Previous approaches using active membrane systems to solve the N-queens problem defined many membranes with just one rule inside them. This resulted in many communication rules utilised to communicate between membranes, which made communications between the cores and the threads a very time-consuming process.The proposed approach reduces unnecessary membranes and communication rules by defining two membranes with many objects and rules inside each membrane. With this structure, objects and rules can evolve concurrently in parallel, which makes the model suitable for implementation on a Graphics processing unit(GPU). The speedup using a GPU with global memory for N=10 is 10.6 times, but using tiling and shared memory, it is 33 times. Previous approaches using active membrane systems to solve the N-queens problem defined many membranes with just one rule inside them. Which resulted in communications between the cores and the threads a very time-consuming process The proposed approach reduces unnecessary membranes and communication rules by defining two membranes with many objects and rules inside each membrane. With this structure, objects and rules can evolve concurrently in parallel, which makes the model suitable for implementation on a Graphics processing unit (GPU The speedup using a GPU with global memory for N = 10 is 10.6 times, but using tiling and shared memory, it is 33 times.