The Australian Bureau of Meteorology is acquiring a new supercomputer designed to enhance its weather and climate forecasting services.

NEC Australia secured the contract to supply the Bureau of Meteorology's (BoM) next supercomputing project, which will include an NEC SX-6 parallel vector supercomputer. The computer, whose processing power will begin at 1.15 TFLOPS (one trillion floating-point operations per second), will be operating at the Joint Bureau of Meteorology/CSIRO High Performance Computing and Communications Centre (HPCCC) by the end of the year.

"The increased power will allow the weather models to go to higher resolution, [which will allow us to] take more things into account and therefore provide a better outcome," Philip Tannenbaum, manager of the Bureau/CSIRO HPCCC, told ZDNet Australia  . "The current model uses 29 layers in the atmosphere, the new computer will allow up to 50 layers."

The new computer will allow the Bureau to interpret readings over much narrower areas and provide a more accurate representation of weather conditions. This will allow the bureau to be more specific in predicting where and when particular weather events will happen.

Tannenbaum said the computer would grow to a processing power of 1.8 TFLOPS within a year, and would make the HPCCC the most powerful supercomputing site in Australia. The computer is a commercialised derivation of the Japanese NEC Earth Simulator, which runs at 40 TFLOPS.

Vector supercomputers provide users with a higher percent of their rated peak performance as compared to scalar systems, according to Tannenbaum, explaining that the theoretical processing speed cited (ie, the TFLOPS) is far greater than what can be realistically achieved on any machine.

A scalar processor adds one number to another number and gives a result, with one instruction. Using a vector processor one instruction can perform 256 additions and obtain 256 results. Effectively, a vector processor runs parallel internally, and this is compounded when many processors are run parallel in a supercomputer.

"With sustainable performance, by the end of 2004 this machine will be able to sustain 500GFLOPS or more day in day out, which is a lot more than any of the scalar supercomputers with similar peak performance ratings," said Tannenbaum.

HPCCC currently runs two supercomputers running at 256 GFLOPS, which recently dropped off the list of the 500 fastest computers in the world. The new and old computers will run in tandem for some time, before the new computer completely replaces the old.

There has been a spate of supercomputer acquisitions in Australia recently, mostly to support biotechnology research. Last week the University of Queensland's Centre for Computational Molecular Science announced the purchase of a 1.4 TFLOP computer that would be used by approximately 15 research groups at the university to carry out molecular-scale computations in new and developing areas of technology.

Similarly, Adelaide University recently obtained a 1.2 TFLOP supercomputer and late last year Swinburne University became the second institution in Australia to run a one TFLOP supercomputer.

Copyright © 2009 CBS Interactive, a CBS Company. All Rights Reserved.
ZDNET is a registered service mark of CBS Interactive. ZDNET Logo is a service mark of CBS Interactive.