This paper presents a performance examination of the Coarse Mesh Finite Difference (CMFD) method for a reduction of inter-cycle correlations in Monte Carlo (MC) active cycle simulations. Sensitivity tests using the UNIST in-house MC code 'MCS' reveal the characteristics of CMFD effectiveness depending on problem sizes, dimensions, and cross-section dependency on neutron energy (multi-group vs. continuous). Also, by applying this method to the three-dimensional BEAVRS whole core problem, it is successfully demonstrated that the CMFD can reduce the inter-cycle correlations of practical, large size, and high dominance ratio problems, and that the figure of merit of efficiency for a practical reactor problem increases by a factor of six.