Modern ships are usually redundantly equipped with one-axis indexing ring laser gyro (RLG) marine inertial navigation system (MINS) and several two-axis indexing RLG MINS. In order to online evaluate the relative performance of the redundant two-axis indexing RLG MINS, an online evaluation method based on joint rotation and modulation was proposed without using external benchmark information. Several joint error Kalman filters were constructed. The system states of each Kalman filter included the position error differences, the velocity error differences and the attitude error differences between the one-axis indexing RLG MINS and corresponding two-axis indexing RLG MINS, as well as the gyro drifts and horizontal accelerometer biases. With the position and velocity difference between the systems being observations, the observability analysis showed that all states including azimuth gyro drift of the one-axis indexing MINS were observable if the relative attitude between the systems was changed by the joint rotation and modulation. The azimuth gyro drift estimation standard deviation of the one-axis indexing MINS were defined as the assessment criteria to online evaluate the random errors of the two-axis indexing MINS. The semi-physical simulations and the experimental results show that the RLG random error difference of the two-axis indexing MINS can be distinguished at a level of 10% noise and the position accuracies of different two-axis indexing MINS are online evaluated according to the azimuth gyro drift standard deviation of the one-axis indexing MINS during 144 h navigation time. The proposed method provides theoretical basis for optimum system choosing in the case of one-axis and two-axis indexing RLG MINS redundant configuration.