Top Supercomputers Revealed

A U.S. supercomputer maintains top spot in the latest list of Top 500 Supercomputer Sites, which was released today at the International Supercomputer Conference in Heidelberg, Germany.

The Department of Energys IBM Corp. (NYSE: IBM) Blue Gene/L (BG/L) supercomputer at the Lawrence Livermore National Laboratory remains in pole position, having nearly doubled in size since the most recent list was released in November.

The system is now capable of 136.8 Teraflops, or trillions of calculations per second, up from 70.72 Teraflops (see DOE Gains Supercomputing Crown and IBM Dominates Supercomputing).

Compare that with the bottom supercomputer on the list: an un-named U.S. government supercomputer capable of 1.166 Teraflops. It's a gauge of supercomputing's progress that this same deployment clinched 299th position six months ago. At that time, the 500th supercomputer on the list was a machine at SBC Service Inc., capable of a maximum performance of 850.6 Gigaflops, or billions of operations per second.

But, for a real indication of how supercomputing is growing, consider this: The bottom-placed supercomputer has about the same computational power as all 500 systems combined, when the list was first created in June 1993.“Every six months when the list is published we see a tremendous amount of turnover,” says Dr. Nan Boden, executive vice president of Myricom Inc., which provides the interconnection technology for around 140 systems on the list. “What it takes to even get onto the Top 500 is an incredibly strong IT system.”

Today's list indicates plenty of sites worldwide are eager to qualify. There are five new entries amongst the world’s top 10 supercomputers. These include another Blue Gene supercomputer at IBM’s Thomas J. Watson Research Center and a Cray Inc. (Nasdaq: CRAY) deployment at Sandia National Laboratories. In Europe, the University of Groningen and the Ecole Polytechnique Federale de Lausanne made their top 10 debuts, along with Japan’s Computational Biology Research Center.

The launch of the latest Top 500 list is also eagerly awaited by vendors keen to get their technology rubber-stamped in high-end environments. Both IBM and HP, for example, have used this week’s Heidelberg conference as the springboard for clustering announcements (see IBM's Cluster Bluster and HP Polishes Its Clusters).

This message appears to be getting through: 304 of the supercomputers in the current list are clusters, up from 296 in November 2004 and 291 a year ago. In November 2003, clusters accounted for only 208 of the supercomputers on the list.

Another vendor gleefully rubbing its hands today is Intel Corp. (Nasdaq: INTC). A total of 333 systems are now using Intel processors, up from 320 six months ago and just 287 in June 2004. IBM is lagging some way behind in second place, with 77 systems using its Power processors, while HP’s PA-Risc and Advanced Micro Devices (NYSE: AMD) processors account for 36 and 25 deployments, respectively.The list is due out again in six months. It is updated twice yearly by representatives of the University of Mannheim, the University of Tennessee, and the National Energy Research Scientific Computing Center.

— James Rogers, Site Editor, Next-Gen Data Center Forum