Efficiency Analysis of Intel, AMD and Nvidia 64-Bit Hardware for Memory-Bound Problems: A Case Study of Ab Initio Calculations with VASP

Nowadays, the wide spectrum of Intel Xeon processors is available. The new Zen CPU architecture developed by AMD has extended the number of options for x86_64 HPC hardware. Moreover, Nvidia has released a custom 64-bit Denver architecture based on the ARM instruction set. This large number of options makes the optimal CPU choice for perspective HPC systems not a straightforward procedure. Such a co-design procedure should follow the requests from the end-users community. Modern computational materials science studies are among the major consumers of HPC resources worldwide. The VASP code is perhaps the most popular tool for these research. In this work, we discuss the benchmark metric and results based on a VASP test model that give us the possibility to compare different hardware and to distinguish the best options with respect to energy-to-solution criterion.

[1]  Raffaele Tripiccione,et al.  Energy-Performance Tradeoffs for HPC Applications on Low Power Processors , 2015, Euro-Par Workshops.

[2]  Vladimir V. Stegailov,et al.  HPC Hardware Efficiency for Quantum and Classical Molecular Dynamics , 2015, PaCT.

[3]  Lars Koesterke,et al.  PerfExpert: An Easy-to-Use Performance Diagnosis Tool for HPC Applications , 2010, 2010 ACM/IEEE International Conference for High Performance Computing, Networking, Storage and Analysis.

[4]  Torsten Hoefler,et al.  Scientific Benchmarking of Parallel Computing Systems Twelve ways to tell the masses when reporting performance results , 2017 .

[5]  Kresse,et al.  Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. , 1996, Physical review. B, Condensed matter.

[6]  Gary Brown,et al.  Denver: Nvidia's First 64-bit ARM Processor , 2015, IEEE Micro.

[7]  Mariana Rossi,et al.  Ab initio molecular dynamics , 2013 .

[8]  James C. Browne,et al.  Enhancing Performance Optimization of Multicore/Multichip Nodes with Data Structure Metrics , 2014, ACM Trans. Parallel Comput..

[9]  Vladimir V. Stegailov,et al.  Floating-point performance of ARM cores and their efficiency in classical molecular dynamics , 2016 .

[10]  Andreas C. Döring,et al.  Power measurements and cooling of the DOME 28nm 1.8GHz 24-thread ppc64 μServer compute node , 2015, 2015 International Conference on IC Design & Technology (ICICDT).

[11]  Vladimir V. Stegailov,et al.  Efficiency of the Tegra K1 and X1 systems-on-chip for classical molecular dynamics , 2016, 2016 International Conference on High Performance Computing & Simulation (HPCS).

[12]  Vladimir V. Stegailov,et al.  GPU-Accelerated Molecular Dynamics: Energy Consumption and Performance , 2016 .

[13]  John Shalf,et al.  Exascale Computing Trends: Adjusting to the "New Normal"' for Computer Architecture , 2013, Computing in Science & Engineering.

[14]  G. Kresse,et al.  Ab initio molecular dynamics for liquid metals. , 1993 .

[15]  Thomas Scogland,et al.  Node variability in large-scale power measurements: perspectives from the Green500, Top500 and EEHPCWG , 2015, SC15: International Conference for High Performance Computing, Networking, Storage and Analysis.

[16]  Lucas Mello Schnorr,et al.  Characterizing the Performance of Modern Architectures Through Opaque Benchmarks: Pitfalls Learned the Hard Way , 2017, 2017 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW).

[17]  G. Kresse,et al.  Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set , 1996 .

[18]  Leonel Sousa,et al.  Energy‐aware mechanism for stencil‐based MPDATA algorithm with constraints , 2017, Concurr. Comput. Pract. Exp..

[19]  Hafner,et al.  Ab initio molecular-dynamics simulation of the liquid-metal-amorphous-semiconductor transition in germanium. , 1994, Physical review. B, Condensed matter.