Santa Clara (CA) - Intel and supercomputers is a combination of words we have been hearing quite a bit lately. Today we learned that the company has secured another contract with NASA, expanding on existing relationship that began with the deployment of the Columbia supercomputer in 2004. The new "Pleiades" system is planned to deliver 1 PFlops of computing power by 2009 and 10 PFlops by 2012.
Under the terms of a Space Act Agreement, NASA said that will work closely with Intel and SGI to increase computational capabilities for modeling and simulation at the NASA Advanced Supercomputing (NAS) facility at NASA’s Silicon Valley-based Ames Research Center at Moffett Field. As NASA’s Columbia supercomputer, currently listed at #20 in the list of the world’s 500 fastest supercomputers, is past its prime time, the new Pleiades is aimed to deliver the computing capability of new technologies and take NASA back to the top of the list.
Few details about Pleiades have been released, but NASA said that it targets a peak performance of 1 PFlops or 16x the performance of Columbia, which stands at about 61 TFlops. A 2012 is expected to result in a ten-fold increase in performance to about 10 PFlops.
This dramatic increase in performance raises questions how Intel and SGI will be able to deliver this speed, especially if we heard just yesterday that a 10 PFlops system would currently cost about $1 billion to construct. It seems as this system could be another project for Intel that takes advantage of Larrabee accelerator nodes.
Columbia went into operation in 2004 and is based on SGI Altix architecture with 10,240 Itanium 2 (Madison-9M core with 9 MB of L3 cache) processors, with the majority running at 1.5 GHz and some running at 1.6 GHz. Each Itanium 2 processor is rated at a maximum performance of 6.4 GFlops - which is only about 10% of what Xeon quad-core processors are estimated to be capable of today.
Last week, Intel said that it will be developing supercomputers in collaboration with Cray.
