Imagine the computing power of 512 processors working in tandem. That's the resource now available at the NASA Ames Research Center in Moffett Field, Calif. Based on the Linux operating system, this facility's 512-processor SGI Altix single-system image (SSI) supercomputer is the first of its kind.
This ultra-fast machine has achieved a Linpack Rmax rate of 2.45 TFLOPS and a STREAMS Triad rate of 1.007 Tbytes/s for the fastest performance in the world by both ratings for a shared-memory system. It's also the first supercomputer to break the 1-Tbyte limit on the memory bandwidth benchmark. Additionally, its worst-case communication latency is less than a microsecond, which is important for sustained performance when running all 512 processors.
Initially, the computer is being used to deliver high-resolution ocean analysis as part of the ECCO (Estimating the Circulation and Climate of the Ocean) Consortium. The project seeks to better understand ocean circulation and its impact on global climate patterns. According to Bob Ciotti, an Ames research scientist, "The turnaround time users typically see on large ocean simulations can take months. By dedicating half of the machine to the ECCO project and scaling the code to run efficiently on all the processors, we now expect turnaround to be about two to three days."
Using Intel Itanium 2 processors and a 64-bit Linux operating environment for better price/performance, the SGI Altix SSI system is faster and more efficient than its predecessor, a 1024-processor SGI Origin-based supercomputer.
For details, go to www.arc.nasa.gov.