Apple cluster ranks third in supercomputer list

A computer cluster consisting of 1,100 Apple Computer Power Mac G5 desktops connected together in a matter of weeks by Virginia...

A computer cluster consisting of 1,100 Apple Computer Power Mac G5 desktops connected together in a matter of weeks by Virginia Polytechnic Institute and State University is the third fastest computer in the world, according to the latest edition of Top 500.

The cluster, which attracted considerable attention when it was announced earlier this year because of its use of off-the-shelf computers and networking technology, is the highest entry in the twice-yearly ranking that also includes a number of other machines based on clusters of standard computers running the Linux operating system.

The ranking relies on computer owners or makers submitting details of their machines and is based on the results of a standard benchmark called Linpack. Machines are ranked by the maximum number of floating operations per second (flops) achieved during the test.

Top two positions in the latest edition of the list are unchanged, with the NEC Earth Simulator of the Japan Marine Science and Technology Centre ranked number one at 35.9Tflops (gigaflops) and the Hewlett-Packard ASCI-Q machine at Los Alamos National Laboratory ranked number two at 13.9Tflops.

Virginia Polytechnic's Apple cluster gained third place with a 10.3Tflops ranking, and became only the third computer in the world to achieve a performance of more than 10Tflops, according to the list compilers.

Each node on the cluster is a Power Mac computer with dual 2.0GHz PowerPC 970 processors made by IBM, 4Gbytes of memory and 160Gbytes of storage. They are connected together using Infiniband interconnect technology and Gigabit Ethernet switches.

Over the past six months, an additional three machines based on clusters of smaller computers, all of which run the Linux operating system, have been deployed that are powerful enough to take positions in the top 10.

A cluster of Dell PowerEdge 1750 servers with a total of 2,500 Intel Xeon 3.06GHz processors and running Red Hat Linux ranked fourth.

The machine, which is at the National Center for Supercomputing Applications (NCSA) at the University of Illinois, achieved a maximum performance of 9.8Tflops.

Pacific Northwest National Laboratory's MPP2 computer was the ranked fifth with a maximum performance of 8.6Tflops.

The computer consists of Hewlett-Packard Integrity RX2600 computers and includes a total of 1,936 of the 1.5GHz version of Intel's Itanium 2 chip and also runs a Red Hat version of Linux.

The final new entry into the top ten, at position six, is the Lightning cluster at Los Alamos National Laboratory (LANL). The computer is the fastest system in the ranking based on Advanced Micro Devices' Opteron processor and also the first machine to enter the top 10 based on the chip.

It includes 2,816 of the processors and achieved a maximum performance of 8.1Tflops. The computer was supplied by Linux Networx.

There are now 208 computer clusters in the Top 500, up from 149 in the previous study six months earlier.

The latest list also marks the emergence of China as a growing supercomputing power.

Deepcomp 6800, a machine based on 1,024 of Intel's Itanium 2 processors running at 1.3GHz, built by the country's Legend Group, has been ranked the 14th most powerful computer in the world at 4.1Tflops.

It also puts China, in terms of the maximum performance of a single machine, ahead of all other nations except the US and Japan. The computer is installed at the Chinese Academy of Science (CAS).

The previous most powerful computer in China, Deepcomp 1800, also installed at CAS, ranked 82 in the latest list and has achieved a maximum performance of 1.3Tflops.

Martyn Willams writes for IDG News Service

Read more on Server hardware