Supercomputer giant Cray Thursday signaled its intent to revitalize its efforts in midrange supercomputing, but customers may have to pay a bit more.
Cray (NASDAQ:CRAY) has discontinued its previous entry-level offerings -- the Cray CX1 Deskside Supercomputer, with prices starting at $25,000, and the Cray CX1000 Supercomputer, with prices starting under $100,000 -- and introduced a new entry-level system for the midrange market with prices starting at $200,000.
While the CX line stood on its own, the new entry-level midrange offering consists of configurations of Cray's high-end XE6 and XK6 supercomputers. This allows the Cray XE and Cray XK architectures to scale up or configure down.
Cray said the new offering combines the lower system cost and breadth of software application support for the CX1 and CX1000 systems with the petascale technologies and scalable architecture of the XE6 and XK6 lines, including Gemini interconnect, the latest version of the Cray Linux Environment, AMD Opteron 6200 Series processors and NVIDIA Tesla GPUs.
"The Cray CX line was a success for us as it allowed us to reach a new segment of users in a broader set of industries," said Peg Williams, Cray's senior vice president of high performance computing systems. "For some time, customers have been looking for us to combine the strengths of both of our product offerings into a single architecture and we've now accomplished that goal. By adding the final pieces to the puzzle -- low starting prices, broad ISV applications support, scalability and a tightly integrated architecture -- we now have an ideal supercomputing product to successfully meet the needs of the midrange market."
The new Cray XE6m supercomputer is optimized to support scalable application workloads in the midrange high performance computing (HPC) market where applications require between 700 and 13,000 cores of processing power. Cray said it is engineered to meet the needs of capability-class HPC applications.
The XE6m uses a two-dimensional torus architecture that Cray said is optimized for superior application performance between 700 and 13,000 processing cores. It incorporates two types of dedicated nodes: compute nodes and service nodes. The compute nodes are designed to run parallel MPI or OpenMP tasks. Each node is composed of two AMD Opteron microprocessors (16 core) and direct attached memory, coupled with a dedicated communications resource. Cray designed the XE6m to support future compute notes, including GPU accelerators. The service nodes are designed to provide scalable system and I/O connectivity. They can also serve as login nodes from which applications are compiled and launched, eliminating the scheduling complexities and asymmetric performance problems associated with common cluster designs. Cray said it also ensures that performance is uniform across distributed memory processes.
"Cray's new entry-level configurations leverage its deep HPC technology portfolio to create purpose-built systems for the departmental technical computing market segment," said Earl Joseph, IDC program vice president for HPC. "This segment was worth around $3 billion in 2011 and IDC projects that it will grow at a healthy 7 percent to 8 percent CAGR through 2015."