November 2000
The 16th TOP500 list was be introduced at the SC2000 Conference in Dallas, TX (04-10, November 2000).
After IBM became the leader in the TOP500, IBM is now able to gain a share of 43% with respect to the number of installed systems as well as installed performance up from 29% only 6 months ago. Second in the number of systems is Sun with 18% while Cray Inc. is holding on to the number 2 spot in performance with 16%.
US DOE ASCI systems are claiming the first four positions of the TOP500. The new IBM ASCI White system at Lawrence Livermore National Laboratory is the new number one with 4.9 TFlop/s Linpack performance. This system is built with 512 nodes, each of which contains 16 IBM Power3 processors using a shared memory. This type of hierarchical architecture is more and more common for systems used in HPC.In addition to the four DOE ASCI systems there are now three more systems exceeding the 1 TFlop/s level on the Linpack benchmark (2 built by IBM and 1 by Hitachi). The TOP500 already contains 18 systems exceeding 1 TFlop/s peak performance. One of these systems is in commercial use. It is an IBM SP system at position 15 installed at Charles Schwab.
The list exhibits an unbroken growth at all ranges of performance. 231 systems fell off the TOP500 since last June due to the increase in the entry level. The last entry of the TOP500 list shows now a performance rate of 55.1 GFlop/s compared to 43.82 GFlop/s 6 months ago and 33.09 GFlop/s in November 1999.
All remaining vector systems on the TOP500 are built by Japanese manufacturers. These are 23 NEC SX system, 17 Fujitsu VPP systems, and 10 Hitachi SR8000 pseudo-vector systems. The NEC SX systems are also the only remaining systems with physically flat shared memory.
The number of "Clusters" built with commodity parts and small SMP nodes with less then 16 processors is on the rise with now 28 such systems listed compared to only 7 one year ago. These systems include clusters of Compaq AlphaServer, Sun HPC systems, the IBM PC based LosLobos cluster, and 7 self-made clusters built with Alpha or Intel based PCs.
Highlights from the Top 10:
- ASCI White is new #1 with 4.9 TF/s on the Linpack
- DOE ASCI systems hold the first 4 positions
- 7 systems have Linpack performance above 1 TFlop/s
- 18 systems have peak performance above 1 TFlop/s including one commercial system (#15 at Charles Schwab)
- 0.89 Tflop/s is the entry point for the Top 10
General highlights from the Top 500:
- 231 systems dropped off the TOP500 since last June This is the second highest replacement rate ever.
- Total accumulated performance is 88.1 TFlop/s compared to 64.2 TFlop/s 6 months ago
- Entry level is now 55.1 GF/s compared to 43.8 GF/s 6 months ago
- The entry point for the top100 moved to 181 GF/s.
TOP 10 Sites for November 2000
For more information about the sites and systems in the list, click on the links or view the complete list.
| Rank | Site | Computer |
|---|---|---|
| 1 | Lawrence Livermore National Laboratory United States | ASCI White, SP Power3 375 MHz IBM |
| 2 | Sandia National Laboratories United States | ASCI Red Intel |
| 3 | Lawrence Livermore National Laboratory United States | ASCI Blue-Pacific SST, IBM SP 604e IBM |
| 4 | Los Alamos National Laboratory United States | ASCI Blue Mountain SGI |
| 5 | Naval Oceanographic Office (NAVOCEANO) United States | SP Power3 375 MHz IBM |
| 6 | National Centers for Environmental Prediction United States | SP Power3 375 MHz IBM |
| 7 | Leibniz Rechenzentrum Germany | SR8000-F1/112 Hitachi |
| 8 | UCSD/San Diego Supercomputer Center United States | SP Power3 375 MHz 8 way IBM |
| 9 | High Energy Accelerator Research Organization /KEK Japan | SR8000-F1/100 Hitachi |
| 10 | Government United States | T3E1200 Cray Inc. |