Model outputs of a high resolution global coupled ice-ocean model with 1/6 by 1/4 degrees and 48 vertical layers running on the ‘Earth Simulator’ supercomputer were evaluated to determine the model stability, physics soundness, and its sensitivity to different process parameterizations. The coupled model consists of the 3-D s-coordinate ocean model of the Tokyo University and the Las Alamos sea ice model (CICE). The massive data output used in this study was sorted by model year (27 to 38) and experiments with or without GM (Gent and McWiliams, 1990) parameterization to the north of 40N. The statistical time series of the total oceanic and ice kinetic energy and ice areas suggest that the model is adjusting towards stability without surface T/S restoring. The model climatology (mean over all the model years) was examined and compared with the seasonal cycle of available observed climatology, such as ice area, temperature and salinity at certain key depths. The mixed layer depth and convection depth shows significant differences between the results with and without GM scheme, so does the deep water properties, such as the Intermediate Atlantic Water in the Arctic Ocean, which suggests that GM parameterization of cross-isopyconal diffusion is important to prevent deep water properties from excessive numerical diffusion in a climate model. Several important physical features in the northern hemisphere, such as the thermohaline in the Arctic Ocean, Atlantic Water, meridional thermohaline overturning, transport at Bering Strait, Fram Strait etc., were examined to determine physical soundness of the model.