Dec. 2006 / Jan. 2007 | As expected, the IBM Blue Gene/L system installed at the Department of Energy’s Lawrence Livermore National Laboratory retained the premier spot in the latest edition of the Top 500 list of the world’s most powerful supercomputers. The rankings were released last month at SC06 — the International Conference for High Performance Computing, Networking, Storage and Analysis.
The Top 500 list is based on a single benchmark — the number of calculations possible per second. While critics view the ranking as of limited use for determining which systems would perform the best on specific applications, the single benchmark allows comparisons from list to list, and is useful in identifying high-performance computing (HPC) technology trends.
The top-ranked IBM Blue Gene system scored a Linpack performance benchmark of 280.6 teraflops (280.6 trillions of calculations per second).
The second place system, the Cray-built Red Storm at Sandia National Laboratories, jumped from ninth place six months ago. In that time, a fifth row of cabinets was added and the entire system upgraded to use AMD dual-core processors. The Cray Linpack benchmark is now 101.4 teraflops, which makes the system only the second installation ever to exceed 100 teraflops.
Perhaps the most notable trend in the new list — one that labs should be seizing upon — is the rapid adoption of multi-core processors for high performance computing. Dual-core processors from Intel and AMD have only been available for about a year, yet already feature in more than 100 of the systems on the current list.
The adoption of multi-core chips in HPC systems is expected to shake up the list, and the industry, in 2007. Typically more powerful computers displace about half of the systems every six months. That trend is expected to continue with some organizations already planning systems based on quad-core processors.
For instance, one announced project involves the Louisiana Optical Network Initiative (LONI), a network of computers run by eight state research universities. By year end, the group plans to have Dell servers equipped with a total of 1,440 Intel quad-core processors up and running and delivering about 50 teraflops of power.
The University of Texas has a multi-core plan in place as well. Built by Sun Microsystems, the proposed system will use 13,000 AMD quad-core processors. The university says the system, once operational in mid-2007, will deliver more than 400 teraflops. If such a system were in existence today, it would be the most powerful computer in the world.
A good number of the systems on the Top 500 list are used for life science research. For example, the number three system, an IBM eServer Blue Gene Solution at the IBM Thomas J. Watson Research Center has given researchers there the ability to do computer simulations that were not possible before. In fact, over the past 18 months, results from the system have been at the heart of several published research papers looking into the interaction between proteins and membrane layers.
Other installations of note that are used for life science research include system number 5, an IBM BladeCenter cluster at the Barcelona Supercomputing Center; system 12, a Dell PowerEdge cluster at the Texas Advanced Computing Center at the University of Texas; system 20, a Dell cluster named Darwin, installed at Cambridge University; and system 21, an IBM eServer Blue Gene system at the Computational Biology Center of Japan’s Advanced Institute of Science and Technology.
For most labs, the Top500 systems are not practical or affordable. However, the trends show the rapid adoption of technologies that are within reach of virtually all labs.
Email Salvatore Salamone.
Subscribe to Bio-IT World magazine.