The validation of complex nonlinear numerical models such as the National Center for Atmospheric Research thermospheric general circulation model (NCAR TGCM) requires a detailed comparison of model predictions with data. The Equinox Transition Study (ETS) of September 17–24, 1984, provided a unique opportunity to address the verification of the NCAR TGCM, since unusually large quantities of high-quality thermospheric and ionospheric data were obtained during an intensive observation interval. In a companion paper (paper 1) by Crowley et al. (this issue) a simulation of the September 18–19 ETS interval was described. Using a novel approach to modeling, the TGCM inputs were tuned where possible with guidance from data describing the appropriate input fields, and the arbitrary adjustment of input variables in order to obtain thermospheric predictions which match measurements was avoided. In the present paper the winds, temperatures, and densities predicted by the TGCM are compared with measurements from the ETS interval. In many respects, agreement between the predictions and observations is good. The quiet day observations contain a strong semidiurnal wind variation which is mainly due to upward propagating tides. The storm day wind behavior is significantly different and includes a surge of equatorward winds due to a global propagating disturbance associated with the storm onset. The density data confirm the existence of the newly discovered four-cell high-latitude density anomaly described in paper 1. A quantitative statistical comparison of the predicted and measured winds indicates that the equatorward winds in the model are weaker than the observed winds, particularly during storm times. This is consistent with predicted latitudinal temperature gradients and storm time density increases which are much smaller than the observed values. Soft particle precipitation or high-altitude plasma heating is invoked as a possible source of the additional high-latitude heating required by the model. A quiet day phase anomaly in the measured F region winds which is not reproduced by the model suggests the occurrence of an important unmodeled interaction between upward propagating semidiurnal tides and high-latitude effects. Wind data from altitudes below 100 km indicate shortcomings in the generic equinox solar minimum tidal specification used in the TGCM. The lack of appropriate data to specify input fields seriously impairs our ability to generate realistic global thermospheric simulations. The problem is particularly acute in the southern hemisphere. Furthermore, disturbances generated in the southern hemisphere simulation propagate northward and degrade the northern hemisphere predictions. Thus, if the thermosphere in the northern hemisphere is ever to be fully understood, the southern hemisphere needs to be observed and simulated more accurately. Several improvements are suggested for future realistic time-dependent simulations of specific intervals.