Compaq Computer has been awarded contracts to build what it claims will be the largest supercomputers in both Australia and Japan, the company said last week.
Compaq said it has been contracted to build a supercomputer with more than 450 AlphaServer SC processors for the Australian Partnership for Advanced Computing (APAC), based in Canberra at the Australian National University (ANU). Claiming that it will be the largest supercomputer in Australia, Compaq said the system will be used by researchers for large-scale scientific and engineering research in the areas of chemistry, physics, environmental science and biotechnology. The machine will be switched on this month and be fully implemented by October, Compaq spokesman Dick Calandrella said.
The Victorian Partnership for Advanced Computing (VPAC) in Australia also awarded Compaq a contract, to provide an AlphaServer SC 128-processor system for use in the areas of molecular modeling for new drugs and pattern discovery for fraud detection, Compaq said.
When Compaq claims that the machines are the largest in both Australia and Japan, the company is basing that on the TeraOps of the supercomputers, or the number-crunching capabilities of each machine, Calandrella said. Financial details of each of the three supercomputer contracts were not disclosed, he said.
Meanwhile, Japan's Atomic Energy Research Institute at the Kansai Research Establishment's Advanced Photon Research Center will become the home of a 1.5 TeraOPs supercomputer powered by 908 AlphaServer SC processors, Callandra said. In laymen's terms, that means the machine will be able to perform 1.5 trillion operations per second, he said.
The computer will be used for research activities like X-ray and medical diagnosis microscopy, ultra-precision machining and medical diagnosis and treatment. The supercomputer is currently being built, Calandrella said.
Compaq's supercomputers are massively parallel systems, which means they allow multiple processors to work on the same program or problem at the same time. An alternative is to vector supercomputers, which typically use a smaller number of more powerful processors, usually between one and 64 processors to a machine.
The RISC (reduced instruction set computer) processors used in some massively parallel systems have improved to a point where they can improve on some tasks more commonly performed by vector systems, Rabe said. Design and cost factors also favor massively parallel systems, he said. The performance of a vector computing system is still better with a small number of applications, but, overall, Rabe suggests that RISC systems provide a better price, scalability, more standard parts and higher reliability.