Abstract: The authors evaluated the performance of two versions of the Grid-point Atmospheric General Circulation Model of IAP/LASG (GAMIL1.0 and 2.0), by comparing the climatological patterns of the cloud-radiative forcings (CRFs), i.e., shortwave cloud radiative forcing (SF), longwave cloud radiative forcing (LF), and their ratio N=-SF/LF, as well as their contrasts between the 1997/98 El Niño years and the normal years over the tropical Pacific region in both the observations and simulations. The result shows that the spatial distributions of multi-year averaged cloud radiative forcing (CRF) and the anomalies of the strong 1997/98 El Niño over the tropical Pacific region is well produced by GAMIL1.0, especially over the western Pacific warm pool region. However, the magnitudes of climatological SF, LF, and their ratio N are all overestimated, and the differences between the strong 1997/98 El Niño years and the normal years are underestimated. Although GAMIL2.0 has better spatial patterns than GAMIL1.0, it loses the anomalies of strong 1997/98 El Niño in the warm pool. Analysis shows that the discrepancies in CRFs are caused by unrealistic cloud vertical structure (i.e. overestimates in deep convective clouds and middle clouds while underestimate in cirrus), in-cloud water path, and high cloud fraction. It is found that the simulation of LF is better than that of SF in GAMIL1.0. The better simulation of LF results from the opposite effect of biases in cloud vertical structure and high could fraction to LF and the joint effect of biases to SF. In GAMIL2.0, the overestimated in-cloud water path compensates the contributions of underestimated cloud cover and then leads to stronger SF. However, it has fewer effects on LF that mainly reflect the impact from the underestimated cloud cover. As a result, the value of N is overestimated by these two versions. In addition, the large discrepancies in the clear-sky fluxes at the top of atmosphere in GAMIL also lead to the overestimation of N value and the SF.