![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
Top 10 supercomputers
Nov. 21st, 2012 01:07 pm1. The Titan is a Cray (USA) XK7 powerhouse, producing 17.59 petaflops of performance using 560,640 AMD Opteron processor cores and 261,632 Nvidia K20x accelerators. It operates at the Oak Ridge National Laboratory (USA).
2. The first million-core system, Lawrence Livermore National Laboratories(USA) Sequoia was the top dog in the last Top500 list. It cranks out a whopping 16.32 petaflops with its 1,572,864 processor cores. Sequoia is the fourth and last IBM Blue Gene/Q system on the latest list.
3. Dropping one place into third from the last list is the Fujitsu K Computer, at Japan's RIKEN Advanced Institute for Computational Sciences (Japan). Using 705,024 Oracle (USA) SPARC64 cores, it produced a Linpack score of 10.51 petaflops.
4. Also using the IBM Blue Gene/Q architecture is Mira, of the Department of Energy's Argonne National Laboratories (USA). However, it packs substantially more cores than JUQUEEN -- 786,432, to be exact -- in return for a nearly doubled performance return of 8.16 petaflops.
5. Juelich-based JUQUEEN eclipsed its German rival SuperMUC to capture the fifth spot on the November list, posting a 4.14-petaflop mark on the Linpack test. Unlike SuperMUC, it's powered by a 393,216-core IBM (USA) Blue Gene/Q system.
6. A hardy perennial of the Top500 list, SuperMUC is based at the Leibniz Supercomputing Center near Munich (Germany). Clocking in at 2.89 petaflops, it's powered by 147,456 Intel (USA) Sandy Bridge processors.
7. New to this top 10 is Dell's Stampede, which rode its new Intel Xeon Phi processors -- a total of 204,900 cores' worth -- to a 2.66 petaflop benchmark. Installed at the University of Texas in Austin (USA), Stampede also packs 112,500 accelerator cores as part of the Intel Xeon Phi platform.
8. The only Chinese entry into this top 10, Tianhe-1A turned in a 2.56 petaflop performance mark, on the strength of its 186,368 Intel (USA) Xeon processor cores. It's also the first machine on this list to use co-processors for additional performance -- 100,352 Nvidia (USA) 2050 cores, to be precise.
9. Based at Italy's CINECA joint venture, Fermi is the first IBM (USA) Blue Gene/Q-based system on our list, clocking in at 1.72 petaflops driven by 163,840 IBM PowerPC cores.
10. IBM (USA) Development Engineering DARPA Trial Subset - this IBM Power7-driven entry clocked in at 1.5 petaflops on Top500's tests, using just 63,360 total processor cores -- by far the smallest number in the top 10.
2. The first million-core system, Lawrence Livermore National Laboratories(USA) Sequoia was the top dog in the last Top500 list. It cranks out a whopping 16.32 petaflops with its 1,572,864 processor cores. Sequoia is the fourth and last IBM Blue Gene/Q system on the latest list.
3. Dropping one place into third from the last list is the Fujitsu K Computer, at Japan's RIKEN Advanced Institute for Computational Sciences (Japan). Using 705,024 Oracle (USA) SPARC64 cores, it produced a Linpack score of 10.51 petaflops.
4. Also using the IBM Blue Gene/Q architecture is Mira, of the Department of Energy's Argonne National Laboratories (USA). However, it packs substantially more cores than JUQUEEN -- 786,432, to be exact -- in return for a nearly doubled performance return of 8.16 petaflops.
5. Juelich-based JUQUEEN eclipsed its German rival SuperMUC to capture the fifth spot on the November list, posting a 4.14-petaflop mark on the Linpack test. Unlike SuperMUC, it's powered by a 393,216-core IBM (USA) Blue Gene/Q system.
6. A hardy perennial of the Top500 list, SuperMUC is based at the Leibniz Supercomputing Center near Munich (Germany). Clocking in at 2.89 petaflops, it's powered by 147,456 Intel (USA) Sandy Bridge processors.
7. New to this top 10 is Dell's Stampede, which rode its new Intel Xeon Phi processors -- a total of 204,900 cores' worth -- to a 2.66 petaflop benchmark. Installed at the University of Texas in Austin (USA), Stampede also packs 112,500 accelerator cores as part of the Intel Xeon Phi platform.
8. The only Chinese entry into this top 10, Tianhe-1A turned in a 2.56 petaflop performance mark, on the strength of its 186,368 Intel (USA) Xeon processor cores. It's also the first machine on this list to use co-processors for additional performance -- 100,352 Nvidia (USA) 2050 cores, to be precise.
9. Based at Italy's CINECA joint venture, Fermi is the first IBM (USA) Blue Gene/Q-based system on our list, clocking in at 1.72 petaflops driven by 163,840 IBM PowerPC cores.
10. IBM (USA) Development Engineering DARPA Trial Subset - this IBM Power7-driven entry clocked in at 1.5 petaflops on Top500's tests, using just 63,360 total processor cores -- by far the smallest number in the top 10.
