Tuesday, August 9, 2011

IBM abandons its Blue Water Project (10 Pflops)

IBM and the National Center of supercomputers at the University of Illinois announced abandon their Blue Waters project which aimed to create a machine with a maximum computing power of more than 10 Petaflops which started in 2008. It was the first to have a sustained computing power of more than 1 Pflops 2008 was a symbolic year for IBM since he was the first to leave a petaflops supercomputer in excess.

Cell embedding processors and AMD Opteron, he took the top spot in June 2008 the top 500 supercomputers. The landscape today is very different. First place is occupied by the K of Computers Fujitsu SPARC64 VIIIfx boarding and provides 8.7 Pflops.

In the ranking of June 2011, IBM can only tenth. Second place is occupied by the Chinese machine Tianhe-1A, which reached 4.7 Pflops. Third place belongs to the Jaguar Cray running on six core Opteron, which has a maximum power of 2.3 Pflops.

Blue Waters was supposed to be operational next year. Made from 200 000 to 10 cores Power7 Po addressable disk space, it would have allowed IBM to take over the lead. The factors behind the abandonment of the project are vague. Both sides speak of a greater complexity than expected and a significant additional cost to IBM.

Finally, they claim they could not negotiate a shared responsibility. According to Rick Doherty, an analyst quoted by IDG, IBM was extremely invested in this project. It provided the hardware, software and personnel. NCSA seemed very distant and in the end, the benefits that Big Blue could get such a project were not up to the costs involved in its design.

Michael Feldman asks HPC Wire, meanwhile, if the reasons for abandonment are not located at PERC, the software architecture inspired by HPCS, which was supposed to create a viable business model for the supercomputers of TODAY 'hui. This could explain why IBM withdrew from the project when he had already started shipping the first racks.

IBM continues to work on large supercomputers. It prepares Sequoia, a machine with computing power of 20 Petaflops. It was commissioned by the Lawrence Livermore National Laboratory. Meanwhile, IBM will reimburse the $ 30 million put forward by the NSCA and servers shipped will be returned to Big Blue.

No comments:

Post a Comment