Cray Gets $174 Million U.S. Supercomputer Contract

The company's next-generation XC system, powered by Intel Xeon and Xeon Phi chips, will help protect the country's nuclear stockpile.


Cray has been awarded a $174 million contract with the federal government for a supercomputer that will be used to help manage the country's nuclear stockpile.

The contract with the National Nuclear Security Administration (NNSA), announced July 10, calls for Cray to deliver its next-generation XC supercomputer—code-named "Trinity"—and Sonexion storage system that will be housed at the Los Alamos National Laboratory and also will support two other facilities, Sandia National Laboratories and Lawrence Livermore National Laboratory.

The contract is one of the largest in Cray's history, according to company officials. The NNSA currently uses a Cray XE6 supercomputer called "Cielo." Trinity, armed with upcoming new Xeon and Xeon Phi processors from Intel, will offer more than eight times the applications performance of Cielo. Cray President and CEO Peter Ungaro said previous collaborations with the NNSA has helped drive the development of other supercomputing systems, including XT3 that was the basis of the Red Storm project, which was deployed in 2005 and hit the No. 2 spot a year later on the Top500 list of the world's fastest supercomputers.

"The NNSA has consistently deployed the world's most advanced supercomputing systems to support their critical mission of ensuring the health of our nation's nuclear stockpile," Ungaro said in a statement.

Trinity will be used to run massive simulations that will look at the security and effectiveness of the United States' nuclear stockpile. The bulk of the new system will be delivered in late 2015 and into 2016, according to Cray officials.

The multi-petaflop supercomputer will be powered by Intel's upcoming Xeon "Haswell" processors and the Xeon Phi "Knights Landing" chips, which will offer more than 60 processing cores. Intel's initial Xeon Phi coprocessors—dubbed "Knights Corner"—were introduced as alternatives to the increasingly popular GPU accelerators from Nvidia and Advanced Micro Devices that are being used in high-performance computing (HPC) environments to increase the performance of the systems while keeping down the power consumption.

However, Intel is making significant enhancements to the Knights Landing chips, including the capability of using them as either coprocessors or host processors. In addition, the 14-nanomter chips—which will begin to appear in systems in the second half of 2015 and will offer three times the performance of the current Xeon Phis—will be able to run single-threaded and parallel-processing workloads, and will come with improved on-package memory and interconnect.

The XC30 portfolio of systems includes the Aries system interconnect, Dragonfly network topology and Cray's high-performance Linux Environment.

The Sonexion storage solution will include 82 petabytes of capacity and 1.7 terabytes per second of sustained performance. Sonexian systems also come with Cray's Lustre expertise and the ability to scale without any degradation in performance, according to company officials.