To find the distribution patterns of dynamic amplification coefficients for dams subjected to earthquake, 3D seismic responses of concrete-faced rockfill dams with different heights and different shapes of river valley were analyzed by using the equivalent-linear model. Statistical analysis was also made to the seismic coefficient, and an empirical formula for calculating the maximum acceleration was provided. The results indicate that under the condition of the same dam height and the same base acceleration excitations, with the increase of the river valley width, the position of the maximum acceleration on the axis of the top of the dam moves from the center to the riversides symmetrically. For the narrow valleys, the maximum acceleration occurs in the middle of the axis at the top of the dam; for wide valleys the maximum acceleration appears near the riversides. The result negates the application of 2D dynamical computation for wide valleys, and shows that for the seismic response of high concrete-faced rockfill dams, the seismic coefficient along the axis should be given, except for that along the dam height. Seismic stability analysis of rockfill dams using pseudo-static method can be modified according to the formula. To find the distribution patterns of dynamic amplification coefficients for dams subjected to earthquake, 3D seismic responses of concrete-faced rockfill dams with different heights and different shapes of river valley were analyzed by using the equivalent-linear model. Statistical analysis was also made to the seismic coefficient, and an empirical formula for calculating the maximum acceleration was provided. The results that under the condition the the dam same and the same base acceleration excitations, with the increase of the river valley width, the position of the maximum acceleration on the axis of the top of the dam moves from the center to the riversides symmetrically. For the narrow valleys, the maximum acceleration occurs in the middle of the axis at the top of the dam; for wide valleys the maximum acceleration appears near the riversides. The result negates the application of 2D dynamical computation for wide valleys, and shows that for the seismic response o f high concrete-faced rockfill dams, the seismic coefficient along the axis should be given, except for that along the dam height. Seismic stability analysis of rockfill dams using pseudo-static method can be modified according to the formula.