NICS Accelerates HPC Performance with Intel® Technology

Optimizing applications with a cluster based on Intel® Xeon Phi™ coprocessors

Leider ist diese PDF-Datei nur im Download-Format verfügbar.

Accelerating Science and Engineering Application Performance

With funding from the National Science Foundation (NSF) and the University of Tennessee (UT), along with assistance from Intel, the National Institute for Computational Sciences (NICS) built Beacon—a cluster designed to help researchers optimize science and engineering applications for the Intel® Many Integrated Core Architecture (Intel® MIC Architecture). Nodes are equipped with the Intel® Xeon® processor E5 family and Intel® Xeon Phi™ coprocessors. By optimizing code, the NICS team and associated researchers found they could achieve more than 2.25 times better performance using a single Intel Xeon Phi coprocessor compared with two processors from the Intel Xeon processor E5 family. Building similar clusters will enable researchers to solve larger, more complex problems while controlling costs.

Challenges
• Optimize performance. Provide a high-performance computing (HPC) cluster that would enable researchers to optimize their code for the Intel MIC Architecture.
• Control costs. Find ways to build more efficient clusters that can help researchers solve larger, more complex problems without significantly increasing hardware acquisition expenses, energy consumption, or software development costs.

Solutions
• Cray CS300-AC* cluster supercomputer with the Intel Xeon processor E5 family and Intel Xeon Phi coprocessors. Beacon comprises 48 compute nodes and 6 I/O nodes with a total of 768 conventional cores and 11,520 accelerator cores. Compute nodes are equipped with Intel Xeon processors E5-2670 and Intel Xeon Phi coprocessors 5110P. Intel® Solid-State Drives (Intel® SSDs) are integrated into the storage environment.
• Intel® Software Development Tools. Software developers used the Intel® Cluster Studio XE suite to optimize code.

Technology Results
• Increased performance. Optimized code running on an Intel Xeon Phi coprocessor can deliver more than 2.25 times better performance than with two Intel Xeon processors E5-2670.

Read the full Accelerating Science and Engineering Application Performance Case Study.