NASA collaborates with Intel and SGI on forthcoming petaflops super computers

NASA, Intel and SGI have announced in their "Space Act Agreement" plans to reach 10 petaflops. Although they do not expect their HPC "Pleiades" system to reach this target until 2012, they intend to have an intermediate system with a peak performance of 1 petaflops – presumably the Linpack Rpeak – ready as early as 2009.

Advertisement

The Pleiades machines will be put to work in NASA's advanced supercomputing center in the Ames Research Center at the Moffett Field military airport in California, not far from Silicon Valley. Since 2004, this facility has been running a Silicon Graphics Columbia system equipped with 14,366 Itanium cores that has been involved in projects such as carrying out calculations for space shuttle missions.

NASA installed two smaller HPC clusters in 2007, the Schirra with 320 Power5+ processors (640 cores) from IBM and another SGI cluster named RTJones with 1024 quad-core Xeons from Intel.

Intel has recently enjoyed a string of successes with its supercomputer clusters: 354 systems in the latest TOP 500 list are equipped with mainly two- and four-core Xeons, but also with Itaniums. A few months ago, Intel also signed an agreement with Cray, Bull and SGI on plans to install new systems in France and Germany – presumably with Nehalem processors.

(trk)