We performed an assessment of a database of synthetic ground motion generated by a suite of dynamic rupture simulations to verify compatibility of the peak ground amplitudes with respect to empirical ground‐motion prediction equations (GMPEs). The dynamic rupture model database, developed by Dalguer and Mai (2011), is composed of 360 earthquake scenarios with moment magnitudes in the range of 5.5–7, for three mechanisms of faulting (reverse, normal, and strike slip) and for both buried faults and surface rupturing faults, having depth‐ and non‐depth‐dependent normal stresses. Initial shear‐stress distribution follows a von Karman stochastic distribution. Overall, we show quantitatively that the upper frequency limit of the suite of simulations is 1.0 Hz. Up to this frequency, the synthetic data are compatible with the empirical model, which means that the residuals (defined as the differences between observed and predicted ground motions) fall in the range of ±σ of the empirical GMPE. Characteristics of the mean values along distance (10–45 km) and period T≥1.0 s, including standard deviation of the synthetic response spectra and peak ground velocity, are comparable to their counterpart empirical GMPE. At very near source (<1–2 km), synthetics show supersaturation of the mean peak values, which is different from the saturation features predicted by the GMPEs. The effect of source parameters (such as stress drop, peak slip velocity, and rupture speed) and of surface and buried rupture, as well as hanging wall and footwall, are highly sensitive to ground motion, suggesting that these effects contribute to the variability of ground motion near the source and that inclusion of them in the source terms may contribute to reducing uncertainties in GMPEs.