The time and amplitude characteristics of the transient power overshoot caused by channels adding up are studied both experimentally and numerically in the backward-pumped Raman amplified wavelength division multiplexing (WDM) system. The results show that the shorter the signal wavelength is, the smaller the last time of the transient power overshoot will be. The maximal power overshoot is small and does not deadly damage the optical component, and the channel distribution of the maximal value of the power overshoot has similar curve in shape with that of the system ON-OFF gain.