monitor the system are not part of the limited power budget.

The students will be taking advantage of existing software and had access to the applications prior to the final competition. The applications utilize standard mathmatical and message-passing libraries. Students where able to work with sponsors and supervisors prior to the competition but are on their own with the final data.

Students configure their own systems but typically their sponsors provide most of the hardware. The University of Texas teams had access to the TACC (Texas Advanced Computing Center). Dell is their sponsor and supplied the team with ninem two-socket, hex core Intel Westmere systems, 108 cores total. The total system memory is 432 Gbytes. A pretty hefty complement of computing power.

The Nizhni Novgorod State University from Russia and the Stony Brook University team from New York included GPUs in their mix. The Russian team's five node system has two NVidia GPUs per node. Stony Brook uses one NVivia GPU per node. They both take advange of NVidia's CUDA. If these teams can get the applications to take advantage of the GPUs then it could provide them with a significant advantage.

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.