WHITE PAPER Intel® True Scale Fabric Architecture Supercomputing

Intel® True Scale Fabric Benchmark Conclusions

Three United States national laboratories, known for their work in supercomputing, recently benchmarked new systems delivered to each of them. These systems are built on the Intel® Xeon® processor E5-2600 family and Intel® True Scale Fabric, based on InfiniBand* and the open source Performance Scale Messaging (PSM) interface. The scientists performing the benchmarks concluded in individual reports that Intel True Scale Fabric:
• Contributed to “unprecedented scalability and performance” in their systems, and it is allowing them to change how they work.
• Outperforms on some tests one of the most powerful, customized supercomputers in the world, rated 18 in the November, 2012 Top500 list.
• Delivers a level of performance they had not seen from a commodity interconnect before.

The new systems, named Chama (at Sandia National Laboratories), Luna (at Los Alamos National Laboratory), and Zin (at Lawrence Livermore National Laboratory), are part of the Tri-Labs Linux* Capacity Clusters 2 (TLCC2) in the Advanced Simulation and Computing (ASC) program under the National Nuclear Security Administration (NNSA). This paper summarizes the findings of the reports from these three laboratories.

Read the full Intel® True Scale Fabric Benchmark Conclusions White Paper.