The performance of IAP (Institute of Atmospheric Physics) Atmospheric General Circulation Model Version 4.0 (IAP AGCM4.0) in simulating the Madden-Julian Oscillation (MJO) is examined in this paper using the 30-year model integration results during 1979-2008. It is found that the IAP AGCM4.0 can reproduce the observed wave number-frequency power spectrum of MJO to some extent, with dominant spectrum power at wavenumber 1 and periods of 30-80 days. Meanwhile, the IAP AGCM4.0 can generally reproduce the observed coherent eastward propagating signals at the intraseasonal time scale, with the power of eastward moving waves much stronger than that of the westward moving waves. The RMM (Real-time Multivariate MJO) index is further applied to evaluate the simulated MJO structure. It is found that IAP AGCM4.0 can well reproduce the observed intraseasonal signals of 850 hPa and 200 hPa zonal winds and the enhanced convection structure of MJO in the tropical regions. However, the simulated eastward propagation is generally too fast, and the simulated westward propagation is stronger than the observation. IAP AGCM4.0 also splits the intraseasonal convective anomalies into two centers straddling the equator, and produces weaker convection. The vertical profile of diabatic heating simulated by the IAP AGCM4.0 has a similar structure to the observation, but in the Indian Ocean and western Pacific Ocean, positive maximum heating occurs later than the observation in phases. Numerical experiments are conducted by using different RHc (relative humidity criterion) values of 85%, 90%, 95%, and 100% for triggering the convection. It is found that the vertical diabatic heating profiles for experiments with different RHc vary considerably, which can lead to differences in the simulated MJO features. Comparison of results further shows that both the main features of MJO and vertical diabatic heating profiles are best simulated when RHc is set to 90%. This suggests that proper specifications of the values for key parameters in the convective parameterization scheme might help improve the model capability in simulating the observed features of MJO.