Two surface wind stress datasets for 1979–91, one based on observations and the other from an integration of the COLA atmospheric general circulation model (AGCM) with prescribed SST, are used to drive the GFDL ocean general circulation model. These two runs are referred to as the control and COLA experiments, respectively. Simulated SST and upper-ocean heat contents (HC) in the tropical Pacific Ocean are compared with observations and between experiments. Both simulations reproduced the observed mean SST and HC fields as well as their annual cycles realistically. Major errors common to both runs are colder than observed SST in the eastern equatorial ocean and HC in the western Pacific south of the equator, with errors generally larger in the COLA experiment. New errors arising from the AGCM wind forcing include higher SST near the South American coast throughout the year and weaker HC gradients along the equator in boreal spring. The former is associated with suppressed coastal upwelling by weak alongshore AGCM winds, and the latter is caused by weaker equatorial easterlies in boreal spring. The low-frequency ENSO fluctuations are also realistic for both runs. Correlations between the observed and simulated SST anomalies from the COLA simulation are as high as those from the control run in the central equatorial Pacific. A major problem in the COIA simulation is the appearance of unrealistic tropical cold anomalies during the boreal spring of mature El Nin˜o years. These anomalies propagate along the equator from the western Pacific to the eastern coast in about three months, and temporarily eliminate the warm SST and HC anomalies in the eastern Pacific. This erroneous oceanic response in the COIA simulation is caused by a reversal of the westerly wind anomalies on the equator, associated with an unrealistic southward shift of the ITCZ in boreal spring during El Nin˜o events. In both experiments, the annual cycle of the upper equatorial ocean (upper 45 m) is dominated by the annual change of surface heat flux and advection with its vertical and horizontal components dominant in the eastern and central Pacific, respectively. The El Nin˜o development, on the other hand, is mainly due to advection of anomalous zonal current in the west, the meridional advection in the central, and the displacement of thermocline in the east Pacific. The weakening of the warm anomalies in COIA in the middle of major warm events corresponds to changes of all these processes. The implication of the results derived from this study to the coupled ocean-atmosphere general circulation models is discussed.