Sequoia, an IBM BlueGene/Q Supercomputer System using 1,572,864 processor cores, scored 16.32 petaflop/s on the Linpack Benchmark, the yardstick used to measure such things.
The incumbent top dog, Fujitsu’s “K Computer” in Kobe, Japan, came in second with a 10.51 Pflop/s score using 705,024 SPARC64 processing cores. Another IBM BlueGene/Q system called Mira, which scored 8.15 petaflop/s using 786,432 cores, was ranked third.
China, which had the top-ranked supercomputer in its Tianhe-1A supercomputer two years ago, now has two systems in the top ten.
The IBM systems led the list accounting for 213 or 42.6 percent of the top 500, followed by Hewlett-Packard with 138 systems (27.6 percent). Cray was next with 5.4 percent. Appro, SGI and Bull rounded out the top vendors on the list.
Some other fun facts: Fifty-eight of the top 500 use accelerators or co-processors, up from 39 six month ago. Fifty-three of those supercomputers use NVIDIA chips. Two of the others use Cell or ATI Radeon processors and the other uses Intel MIC technology.
Nearly three-quarters of the top 500 supercomputers (74.4 percent) run Intel chips, down from 384 (or 76.8 percent). Sixty three systems (12.6 percent) run AMD Opteron chips. And, the number of supercomputers running IBM Power processors share grew to 58 up from 49 last year.
Related research and analysis from GigaOM Pro:Subscriber content.