In order to solve the problem of slow computation of point source model, we designed a real-time computer holographic generation system based on a multi-core CPUs and graphics processing unit (GPU). This system makes full use of the GPU's powerful parallel computing capabilities and CPU logic computing capabilities. It has been verified through experiments that the system is effective and feasible. At the same time, we use the Compute Unified Device Architecture (CUDA) platform to program an algorithm for the parallel computation of holograms in a graphics processing unit. In this paper, we have implemented a point source model to generate compute-generated holograms. We also compared computational performance in CPUs, GPUs, multi-core CPUs and GPUs. Among them, the multi-core CPU and GPU systems have the fastest computational holograms, which can at least increase the hologram calculation speed by 120 times compared with the equivalent CPU system, and also can increase the speed of calculation by 2 to 10 times compared with the GPU system. Therefore, the system which we designed provides a new method for real-time calculation of holograms.