Scientific Application Based Performance on Magic Cube
Magic Cube at Shanghai Supercomputing Center is the fastest supercomputer in east Asia. It is configured with 1920 16-Way SMP compute nodes, which gives 233 TFLOPS theoretical peak performance. The compute nodes are connected using Infinib and switches. In this study, we use STREAM micro-benchmark, r andom access in HPCC benchmark suite to underst and the memory performance of Magic Cubes SMP compute node. The multi-zone version of NAS Parallel Benchmark suite (NPB-MZ) servers as the application performance evaluation tools.
Ying Xu D.D. Zhang Lei Xu
Shanghai Supercomputing Center, Shanghai, China, 201203
国际会议
The Second International Conference on High Performance Computing and Applications(第二届高性能计算及应用国际会议)
上海
英文
485-490
2009-08-10(万方平台首次上网日期,不代表论文的发表时间)