Cray today is announcing one of its largest contracts to date – a $188 million contract for a massive new system that the National Science Foundation will use for scientific research.
The announcement comes on the opening day of a major supercomputing conference in Seattle at the Washington State Convention & Trade Center. More than 10,000 attendees are expected at SC11, which is formally the International Conference for High Performance Computing, Networking. It begins today and runs through Friday.
Cray’s new system for the NSF’s Blue Waters project will be delivered next year to the University of Illinois’ National Center for Supercomputing Applications.
It’s incredibly fast but not designed to break computing speed records. Instead it’s intended to perform sustained operations at ultrafast speeds. It’s capable of 11.5 petaflops peak speed and sustained speeds of 1 petaflop.
Six applications have run at sustained petaflop speeds, and five are on Cray “Jaguar” systems, a Cray spokesman said.
The Blue Waters system is a Cray XE6 combined with an XK6.
The specs, from Cray’s release:
– Cray’s scalable Gemini high-performance interconnect, providing a major improvement in message throughput and latency.
– AMD Opteron 6200 Series processors (formerly code-named “Interlagos”).
– Cray XK6 blades with NVIDIA’s next-generation Kepler GPU, which is expected to double the performance of the Fermi GPU on double-precision arithmetic.
– 1.5 petabytes of total memory (or four gigabytes per AMD Opteron 6200 Series processor core).
– Cray’s scalable Linux Environment (CLE) and HPC-focused GPU/CPU Programming Environment (CPE).
– A Cray integrated Lustre parallel file system with more than one terabyte-per-second of aggregate storage bandwidth and more than 25 petabytes of user accessible storage.
– Up to 500 petabytes of near-line storage and up to 300 gigabits per second of wide area connections.