Will Japan beat U.S. in supercomputers?

The top 500 supercomputers in the world are from IBM, SGI, Sun, and HP. But this does not necessarily mean a U.S. lead, as Japanese makers come on strong.

Brooke Crothers Former CNET contributor
Brooke Crothers writes about mobile computer systems, including laptops, tablets, smartphones: how they define the computing experience and the hardware that makes them tick. He has served as an editor at large at CNET News and a contributing reporter to The New York Times' Bits and Technology sections. His interest in things small began when living in Tokyo in a very small apartment for a very long time.
Brooke Crothers
3 min read
The top 500 supercomputers in the world are from IBM, Silicon Graphics, Sun Microsystems, and Hewlett-Packard--but this may not necessarily indicate U.S. strength in the supercomputing world, as Japanese makers come on strong.

SGI took the lion's share of the top 100 of the total 500 machines, and, surprisingly, Intel tops the list despite the fact supercomputing is not one of its major markets.

As their name implies, supercomputers are the most powerful computing machines in the world, and they are typically used for extremely intensive tasks such as forecasting weather or simulations for aircraft design. They are a crucial indicator of a country's state of computer technology.

Equally interesting is that machines from IBM outnumber the combined totals of supercomputer stalwarts like Hitachi and NEC in the top 50.

The IBM machines use a technique known as parallel processing while the more traditional supercomputers use a method called vector processing. Basically, parallel processing spreads computing tasks among many processors running at the same time, whereas vector processing relies on fewer, faster processors.

But some in the industry say that supercomputer development in the United States is in a perilous state because the large Japanese makers of vector computers have increased their performance lead as U.S. makers such as SGI emphasize alternative technologies. SGI bought out Cray, once the world leader in vector processing.

The top 500 list is a good indication of peak performance but not a good benchmark necessarily for how well an application will perform in the real world, said Richard Russell of Tera Computer in Seattle. Tera's computers run vector code but the company describes them as multi-threaded supercomputers.

"Many think there is a crisis in the U.S.," Russell said. "The Japanese are building very fast vector machines and the U.S. is falling behind." Russell said that critical applications for the automotive and aerospace industries, for example, run better on traditional vector machines than the new parallel processing supercomputers coming from U.S. companies.

However, parallel machines do show up at large companies. For example, aerospace firm Lockheed Martin and automotive giant DaimlerChrysler both have SGI Origin machines, and insurance company Aetna and retailer Sears both use IBM parallel machines.

Machines from Sun Microsystems and Hewlett-Packard begin to appear in the list regularly in the latter two-thirds of the top 500 list as do SGI Origin systems--which are not Cray supercomputers.

IBM seems to stand out as one of the strongest new supercomputer suppliers in terms of the raw number of machines appearing on the list. IBM said separately in a statement that it posted double digit growth on the list and captured the No. 2 overall ranking.

The Web site that lists the supercomputers in its introduction qualifies the ranking by stating that they list computers ranked by their performance on the Linpack Benchmark and goes on to say: "While we make every attempt to verify the results obtained from users and vendors, errors are bound to exist and should be brought to our attention."

The site also states that the list has been compiled twice a year since June 1993 "with the help of high-performance computer experts, computational scientists, manufacturers, and the Internet community."

The Intel supercomputers, of which there are only a few, are part of a Department of Energy supercomputer program.

News.com's Stephen Shankland contributed to this report.