This article empirically assesses a coarse-fine approach for diversity tuning in cellular Genetic Algorithms (cGAs). The coarse tuning is performed through the constant reconfiguration of the grid while the fine tuning is locally achieved through dynamic anisotropic selection which considers individuals' locations in the local neighborhood. Benchmark problems including continuous, real-world and combinatorial problems are evaluated. The experimental results show an improvement in cGAs performance when compared to having a fixed topology configuration or to independently applying dynamic lattice reconfiguration or dynamic anisotropic.