Two life science computers are among many new arrivals in the latest version of the twice-annual ranking of the world’s 500 fastest supercomputers, released at the International Supercomputing Conference this week in Dresden, Germany.
To be eligible for the current Top500 list, systems needed to reach 4.005 teraflops on the Linpack benchmark, compared to 2.737 teraflops for the last version, released in November [BioInform 11-17-07].
This boost in performance meant that more than half the computers on the November list were ineligible for the current ranking — the “largest turnover rate between lists” in the project’s 15-year history, according to a statement released by its organizers, Hans Meuer of the University of Mannheim; Erich Strohmaier and Horst Simon of NERSC/Lawrence Berkeley National Laboratory; and Jack Dongarra of the University of Tennessee.
Among the newcomers are a 2,208-processor, 15.6-teraflop Dell system at Stanford University’s Biomedical Computational Facility dubbed “Bio-X2,” which claimed the No. 54 spot, and a 2,016-processer, 4.4-teraflop IBM BladeCenter system at Cold Spring Harbor Laboratory, which landed the No. 454 spot.
One life science system on the last Top500 list, a 3.1-teraflop Hewlett-Packard cluster at India’s Institute of Genomics and Integrative Biology, was among the 284 computers that did not make the 4.005 teraflop cutoff for the current version.
As a result, the total number of life science systems on the list increased from four to five over the last six months (see table 1, below, for details).
For the fourth straight time, the No. 1 spot was held by a 280.6-teraflop BlueGene/L system developed by IBM and the Department of Energy’s National Nuclear Security Administration and installed at DOE’s Lawrence Livermore National Laboratory.
Six of the top 10 supercomputers in the current list are from IBM and four of those are BlueGene systems, including two new machines: one at the New York Center for Computational Science in Stony Brook, NY, which is ranked No. 5, and another at the Computational Center for Nanotechnology Innovations at Rensselaer Polytechnic Institute in Troy, NY, which is ranked No. 7.
Despite IBM’s dominance of the highest-ranking systems, HP regained its lead over IBM as the manufacturer with the most systems on the list. HP’s system count grew to 202, a 40.6-percent increase over the November list. By comparison, IBM’s count, at 192, declined 38.4 percent from the November listing (see table 2, below, for details).
IBM remains ahead of HP in terms of total performance, however, with 41.9 percent of installed performance, or 2.06 petaflops, compared to HP’s 24.5 percent, or 1.19 petaflops.
After dipping slightly in November, cluster architectures are gaining ground again and remain the leading type of system on the list, with 373, or nearly 74.6 percent of all systems (see table 3, below, for details).
Likewise, Linux is holding steady as the operating system of choice, moving to 389 systems, or 77.8 percent of the total list, from 376 systems six months ago. By comparison, Unix usage fell from 86 systems in the November list to 60 in the current version.
“Mixed” operating systems grew 31 percent to 42 systems from 32 six months ago. BSD, Mac OS, and Windows rounded out the list with four, three, and two systems, respectively.
Intel regained its dominance among chip manufacturers, with 289 systems, or 57.8 percent, of the Top500 systems using Intel processors, up slightly from 261 systems six months ago. This is largely the result of a surge in adoption of the company’s Woodcrest dual-core chip, which grew to 205 systems from 31 six months ago (see table 4, below, for details).
The second most-commonly used processors are the AMD Opteron family, which lost 7 percent of its standing since November and now account for 21 percent of total systems, or 113 units. Of those, 90 use Opteron dual core processors, up from 75 six months ago.
IBM Power processors are in 85 systems, down from 93 systems six months ago.
The complete Top500 list is available here.
