IBM has left a scheme to develop a supercomputer for US academic research due to ballooning financial and technical support costs.
The termination of the development contract was announced by IBM and the National Centre of Supercomputing Applications (NCSA) on Monday. It means the Blue Waters scheme to build a general purpose supercomputer for scientific research has lost its main contractor, and a new one will have to be found.
"The innovative technology that IBM ultimately developed was more complex and required significantly increased financial and technical support by IBM beyond its original expectations," IBM said in the joint statement on Monday. "NCSA and IBM worked closely on various proposals to retain IBM's participation in the project but could not come to a mutually agreed-on plan concerning the path forward."
Work started on the Blue Waters project in 2006. It received a $208m (£127m) grant from the US National Science Foundation and IBM was selected as the supercomputer vendor in 2007. IBM expected to base the computer around its Power7 processors.
There were some innovative technologies proposed that became more difficult for IBM to implement.
– John Melchi, NCSA
When IBM announced its withdrawal from the project at midnight on Saturday it had been paid about $30m from the grant, which it will return, John Melchi, director of administration for the NCSA told ZDNet UK.
"There were some innovative technologies proposed that became more difficult for IBM to implement," he said. "The cost incurred by IBM went up substantially."
The NCSA is "diligently" working to select another contractor, he said, with the grant remaining at the same value. "There's been no discussion about increasing that with inflation," he said.
Sustained versus peak performance
The goal for the Blue Waters scheme is to develop a supercomputer with a peak performance of around 10 petaflops and a reliable performance of a petaflop for general applications.
The main benchmark for peak performance of supercomputers is the Linpack, which is used by the Top500 list to rank the top speed of the world's supercomputers. This scores the world's fastest computer as the Fujitsu-designed K Computer with an Rmax score of 8.16 petaflops.
Blue Waters's goal is to create a computer that peaks at 10 petaflops with a sustainable one petaflop, Melchi said. Sustained means the type of performance a typical science application could expect to achieve on the machine.
"Rather than taking a simple algorithm that you know will scream on a large machine, we're taking real world applications and putting them on a supercomputer," he said. "Our emphasis has been on doing real world science rather than what I would categorise as simple benchmarks on a system [such as Linpack peak scores]."
Once completed, which Melchi anticipates will be by the end of 2012 even with a new contractor, the Blue Waters computer will run a range of applications such as weather simulations, weather prediction, cosmology and the analysis of complex biological systems.
