Performance and accuracy of criticality calculations performed using WARP – A framework for continuous energy Monte Carlo neutron transport in general 3D geometries on GPUs

•WARP has been developed and benchmarked against production codes MCNP and Serpent.•WARP produces statistically identical results for a number of test cases.•WARP results also differ significantly for the most geometrically complex cases.•WARP performs like 1-7 supercomputer nodes, depending on hardware and problem type.•Hardware and running cost savings could also be realized with WARP.•Improvements have been made compared to previous publications about WARP. In this companion paper to “Algorithmic Choices in WARP – A Framework for Continuous Energy Monte Carlo Neutron Transport in General 3D Geometries on GPUs” (doi:http://dx.doi.org/10.1016/j.anucene.2014.10.039), the WARP Monte Carlo neutron transport framework for graphics processing units (GPUs) is benchmarked against production-level central processing unit (CPU) Monte Carlo neutron transport codes for both performance and accuracy. Neutron flux spectra, multiplication factors, runtimes, speedup factors, and costs of various GPU and CPU platforms running either WARP, Serpent 2.1.24, or MCNP 6.1 are compared. WARP compares well with the results of the production-level codes, and it is shown that on the newest hardware considered, GPU platforms running WARP are between 0.8 and 7.6 times as fast as CPU platforms running production codes. The GPU platforms running WARP are also shown to be between 15% and 50% as expensive to purchase and between 80% and 90% as expensive to operate as equivalent CPU platforms performing at an equal simulation rate.