Based on comparison with observational and reanalysis data, we assess the performances of two versions of the IAP/LASG Flexible Global Ocean-Atmosphere-Land System (FGOALS), FGOALS-g2 and FGOALS-s2, in simulating the climatology and interannual variability of South Asian summer monsoon (SASM). Moisture budget analysis is applied to explain the precipitation biases. FGOALS-g2 and FGOALS-s2 both underestimate precipitation over the continental monsoon trough but overestimate precipitation over the adjacent ocean. The northward seasonal migration of continental convergence zone is weaker than observation. The east-west sea surface temperature (SST) biases in the equatorial Indian Ocean (IO) simulated by FGOALS-g2 lead to weak southern intertropical convergence zone (ITCZ) over the eastern equatorial IO, while the south-north SST biases over the IO simulated by FGOALS-s2 result in southwestward shift of the ITCZ. Moisture budget analysis shows that precipitation biases in the FGOALS models are mainly attributed to the convergence of vertically integrated moisture flux biases, especially biases in the vertical dynamic moisture transport term. On the one hand, cold SST biases in the Arabian Sea and the Bay of Bengal along with warm SST biases in the tropical western IO reduce moisture flux over the Indian subcontinent in both models. On the other hand, cold biases of tropospheric temperature in the FGOALS models are most prominent in the upper troposphere over northern India. The FGOALS models also simulate weak longwave cloud radiative effects over the monsoon trough region due to their negative biases of cloud fraction over South Asia. The subsiding branches linked with the reduced meridional tropospheric temperature gradient and strengthened gross moist stability decrease climatological precipitation in the continental monsoon trough region. The FGOALS models cannot reasonably simulate the ENSO-SASM relationship at interannual time scale. The descending branch of the anomalous Walker circulation and corresponding negative precipitation anomalies are shifted to the tropical central-western IO to the south of the equator. The heating anomalies asymmetric about the equator enhance the northward cross-equatorial monsoon circulation and further cause erroneous positive precipitation anomalies over southern India. The shifts of the anomalous Walker circulation and negative precipitation anomalies are associated with the model biases in simulating climatological precipitation over the southern tropical IO. Our results show that reducing IO SST biases, tropospheric temperature biases and cloud biases is necessary for better simulation of mean state SASM by climate system models. On interannual time scale, the reasonable simulation of ENSO-monsoon relationship relies on successful simulation of climatological precipitation over the tropical IO and ENSO related circulation anomalies.