Benchmarking Radiation Transport Monte Carlo Simulations with MCNP and Geant4 Using High Performance Computing.

The objective of this paper is to compare the performance of several high-performance computing systems in order to inform decisions regarding their use for Monte Carlo simulations of radiation transport. Gamma ray emission from 131 I in the human thyroid and detection using a personal radiation detector were modeled using the MCNP and Geant4 Monte Carlo software. These simulations were benchmarked by recording the computing time needed to run the simulation as a function of the number of parallel computing threads used. Simulations were run using a virtual machine, two desktop PCs, a CX-1 supercomputer, the Government of Canada General Purpose Science Cluster, and cloud computing. Using a higher number of parallel threads on these high-performance computing systems was found to reduce the computing time needed to run the MCNP and Geant4 simulations. The optimal configuration for running the simulations on cloud computing was evaluated, considering the number of available processors, the computing time, and the cost. Cloud computing was found to be a cost-effective, on-demand, high performance computing option for Monte Carlo simulations.