A Performance Comparison of HPCx Phase 2a to Phase 2

The results of benchmarking communications and several popular applications on HPCx Phase 2a are presented and compared with those results from running on HPCx Phase 2. In terms of communications, there is little performance dierence between the platforms. In general, the majority of application types perform up to a factor of 2 better on Phase 2a, due to the improved memory architecture. An exception is the classical molecular dynamics class of applications, several of which are studied in this paper. This type of code is not performing as well as expected on Phase 2a, sometimes slightly slower than on Phase 2. This may be explained by sensitivity to an increased latency in some part of the memory subsystem.