Intel Broadwell Xeon Debuts on Cray XC and CS Series Supercomputers

Your next demanding compute challenge: Does it seem insurmountable? Need more cores? Higher base processor frequency? Faster memory? All of the above?

Take a look at Cray’s integration of next-generation Intel® Xeon® E5-2600 v4 processors, previously codenamed “Broadwell.” The new processor enables theoretical peak performance boosts of up to 30 percent for new and existing Cray® XC™ series supercomputers and Cray® CS™ series cluster supercomputers. The Broadwell “new process technology tick” implementation is all about improving performance, delivering higher peak flops than previous-generation “new micro-architecture tock” Haswell Xeon processors. Built on Intel’s 14nm Grantley platform, the new device family offers configurations supporting more cores per processor (up to 22), faster base frequencies and higher-speed memory (DDR4 2,400 MHz).

Upgradability by design – to protect your investment

In addition to new CS series users like Alfred Wegener Institute (AWI) leveraging these faster devices, existing customers are using the same socket, drop-in replacement to upgrade the performance of their current Cray XC series systems. Continuing on Cray’s vision for adaptive supercomputing, customers like the Swiss National Supercomputing Centre (CSCS), Germany’s National Meteorological Service, the Deutscher Wetterdienst (DWD), the European Centre for Medium-Range Weather Forecasts (ECMWF) and the U.K. Met Office are benefiting from the XC supercomputer’s upgradability-by-design to improve their TCO and extend the lifetime of their computing systems. Via a proactive roadmap of in-place upgrades, customers protect their investment while continuing to deliver improving performance to their hands-on users — engineers, scientists, researchers, analysts and more.

Practice makes perfect – Broadwell’s benchmarks

Collaborating for years prior to general availability announcements, Cray and Intel tested a broad array of application codes on sample processors during the development process. To ensure sustained performance metrics that we can review with customers and prospects in advance of their market requirements, Cray used internal systems to obtain early performance information on specific benchmarks, ISVs codes and customer applications.

Standard system benchmarks that Cray has already run on Broadwell include: HPCC, HPCG, SPECfp, SPECint, Stream, Memory Latency, SWIM, GUPs, DGEMM, HPL and IMB. The diverse example applications include: LS-DYNA, STAR-CCM+, Fluent, RADIOSS, Abaqus, PowerFLOW, OpenFOAM, Qbox, Gromacs, IFS, UM, Harmonie, FESOM, NAMD, NWChem and Quantum Expresso.

Price/performance in production and commercial environments

XC series supercomputers are dominant in the lists of Top10/50/100 systems worldwide because Cray’s HPC-optimization of hardware, interconnect, network and software delivers sustained performance at scale. Additionally, Cray’s 40-plus years of experience in innovative supercomputing power, cooling and packaging design means we deliver systems that excel in running high-utilization capability or capacity workloads in production environments where reliability, availability and serviceability are an imperative. For example, several of above-mentioned Cray/Broadwell customers operate in the increasingly demanding 24/7 environment of real-time weather forecasting.

Government and research organizations are not the only domains to benefit from the continued march of performance improvements. The speed-up advancements across the XC series and CS cluster supercomputers enable a broad range of flexible configuration offerings that span both the performance and price spectrums. That means significant competitive advantages for numerous commercial segments and applications, including cybersecurity, energy, engineering and manufacturing, finance and more.

Learn more about the new XC series upgrade at DWD and the Intel Broadwell family of Xeon processors.

Speak Your Mind

Your email address will not be published. Required fields are marked *