IBM has revealed a supercomputing architecture that can reach 100 petaflops, 10 times more powerful than the fastest publicly disclosed supercomputer today.
IBM has outlined the architecture for its Blue Gene/Q supercomputer, which will be able to reach 100 petaflops. Image credit: IBM
The Blue Gene/Q architecture, announced on Tuesday, is expected to be used to predict the paths of hurricanes, decode gene sequences and perform many other simulation and analysis tasks that require large amounts of processing power.
"The third generation in the Blue Gene family of supercomputers, Blue Gene/Q operates at an order of magnitude faster than previous systems, deploying 16 multi-processing core technology and a scalable peak performance up to 100 petaflops," IBM said in a statement. "Blue Gene/Q, will provide an ultra-scale technical computing platform to solve the most challenging problems facing engineers and scientists at faster, more energy efficient, and more reliable rates than ever before."
Blue Gene/Q is the successor to Blue Gene/P, which scaled to three petaflops and was launched in 2007. In the past year Cray and Fujitsu have both announced supercomputing architectures that can scale to levels far above the current world leader — the 10-petaflops Fujitsu-designed Japanese K Computer — with systems capable of scaling to 50 and 20 petaflops, respectively.
A European project to build a prototype low-energy high-performance supercomputer evaluated top supercomputing architectures and found that machines based on prototype versions of IBM's Blue Gene/Q architecture demonstrated the greatest performance per watt efficiency (PDF, slide 4.)
The IBM system is based on the PowerPC A2 processing architecture (PDF) with 18 1.6GHz cores in total — 16 compute processors, along with a core that carries out operating system administration and maintenance and a 'spare core' that can be put in if one of the others fail. The chips include 10Gb I/O and two DDR3 memory controllers capable of managing up to 16GB of RAM, and consume 55 watts when delivering their maximum processing performance — 205 gigaflops. Each chip has around 1.47bn transistors.
Pricing and dates of general availability were not disclosed. The company has already lined up a number of institutions, including the Argonne National Laboratory and the Department of Energy's Lawrence Livermore National Laboratory, which will deploy petascale systems built on the architecture in 2012.
Get the latest technology news and analysis, blogs and reviews delivered directly to your inbox with ZDNet UK's newsletters.