Cellular genetic algorithms (cGAs) are mainly characterized by their spatially decentralized population, in which individuals can only interact with their neighbors. In this work, we study the behavior of a large number of different cGAs when solving the well-known 3-SAT problem. These cellular algorithms differ in the policy of individuals update and the population shape, since these two features affect the balance between exploration and exploitation of the algorithm. We study in this work both synchronous and asynchronous cGAs, having static and dynamically adaptive shapes for the population. Our main conclusion is that the proposed adaptive cGAs outperform other more traditional genetic algorithms for a well known benchmark of 3-SAT.