ABSTRACTThis paper will show a comparison between the Kepler, Maxwell and Pascal GPUarchitectures using CUDA-Fortran, with and without dynamic calls, to efficiently solve partialdifferential equations. The target is to show the possibility of using affordable hardware, such asthe GTX670, GTX970 and GTX1080 NVIDIA GPUs, which are commonly found in personaland portable computers, for scientific applications. For simplicity we consider a standardwave equation where we use a second order finite difference method for the spatial and timediscretizations to obtain the numerical solution. We found that, as we increase the spatialresolution of the domain we also increase the performance difference between the GPU and theCentral Processing Unit (CPU)