IBM has unleashed a new iteration of its chess-playing powerhouse Blue Gene, which it also reckons is useful in academic and commercial fields like drug discovery and mineral exploration.
The Blue Gene/P supercedes the Blue Gene/L with faster Power PC cores, more memory and a symmetric multiprocessing (SMP) mode to support multi-threaded apps, as IBM scrambles to keep supercomputing in line with the coding vogue in cluster systems.
IBM says Blue Gene/P is three times more powerful than its predecessor, and seven times more energy efficient than its supercomputing bretheren. The firm has visions of flogging systems to financial institutions. Dave Turek, VP of deep computing said: "This is on course with an adoption cycle - from government labs to leading enterprises – that we’ve seen before in the high-performance computing market."
To that end, the software environment has been tweaked to provide better app support and system management. Blue Gene systems run a version of Linux.
Back on the silicon side of things, each two foot by two foot board in a Blue Gene/P setup comprises 32 chips, which each contain four Power PC 450s clocking 850MHz. The boards are stacked 32-high to build racks which will churn 13.9 teraflops. The petaflop configuration calls on 72 of these racks; a total of 294,912 processing cores. The modular system tops out at a 216-rack cluster delivering three petaflops.
According to IBM, for a pile of laptops to run at the same speed, it would have to be a mile and a half high [please address all pedantry to IBM, Computer Road, Big Corporation City, NY 10101, USA].
The US Department of Energy will be first to wrangle the beast, when Argonne National Laboratory lets it loose later this year. In the UK, the Science and Technology Facilities Council's Daresbury Laboratory in Cheshire is planning a deployment.®