摘要

With the Inspur TS1000 HPC Server of Tsinghua University, a lot of calculations have been done based on the NEA Data Bank full-core benchmark model and EDF 3D pressurized water reactor (PWR) full-core calculations through large-scale paralleling. The performance of MCNP which is the widely used general Monte Carlo code and RMC which is used for reactor analysis and developed by Tsinghua University is compared systematically. It is found that MCNP is unable to calculate the local power density of full-core reactors at the required accuracy because its limits on the parallel model, the performance of tallying and so on, while RMC is well-suited owing to its improvement on those limits. Thus, it can be concluded that the performance of RMC is better than MCNP in terms of detailed power density distributions calculation in full-core reactors.