IBM launches low power supercomputer

IBM and Californian researchers plan to develop a new generation of commercial supercomputers that run at 15 times the speed of the fastest mainframes using only a fiftieth of the energy.

Big Blue is to partner the Lawrence Livermore National Laboratory in California to develop this new generation of commercial supercomputers, codenamed Blue Gene/L.

The project is based on IBM's Blue Gene experimental supercomputer currently being constructed in order to carry out large scale bio-molecular calculations.Although this is not scheduled for completion until 2005, availability of the Blue Gene/L prototype is targeted for 2004.

Tim Jennings, research production director at the Butler Group, said: "We do forever need more powerful computers. But IBM's strategy has been in another direction, namely grid computing. Long term, that would seem to have more applicability."

But a spokesperson from IBM commented: "Blue Gene/L needs to be seen as one of the nodes on the grid, so there is no conflict. It will provide an incredibly powerful node, in fact."

Blue Gene/L will use data-chip cells containing memory, a processor for computing and another for communicating.

This, said IBM, would optimise data access because the delay accessing data from a separate memory store was eliminated.

The prototype will use 65,000 of these cells which will work concurrently to solve complex problems.

"With 65,000 separate computers Blue Gene/L also fits into the self-managing, self-healing technologies that IBM is working on," said the IBM spokesperson.

IBM has been deliberately vague as to what the 'L' stands for, saying it could mean Livermore, low power or lite. But its spokesperson offered Linux as another possibility, perhaps hinting that the open source operating system might form the system's operating kernel.

IBM is also looking for a commercial partner which would benefit from the technology to create a commercial system version.

"We can't tell at this stage, but it could be a life sciences or financial services company that needs to search very large amounts of data," said IBM's spokesperson.