using 1 049 088 cells the application of Conway Game of Life using SFML and CUDA 9.0 run at 5 FPS.And that has to do mostly with the heavy cost of communication between the Host (CPU) and Device (GPU), CUDA Kernels cannot launch Host functions, so at each time the new cells stat matrix is calculated on the GPU it need to be transferred back to the host so it can call some OpenGl libs that would call the GPU so it can be rendered, we can calculate multiple stat matrices then send them in chunks to the host but that has been proven to be too memory inefficient due to the large size of the matrices.