Thursday, March 10, 2011

QLogic TrueScale InfiniBand integrates with NVIDIA Tesla GPUs

ALISO VIEJO, USA: QLogic Corp. announced that a cluster using NVIDIA Tesla graphics processing units (GPUs), QLogic InfiniBand switches and adapters and operated by the National Center for Supercomputing Applications (NCSA) achieved a number three ranking for MFlops/watt on the Green500 list of the world's top supercomputers.

Launched in 2006, the Green500 ranks the most energy-efficient supercomputers in the world using performance-per-watt to encourage HPC vendors and users to deploy more cost-effective computing systems. The NCSA's hybrid cluster incorporated Intel Core i3 2.93Ghz dual core processors with NVIDIA Tesla C2050 GPUs and QLogic TrueScale InfiniBand solutions, producing a score of 933.06 MFlops/watt – nearly four times more efficient than average supercomputers.

"We wanted to expand the frontiers of computational science, and the combination of NVIDIA Tesla GPUs with QLogic TrueScale InfiniBand fabrics is enabling this exploration," said Prof. Wen-mei Hwu of Electrical and Computer Engineering at the University of Illinois at Urbana-Champaign. "In collaboration with NCSA's Innovative Systems Laboratory, the scaling and power efficiency from this combination of technologies has helped to place us near the top of the Green500 list."

"The high performance per watt of Tesla GPUs has made them the architecture of choice for modern supercomputing, as evidenced by the presence of several GPU supercomputers in the Top 10 of the Green500," said Sumit Gupta, product line manager, Tesla business at NVIDIA. "QLogic's TrueScale InfiniBand architecture, coupled with Tesla GPUs, is key to helping increase overall system performance and efficiency of the NCSA cluster."

QLogic's newest TrueScale InfiniBand software release does not require Linux kernel patches or special InfiniBand drivers to integrate with NVIDIA Tesla GPUs, making it very easy to install and maintain GPUs for HPC applications.

"Our interoperability with NVIDIA Tesla GPUs is another demonstration of how easily QLogic TrueScale InfiniBand fabrics support the most advanced cluster performance technologies," said Joe Yaworski, director of global alliances and solution marketing for QLogic. "These results show how our development teams are continually pushing the envelope to create the easiest, most efficient, and most scalable integration of technologies to make real-world differences in cluster scaling and efficiency."

InfiniBand networks powered by QLogic
QLogic offers a comprehensive, end-to-end portfolio of InfiniBand networking products for HPC, including QDR host channel adapters, QDR multi-protocol director and edge switches and intuitive tools to install, operate, and maintain high-performance fabrics.

Delivering the industry's highest message rate, lowest MPI latency and highest effective bandwidth, QLogic InfiniBand adapters enable MPI and TCP applications to scale to thousands of nodes with industry-leading price-performance. As the most comprehensive and flexible HPC cluster interconnect fabric solutions on the market, QLogic QDR InfiniBand switches incorporate the industry's only management tools that enable an administrator to install and boot an InfiniBand fabric in minutes, helping to accommodate the growing demand for high performance computational clusters and grids.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.