Abstract High Performance Computing (HPC) system is an important part of a modern SAR processing system. SAR imaging has its specific characteristics compared with other HPC problems. It performs with great difference on different HPC architectures. The effects on SAR imaging performance by every part of the HPC architecture are analyzed on this paper. Based on this analysis, some benchmarking could be made to more accurately predict the SAR imaging performance on a multi-nodes environment. Prom these benchmarks, the effective way to optimize the SAR imaging could be found.