The opinion evolution results of both dynamic models of group experts round discussion and free discussion in collective discussion are analyzed and compared by computer simulation. Statistical results show that under the same initial conditions, the model of free discussion using the speech sequence rule of authoritative experts first and then experts with the maximum of opinion cumulative change first has less average opinion clusters. It also finds that discussions after 15 rounds have less improved impact on indicators. As the free discussion is followed by serial speeches that are less efficient, in the actual selection of the mode of discussion, when there are fewer experts and plenty of time, it should use free discussion, and vice versa it should adopt discussion by round.