Supercomputers based on Apple’s Xserve technology landed four spots on the newest TOP500 list. The announcement came during the 20th International Supercomputer Conference, held this week in Heidelberg, Germany.
Presented by the Universities of Mannheim and Tennessee and the National Energy Research Scientific Computing (NESRSC) Center, the TOP500 project collects performance benchmarks for the most powerful computing systems on the planet. The TOP500 list is created by comparing the best performance of the “Linpack” benchmark, which tests the system by making it solve a dense system of linear equations. The TOP500 list is generated twice each year.
Four Apple-based supercomputers made the top 200 spots on the TOP500 list this year. The top Apple-based performer (in 14th place) was Virginia Tech’s System X, comprising 1100 dual-processor 2.3GHz Xserve G5 models. The system turned in maximal performance of 12250 gigaflops (billions of floating-point instructions per second), with theoretical peak performance estimated at 20240 gigaflops.
The University of Illinois at Urbana-Champaign (UIUC) was next on the list in position 66 with its Turing Xserve Cluster, made up of 512 dual-processor Xserve G5 systems operating at 2.0GHz. Its maximal performance was measured at 4559 gigaflops, with a theoretical peak performance of 8192 gigaflops.
The University of California Los Angeles’ (UCLA) Plasma Physics Group’s Dawson Xserve cluster, named after deceased UCLA professor and physicist John Dawson, landed in the 162nd spot on the TOP500 list. The Dawson cluster, comprising 256 Apple Xserve G5s — a mix of 2.0 and 2.3GHz systems — turned in maximal performance of 2135 gigaflops and theoretical peak performance of 4403.2 gigaflops.
Bowie State University was not far behind in the 166th spot with its Xseed cluster, a collection of 224 dual-processor Xserve G5s operating at 2.0GHz. Xseed generated maximal performance of 2104 gigaflops, with theoretical peak performance of 3584 gigaflops.
The top system for this roundup is Lawrence Livermore National Laboratories’ BlueGene/L system, an IBM-built supercomputer. BlueGene/L turned in maximal results of 136800 gigaflops, with theoretical peak performance of 183500 gigaflops. Running Linux, BlueGene/L’s is powered by 65,536 PowerPC 440 chips each operating at 700MHz. Six of the top ten systems were built by IBM.