Abstract: In order to improve the computational efficiency of BCC_CSM climate system model, porting and tuning work of BCC_CSM on IBM high performance computing system were carried out. According to the requirement of climate prediction production, the model computational efficiency was greatly improved by tuning. With the help of distributed situation of climate variables’ field and quantitative indicators, model simulation performance was verified. The results indicate that computational speed from the modified model is 1.4 times faster than that of the original one. Annual mean surface air temperature distribution is reasonable in ten years (531-540) based on climate simulation of a CMIP5 piControl experiment, and relative error is less than 0.5%. Both of computational efficiency and simulation performance of the BCC_CSM model are met application requirements.