Table 1: Top-Ranking Life Science Supercomputers, June 2007
|
||||||
Rank June 2007
|
Rank
Nov. 2006 |
Installation Site
|
Manufacturer/
Computer |
Number
of Processors |
Tflop/s (max)
|
Year installed
|
37
|
21
|
Computational Biology Research Center, AIST (Japan) | IBM/"Blue Protein" eServer Blue Gene Solution |
8,192
|
18.7
|
2005
|
54
|
—
|
Stanford University Biomedical Computational Facility | Dell/"Bio-X2" PowerEdge 1950 (2.33 GHz) |
2,208
|
15.6
|
2007
|
188
|
100*
|
Merck & Co. | IBM/BladeCenter LS20 (2.2 GHz Opteron dual core) |
2,688
|
6.6
|
2006
|
320
|
75†
|
Eli Lilly and Co. | IBM/BladeCenter HS21 Cluster (3.0 GHz Xeon dual core) |
1,024
|
5.1
|
2007
|
454
|
—
|
Cold Spring Harbor Laboratory | IBM/BladeCenter LS20 (2.0 GHz Opteron dual core) |
2,016
|
4.4
|
2006
|
* Identical system listed as "undisclosed US biotech" in the Nov. 2006 Top500 list
† Different configuration in Nov. 2006 list: 1,072 processors, 8.6 Tflop/s |
Table 2: Top500 Supercomputer Manufacturer Ranking, June 2007
|
||||||
Manufacturer
|
Rank June 2007
|
Rank Nov. 2006
|
Change in Rank
|
Count† June 2007
|
Count† Nov. 2006
|
Change in Count†
|
Hewlett-Packard | 1 | 2 | +1 | 202 | 158 | +44 |
IBM | 2 | 1 | -1 | 192 | 236 | -44 |
Dell | 3 | 4 | +1 | 23 | 18 | +5 |
SGI | 4 | 3 | -1 | 19 | 20 | -1 |
Cray Inc. | 5 | 5 | — | 11 | 15 | -4 |
Linux Networx | 6 | 7 | +1 | 8 | 7 | +1 |
Sun Microsystems | 7 | 6 | -1 | 7 | 9 | -2 |
Hitachi | 8 | 8 | — | 6 | 6 | — |
Self-made | 9 | 10 | +1 | 5 | 5 | — |
Fujitsu | 10* | 9 | -1 | 4 | 5 | -1 |
NEC | 10* | 12 | +2 | 4 | 3 | +1 |
Atipa Technology | 10* | 11 | +1 | 4 | 4 | — |
Intel | 11* | 14* | +3 | 2 | 1 | +1 |
Appro International | 11* | 13 | +2 | 2 | 2 | — |
Bull SA | 11* | 14* | +3 | 2 | 1 | +1 |
NEC/Sun | 11* | 14* | +3 | 2 | 1 | +1 |
lenovo | 12* | 14* | +2 | 1 | 1 | — |
California Digital Corporation | 12* | 14* | +2 | 1 | 1 | — |
Dawning | 12* | 14* | +2 | 1 | 1 | — |
Hitachi/Fujitsu | 12* | 14* | +2 | 1 | 1 | — |
DALCO AG Switzerland | 12* | 14* | +2 | 1 | 1 | — |
Megware | 12* | — | — | 1 | 0 | +1 |
T-Platforms | 12* | — | — | 1 | 0 | +1 |
* Tie
† Number of installed systems in the Top500 |
Table 3: Top500 Architecture Ranking, June 2007
|
||||||
Computer Architecture
|
Count† June 2007
|
Count† Nov. 2006
|
Change
|
|||
Cluster (Beowulf, NOW, etc.) |
373 | 361 | +12 | |||
MPP (homogeneous architecture) |
107 | 108 | -1 | |||
Constellations (cluster of symmetrical processors) |
20 | 31 | -11 | |||
† Number of installed systems in the Top500
|
Table 4: Top500 Processor Ranking, June 2007
|
||||||
Processor Generation
|
Count†
June 2007 |
Count†
Nov. 2006 |
Change
|
|||
Xeon 51xx (Woodcrest) | 204 | 31 | +173 | |||
Opteron Dual Core | 91 | 76 | +15 | |||
PowerPC 440 | 34 | 29 | +5 | |||
Pentium 4 Xeon | 27 | 118 | -91 | |||
Itanium 2 | 25 | 35 | -10 | |||
POWER5+ | 20 | 12 | +8 | |||
Xeon 53xx (Clovertown) | 19 | 0 | +19 | |||
Opteron | 16 | 37 | -21 | |||
PowerPC 970 | 11 | 18 | -7 | |||
POWER5 | 10 | 17 | -7 | |||
Xeon EM64T | 7 | 76 | -69 | |||
PA-8900 | 7 | 8 | -1 | |||
POWER4+ | 6 | 11 | -5 | |||
NEC | 4 | 3 | +1 | |||
SPARC64 V | 3 | 3 | — | |||
Itanium 2 Dual Core | 3 | 0 | +3 | |||
Alpha | 2 | 3 | -1 | |||
Cray X1 | 2 | 4 | -2 | |||
PA-8700+ | 2 | 9 | -7 | |||
POWER3 | 2 | 2 | — | |||
POWER4 | 1 | 2 | -1 | |||
PA-8800 | 1 | 3 | -2 | |||
Low Voltage Pentium Xeon Xeon | 1 | 1 | — | |||
50xx (Dempsy) | 1 | 1 | — | |||
PowerPC 450 | 1 | 0 | +1 | |||
† Number of installed systems in the Top500
|
||||||