IBM and Cray have come out as the big winners in DARPA's massive supercomputing contest, leaving Sun Microsystems out as the unlucky loser.
The Defense Advanced Research Projects Agency (DARPA) today revealed that Cray will take home $250m in funding, while IBM has earned $244m. These stunning handouts will go towards DARPA's effort to keep the US well ahead of rival nations in terms of computing horsepower. Both IBM and Cray are expected to produce so-called "Petascale" systems by 2010 that will theoretically push the US one computing generation past other countries.
"High productivity computing is a key technology enabler for meeting our national security and economic competitiveness requirements,” said Dr. William Harrod, DARPA program manager. "High productivity computing contributes substantially to the design and development of advanced vehicles and weapons, planning and execution of operational military scenarios, the intelligence problems of cryptanalysis and image processing, the maintenance of our nuclear stockpile, and is a key enabler for science and discovery in security-related fields."
The DARPA contest stood as the greatest prize in the high performance computing field. IBM, Cray and Sun had already fought off a number of other challengers to receive early funding from DARPA back in 2003. Each vendor hoped to secure this additional computing pork to help fund the development of their future computing lines.
DARPA's "Petascale" program is meant to address many of the cost complaints often faced by the labs that need large, specialized machines. The government agency demanded that all the vendor systems be practical for supercomputing tasks at labs as well as commercial HPC jobs.
For that reason, IBM centered its 2010 system around the Power processors that its uses across its current server line. In fact, the IBM system will run on the company's upcoming Power7 chips and taps the company's GPFS clustered file system.
Meanwhile, Cray's 2010 offering has been named "Cascade." It will combine Opteron-based servers, FPGAs, vector systems and massively multi-threaded systems into the same chassis. By melding its currently disparate product lines, Cray expects a dramatic fall in the price of its systems.
The company was depending on the DARPA funding to bring Cascade to fruition.
For its part, Sun had planned to create a box around a future version of the multi-core Rock processor.
The Petascale boxes will be expected to reach two petaFLOPS of sustained performance - where one petaFLOP equals one quadrillion floating point operations per second. The systems should also have a peak performance of more than four petaFLOPS.
Such systems would blow away the current top supercomputers.
An incredible amount of software work and hardware engineering will need to take place over the next three years for IBM and Cray to match their proposals. ®