Cray has added more horsepower to its latest supercomputer, the XC40, and already has scored some big-time customers.
The XC40, announced on Tuesday, is based on Intel’s new Xeon E5-2600 v3 server chip series, code-named Grantley, which was unveiled earlier this month. The supercomputer has a complement of new technologies that provide faster access to memory and storage, which boosts overall computing power.
Cray has made some of the world’s fastest computers, including what is currently the world’s second fastest supercomputer, Titan, which operates at a peak speed of 27 petaflops. The XC40 will serve as the basis for a U.S. Department of Energy supercomputer called Trinity, which is due for deployment in 2016 and will have 82 petabytes of distributed storage.
The processors in XC40 are two times faster than chips in its predecessor, the XC30, according to Cray. The new supercomputer will provide throughput of 1.7Tbps (terabytes per second) for internal data transfers, which is faster than the Titan system.
A new technology in XC40 called DataWarp brings solid-state drives closer to processors, so data will travel a shorter distance, speeding up program execution. SSDs are used as cache, where data is temporarily stored before being dispatched for processing or to long-term storage. The SSDs will plug into PCI-Express 3.0 slots, which provide faster throughput than traditional SATA controllers used for storage.
A revamped version of Cray’s Aries interconnect will hook up server closets, storage arrays and other components. Co-processors such as Intel’s Xeon Phi and Nvidia’s Tesla can also be installed in supercomputers to speed up specific calculations. The Xeon Phi will have support for DDR4 memory, which is faster and more power efficient than widely used DDR3 memory.
The Swiss National Supercomputing Centre in Lugano, Switzerland, will starting using an XC40-based supercomputer in late 2014 for research in areas like geology, biology, material science and meteorology, Cray also noted.