Today Cray announced its highly anticipated line of “Cascade” supercomputers with the Cray XC30 system. Powered by Xeon processors and the Aries Interconnect technology recently purchased by Intel, Cray XC30 systems represent the flagship of the company’s offerings and feature innovative packaging and cooling optimized for ultrascale deployments of up to 100 Petaflops.
After several years of incredibly hard work focused on completing the most ambitious R&D program in our company’s history, today’s unveiling of the Cray XC30 supercomputer is an exciting moment for Cray employees and our customers who have been eagerly anticipating what is an amazing new system,” said Peter Ungaro, president and CEO of Cray. “As a follow on to our most successful, productive line of supercomputers, the Cray XC30 is the realization of our Adaptive Supercomputing vision and will provide researchers, scientists and engineers with a system that can adapt to fit their most demanding applications. We’re off to a great start with more than $100 million in contracts for this system, and we believe the Cray XC30 series of supercomputers will allow a broader base of users to leverage the world’s most advanced supercomputing technology.”
A Major Leap Forward
The Cray XC30 has the rare distinction of being one of the few high-end supercomputer lines that I can think of that started shipping ahead of schedule. As a result, this announcement comes right on the heels of the launch of the Cray XK7 line, which came out last week with the debut of the Titan supercomputer at Oak Ridge. So not to confuse these two lines, think of the XK7 as an upgrade of their XK6 line with the added boost of Nvidia K20X GPUs. With the proven Gemini interconnect and a torus topology, the Cray XK7 still has wide appeal as the only Cray system currently shipping with Kepler acceleration.
On the other hand, the Cray XC30 is the first in the Cascade series and as such represents a major leap in design architecture for the company made possible in part by Cray’s participation in the DARPA High Productivity Computing Systems program. To minimize latency and hops, the XC30 uses a Dragonfly interconnect topology. It’s blade architecture is connected by a backplane and cabinets are interconnected by optical cables, so that complex cabling is minimized.
And in what I found to be the most interesting part of the system, XC30 is cooled from the side with water-cooled chillers between cabinets. That means no more hot isles and cold isles as air departs the system with no temperature delta. As an added bonus, this design makes Cray XC30 much less noisy than typical supercomputers.
Several leading HPC centers have signed contracts to purchase Cray XC30 supercomputers, including:
- The Swiss National Supercomputing Centre (CSCS) in Lugano, Switzerland
- The Pawsey Centre in Perth, Australia, owned by CSIRO and operated by iVEC
- The Finnish IT Center for Science Ltd. (CSC)
- The Department of Energy’s National Energy Research Scientific Computing Center (NERSC) in Berkeley, Calif.
- The Academic Center for Computing and Media Studies (ACCMS) at Kyoto University in Kyoto, Japan
- The University of Stuttgart’s High Performance Computing Center Stuttgart (HLRS) in Germany
Early shipments of the Cray XC30 are starting now, and systems are expected to be widely available in first quarter of 2013.