Performance and energy efficiency in material science simulation on heterogeneous architectures

In HPC applications, the energy efficiency is becoming more and more important, due to architectural constraints. It is therefore of primary interest to measure and evaluate the energy efficiency of current architectures using typical HPC workloads. One of the most used and appreciated codes publicly available for computational material science simulation, and largely used in many high end HPC system is Quantum ESPRESSO [1]. In this work we present the results of a set of benchmarks of Quantum ESPRESSO on the heterogeneous EURORA Supercomputer, featuring both NVIDIA K20 GPU and Xeon PHI co-processors, and measuring the energy to solution, thanks to an innovative monitoring system recently deployed on EURORA.