记得 25th 的时候，还是上着并行计算的课程，觉得很有意思，却学不懂。有意思是因为像是拼积木，学
26th Edition of TOP500 List of World's Fastest Supercomputers Released: DOE/LLNL BlueGene/L and IBM gain Top Positions
Nov 14, 2005, 07:09
MANNHEIM, Germany; KNOXVILLE, Tenn.; & BERKELEY, Calif.; In what has become a closely watched event in the world of high-performance computing, the 26th edition of the TOP500 list of the world's fastest supercomputers was released today (November 14, 2005) at the Supercomputing Conference (SC05) in Seattle, WA.
The new TOP500 list, as well as the previous 25 lists, can be found on the Web at http://www.top500.org/.
The No. 1 position was again claimed by the BlueGene/L System, a joint development of IBM and DOE's National Nuclear Security Administration (NNSA) and installed at DOE's Lawrence Livermore National Laboratory in Livermore, Calif. BlueGene/L also occupied the No. 1 position on the last two TOP500 lists. However, the system was doubled in size during the last six months and reached a new record Linpack benchmark performance of 280.6 TFlop/s ("teraflops" or trillions of calculations per second). No other system has yet exceeded the level of 100 TFlop/s and this system is expected to remain the No. 1 Supercomputer in the world for the next few editions of the TOP500 list.
The pace of innovation and performance improvements seen at the very high end of scientific computing shows no sign of slowing down. This time, four of the TOP10 systems on the June 2005 TOP500 list were displaced by newly installed systems, and the last 221 systems on the list from June 2005 are now too small to be included.
The new No.3 system, also installed at LLNL, is the ASCI Purple system, built by IBM and based on their p575 server. It reached 63.4 TFlop/s.
Two new systems at DOE's Sandia National Laboratories captured the No. 5 and 6 spots, with a Dell PowerEdge-based system slightly outperforming Cray's Red Storm system. The NEC-built Earth Simulator, which has a Linpack benchmark performance of 35.86 TFlop/s and had held the No. 1 position for five consecutive TOP500 lists before being replaced by BlueGene/L last November, has slipped within one year to No. 7.
The No. 10 spot was captured by a Cray's XT3 system at DOE's Oak Ridge National Laboratory with 20.53 Tflop/s. This is also the new entry level for the TOP10, up from just under 10 TFlop/s Linpack performance one year ago.
IBM has now established itself as the dominant vendor of supercomputers with almost half of the list (43.8 percent) carrying its label. Also, five of the TOP10 systems are from IBM. Hewlett-Packard (HP) remains unchallenged at the second position in this survey with 33.8 percent of all systems.
Intel microprocessors are at the heart of two-thirds (333) of all 500 systems. Intel's new EM64T-based processors are very successful in the high performance computing (HPC) market place, with 81 systems using them already. AMD's Opteron processors are also steadily gaining ground, now with 55 systems using them compared to only 25 systems six months ago.
The U.S. is clearly the leading consumer of HPC systems with 305 of the 500 systems installed there, while the European (100 systems) and Asian share (66 systems) is slowly decreasing over time.
Here are some highlights from the newest Top 500:
Only systems exceeding the 1.64 TFlop/s mark on the Linpack benchmark were qualified to make the list this time, compared to 850.6 GFlop/s one year ago. The last system on the latest list would have been listed at position 173 just one year ago.
The entry level for the TOP10 exceeds 20 TFlop/s and the entry point for the top 100 moved from 2.026 TFlop/s one year ago to 3.98 TFlop/s.
Total combined performance of all 500 systems on the list is now 2.30 PFlop/s (petaflops or thousand teraflops), compared to 1.127 PFlop/s one year ago.
Other trends of interest:
A total of 333 systems are now using Intel processors, with 81 one these are already using the new EM64T processors. The second most-commonly used processors are the IBM Power processors (73 systems), ahead of AMD Opteron processors (55).
There are 360 systems now labeled as clusters, making this the most common architecture in the TOP500. Of these, 249 cluster systems are connected using Gigabit Ethernet and 70 system using Myricom's Myrinet.
At present, IBM and Hewlett-Packard sell the bulk of systems at all performance levels of the TOP500. IBM remains the clear leader in the TOP500 list with 43.8 percent of systems and 52.8 percent of installed performance. HP is second with 33.8 percent of systems and 18.8 percent of performance. No other manufacturer is able to capture more than 7 percent in any category.
The U.S. is clearly the leading consumer of HPC systems with 305 of the 500 systems installed there (up from 267 one year ago). The European (100 systems) and Asian share (66 systems) is slowly decreasing.
Dominant countries in Asia are Japan with 21 systems (down from 23 six months ago) and China with 17 systems (down from 19 six months ago).
In Europe, Germany (24 systems) lost the No.1 spot again to UK again (41 systems). Six months ago Germany was in the lead with 40, compared to UK's 32 systems.
The TOP500 list is compiled by Hans Meuer of the University of Mannheim, Germany; Jack Dongarra of the University of Tennessee, Knoxville; and Erich Strohmaier and Horst Simon of NERSC/Lawrence Berkeley National Laboratory.