Sequoia's 96 racks during installation (Photo: Lawrence Livermore National Laboratory)
Using LINPACK benchmarking, Sequoia was found to be approximately 55 percent faster than K, which achieved 10.51 petaflop/s to Sequoia's 16.32 petaflop/s (or 16,320,000,000,000,000 flop/s). Standing for floating point operations per second, flop/s are a more sophisticated measure of computer performance than instructions per second, representative of the scientific calculations such supercomputers are likely to perform.
Impressively, Sequoia is reportedly one of the most energy-efficient computers on the list, consuming 7.9 MW of power to K's 12.7 MW. This gives Sequoia an efficiency of approximately 2.07 teraflop/s/W to K's 0.83 teraflop/s/W. Lower power consumption is central to IBM's ongoing Blue Gene project.
The June 2012 top ten looks like this:
1. IBM: Sequoia (DOE Lawrence Livermore National Laboratory - USA)
2. Fujitsu: K (RIKEN Advanced Institute for Computational Science - Japan)
3. IBM: Mira (DOE/SC/Argonne National Laboratory - USA)
4. IBM: SuperMUC (Leibniz Rechenzentrum - Germany)
5. NUDT: Tianhe-1A (National Supercomputing Center in Tianjin - China)
6. Cray: Jaguar (DOE/SC/Oak Ridge National Laboratory - USA)
7. IBM: Fermi (CINECA - Italy)
8. IBM: JuQUEEN (Forschungszentrum Juelich - Germany)
9. Bull: Curie thin nodes (CEA/TGCC-GENCI - France)
10. Dawning: Nebulae (National Supercomputing Centre in Shenzhen - China)
Perhaps the most notable characteristic of the top ten is that IBM supercomputers make up half the list. This marks the first occasion a US computer has topped the list since November 2009.
Sequoia is used by the Department of Energy's National Nuclear Security Administration to help assess the USA's nuclear deterrent.
Source: TOP500
No comments:
Post a Comment