Based on the LS-Dyna neon_refined_revised TopCrunch benchmark, Scali and Mellanox were able to compare SMP implementations and MPI-based versions of the same application. The Scali MPI Connect based-version, running in conjunction with Mellanox 20Gb/s InfiniBand interconnect, delivered superior performance than alternatives. Additionally, on a jobs per day basis, the data showed that the distribution of the applications can significantly increase throughput. By example, running two jobs across 16 nodes, with 4 cores per node dedicated to each job delivered 43% greater jobs per day than the same jobs run serially with all eight cores dedicated to a single job. The benchmarks results were obtained on the "Helios" cluster (Mellanox Cluster Center).

"Multi-core cluster environments impose high demands on the interconnect and the software protocol", stated Thad Omura, Vice President of Product Marketing at Mellanox Technologies. "The performance, reliability and scalability of Mellanox 20Gb/s InfiniBand, and the efficiency of Scali MPI are essential for driving maximum application performance out of multi-core server platforms."

"As more multi-core nodes with larger core counts are delivered, the ability to balance a system to maximize total throughput is becoming critical", stated Hakon Bugge, CTO of Scali. "Working with Mellanox, we have been able to help define key criteria for understanding how to get the most from these new types of environments."

Scali MPI Connect is a fully integrated message passing interface (MPI) solution, designed to enable applications to run at maximum performance through its unique, high performance MPI implementation. With Scali MPI Connect, application developers can take advantage of leading interconnect technologies to build high performance clusters. Scali MPI Connect supports a wider range of customer environments while lowering the number of binaries required, and delivers faster performance through application enhancement.