13 Sep 2011 Issy Les Moulineaux - The Barcelona Supercomputing Center - Centro Nacional de Supercomputacion (BSC-CNS) now has a new cluster with graphical accelerators which will be used to consolidate its research in programming models, tool development and application porting. This new Bull system, based on bullx nodes, each equipped with Intel processors and NVIDIA GPUs, would be ranked 90 in the world in the June 2011 Top500 list and has a peak performance of 186 Teraflops. This cluster doubles MareNostrum's calculation capacity, consumes seven times less power and occupies 13 times less space. The combination of this new machine and the current MareNostrum system means that the public consortium has tripled the supercomputing resources provided to the Spanish scientific community.
Unlike the MareNostrum, the new BSC-CNS cluster is a special-purpose supercomputer that offers an optimum peak performance for specific applications or programs. Its advantage, compared to
general-purpose machines, is its speed and low energy consumption. The cluster is more difficult to program but this is not a problem as the BSC-CNS relies on recognised in-house expertise on programming models.
Mateo Valero, BSC-CNS Director, stated: "We currently have some of the world's best programming models so we are in an optimal position to provide some specific applications the most efficient usage of the new system, offering a significant increase in their performance. Additionally, our tools enable the analysis of each piece of hardware and software used in HPC (High Performance Computing), to improve the efficiency of the use of these resources." The BSC-CNS is working on developing the Star Superscalar (StarSS) programming model and a set of tools including Paraver, Dimemas and GMAC to enable the usage of new HPC systems including special-purpose processors or components.
BSC-CNS's Associate Director, Francesc Subirada, emphasize that "the BSC's programming model and tools allow the optimum use of supercomputing resources, accelerating the performance of complex applications with a high social impact - genome analysis, seismic imaging, design of new materials, etc. Their increasing use in universities, research centres and companies around the world encourages us to carry on and further their development."
Julio del Valle, Director of Bull Spain, stated: "The trust in our company and our technology shown by a centre with a reputation like the BSC-CNS's is an endorsement of Bull's long term strategy to place our company, and European technology, among the best in high performance computing systems. This is based on our ability to innovate, develop and integrate. We are proud to have numerous bullx systems among the most powerful in the world, many of them operating in different European research institutions. The Bullx systems help these institutions to reinforce Europe's future competitiveness. Bull strongly believes in this competitiveness, which can only be expressed through joining our R+D capabilities. In this sense the BSC-CNS is an example to all of us."
Sumit Gupta, manager of Tesla GPU Computing products at Nvidia, stated: "The dramatic boost in application performance enabled by the GPU is enabling researchers across many fields of science research to increase the pace of their work. We look forward to seeing what advances will be made by BSC users with the addition of these new resources." NVIDIA named BSC a CUDA (Compute Unified Device Architecture) Research Center in 2010 in recognition of its work to further science through the use of GPU Computing.
This cluster will become part of the Spanish Supercomputing Network (RES) allowing Spanish scientists to increase their supercomputing resources via the current Access Committee.
The new Bullx system in BSC-CNS has the most advanced technologies such as InfiniBand QDR and Solid State Disk. It is equipped with rear cooling doors to minimize the carbon footprint.
The new system includes: