Cray launched the Cray CS-Storm, a high-density accelerator compute system chock-full of NVIDIA Tesla GPUs and a peak performance of more than 11 teraflops per node. Based on the Cray CS300 cluster supercomputer the new CS-Storm is a powerful and dense system, packing an eight-to-two ratio of GPUs to CPUs.
The new system makes a 250 Teraflop rack possible, with 22 x 2U servers in a 48U rack, all with 176 NVIDIA Tesla K40 GPU accelerators. To take advantage of this dense GPU environment Cray has specifically tuned its CS programming environment and tools for performance GPU computing. The CS-Storm is targeted at specific HPC markets and needs that can justify the impressive compute power that the CS-Storm can offer. With a NVIDIA K40 GPU costing around $4,000 each, a fully loaded 2U CS-Storm would run $32,000, and a rack would be around $704,000 — and that is just for GPU chips.
Barry Bolding, Cray’s vice president of marketing and business development, said, “The Cray CS-Storm is built to meet the most demanding compute requirements for production scalability, while also delivering a lower total-cost-of-ownership for customers with accelerator workload environments. With the combination of an extremely efficient cooling infrastructure, Cray’s high-productivity cluster software environment and powerful NVIDIA K40 accelerators, the Cray CS-Storm is designed to be a production workhorse for accelerator-based applications in important areas such as seismic simulation, machine learning and scientific computing.”
The CS-Storm also features the Cray CS300 air-cooled system, Intel Xeon E5 2600 v2 processors, and the Cray Advanced Cluster Engine cluster management software. Cray says that the CS300 series of supercomputer clusters are available with air or liquid cooled architectures. The CS-Storm system is designed for HPC workloads in the defense, oil and gas, media and entertainment and business intelligence sectors.