X

IBM, Intel chew up supercomputer list

Big Blue hounds No. 1 HP by increasing the number of its systems on the list from 131 to 158. Meanwhile, the number of systems based on Intel processors jumps from 56 to 119.

Stephen Shankland Former Principal Writer
Stephen Shankland worked at CNET from 1998 to 2024 and wrote about processors, digital photography, AI, quantum computing, computer science, materials science, supercomputers, drones, browsers, 3D printing, USB, and new computing technology in general. He has a soft spot in his heart for standards groups and I/O interfaces. His first big scoop was about radioactive cat poop.
Expertise Processors, semiconductors, web browsers, quantum computing, supercomputers, AI, 3D printing, drones, computer science, physics, programming, materials science, USB, UWB, Android, digital photography, science. Credentials
  • Shankland covered the tech industry for more than 25 years and was a science writer for five years before that. He has deep expertise in microprocessors, digital photography, computer hardware and software, internet standards, web technology, and more.
Stephen Shankland
3 min read
IBM and Intel have edged ahead on a quickly changing list of the 500 fastest supercomputers.

Hewlett-Packard leads the herd with 159 systems on the list, which is set to be released Monday, up from 137 six months ago. But IBM has almost caught up to HP by increasing its numbers to 158 from 131. Meanwhile, the number of systems based on Intel processors--mostly Xeons and a handful of Itaniums--increased from 56 to 119.

Big Blue continues to lead when it comes to the largest fraction of the total computing horsepower. Of the total ability of 375 trillion calculations per second, or teraflops, IBM machines account for 34 percent. More than a tenth of the total comes from the Japanese Earth Simulator, a behemoth with a 36-teraflop rating.

Sun Microsystems, despite a new high-performance computing effort, has only nine systems on the list.

The Top500 list is released twice each year by researchers from Germany's University of Mannheim and the University of Tennessee. And although it always changes, the flux in the new list is higher than usual.

Turnover is high, for one thing. The lowest-ranked system on the new list has a speed of 245 gigaflops, or billion calculations per second. Six months ago, that was enough performance for a comfortable 285th-place ranking.

And there are new varieties of systems arriving, including Cray's new X1, which holds 10 spots starting with No. 112.

Making a rare appearance: a Windows-based system, in the form of a cluster of Dell Computer servers with 384 Xeon processors. It was built at the Cornell Theory Center and is ranked No. 50.

And turning up at No. 362 is an HP system with 64 of Intel's forthcoming Itanium 2 6M Madison processors running at 1.5GHz. The system has a speed of 335 gigaflops. The No. 8 spot went to an HP system with 1,540 Itanium 2 processors at Pacific Northwest National Laboratory.

But Xeon processors are the main reason for Intel's climb.

The company's jump up the list is largely due to increased performance and lower costs, said Tom Gibbs, director of industry marketing at Intel. The chipmaker worked on keeping prices down to prompt a massive change in the high-performance computing market.

"The hypothesis we had three years ago was that by altering the price-performance (ratio) in these machines, we would see a dramatic increase in demand," Gibbs said. "A year ago we had three systems on the list."

One trend Gibbs noted is that the number of processors in the Intel systems is declining because of increasing speeds. In the last version of the list, the smallest Intel machine had 128 Xeons. The current one has 92.

Also changing are some of the high rankings. Second and third place were occupied by two identical HP systems called ASCI Q at Los Alamos National Laboratory that now have been combined into one with a 14-teraflop speed. The system uses 8,192 Alpha processors.

A Linux Networx system with 2,304 Xeon processors at sister lab Lawrence Livermore National Laboratory was remeasured at 7.6 teraflops, enough to move it past ASCI White, an IBM machine also at that lab.

The list uses the Linpack benchmark test. Although it's a handy test, it's an imperfect measurement of real-world performance in many cases.

News.com's Michael Kanellos contributed to this report.