Sign up for our newsletter and get the latest HPC news and analysis.
Send me information from insideHPC:


Hornet Supercomputer Installed at HLRS in Germany

Cray XC40 system “Hornet” of HLRS (High Performance Computing Center, Stuttgart/Germany), Peak Performance: 3.8 PetaFlops

Cray XC40 system “Hornet” of HLRS (High Performance Computing Center, Stuttgart/Germany), Peak Performance: 3.8 PetaFlops

Today the Gauss Centre for Supercomputing (GCS) announced on 13 November 2014 that its member centre High Performance Computing Center Stuttgart (HLRS) successfully completed the installation of HPC system “Hornet”. The new HLRS supercomputer, a CRAY XC40 system which delivers a peak performance of 3.8 PetaFlops, has been declared fully operational and will be available for its scientific and industrial usership as of immediately.

HLRS’s new Cray XC40 system is based on the new Intel Xeon processors, formerly code-named “Haswell”, and the Cray Aries system interconnect. In its current configuration state, Hornet consists of 21 cabinets hosting 3,944 compute nodes, which sums up to a total of 94,656 compute cores. The system’s main memory capacity is 493 Terabyte. Users will specifically benefit from the now quadrupled storage space the HLRS supercomputing infrastructure provides: 5.4 Petabyte of file storage with an Input/Output speed in the range of 150GB/s are available to meet the performance challenges of today’s most demanding HPC users, which come from a wealth of fields ranging from the automotive and aerospace research and industries to medicine and life sciences, astrophysics and geophysics, amongst others.

Hornet is the successor of HLRS’s previous flagship computer Hermit, which after about three years in service is gradually been taken out of operation. The system replacement went according to the earlier agreed HPC system roadmap as defined by GCS. “HPC systems have become an indispensable tool to achieve breakthrough discoveries and innovations. With Hornet, we have taken the next step for HLRS to enable world-class research,“ states Professor Dr.-Ing. Michael M. Resch, Director of HLRS. “It is very important for us that we are now in a position to offer state-of-the-art HPC technology also to our industrial users. Together with our simulation expertise the extended simulation capacities and capabilities of Hornet provide our scientific and industrial users in Germany and Europe with perfect means to continue enabling innovation and quality of the highest degree.”

More Power Output – Less Power Consumption

HLRS’s new high-end HPC system delivers significantly increased computing power – it outperforms Hermit both in peak as in sustained performance by a factor of about 4 – while at the same time excelling in drastically reduced power consumption. In combination with the at HLRS installed energy-efficient cooling system, an optimal cost-of-ownership is achieved which furthers the national HPC centre’s aspirations to be a model example in the area of energy efficiency and sustainability.

Hornet: HPCG-Benchmark

Delivering proof that their HPC clientele enjoys top priority, HLRS underwent all necessary efforts to deliver data for a new kind of benchmark, which from the user’s perspective is of high significance: the HPCG benchmark (High Performance Conjugate Gradient). This new benchmark, which debuted at ISC’14 in Leipzig in June of this year, does not just concentrate on raw CPU performance but stresses the system balance, e. g. floating point and communication bandwidth and latency, and it tightens the focus on messaging, memory, and parallelization—parameters that add up to an “averaged” system performance which is especially from the users’ point of view more beneficial and more important. Hornet delivered about 40 TFlops in this test. The world-wide HPCG-benchmark rankings will be announced at the upcoming Supercomputing Conference 2014 in New Orleans, LA/USA (16-21 Nov.).

According to HLRS, Hornet will be upgraded in 2015 with 20 additional cabinets, boosting the system’s expected peak performance to over 7 Petaflops.

Sign up for our insideHPC Newsletter.

Resource Links: