This paper describes the usage of shaders to make parallel operations by improving the CPU–GPU communication, using both rendering and compute shaders. When the number of spheres is large, the execution becomes slow and requires a lot of space to store particles. We parallelized the frozen method using an efficient inter-process GPU communication to reduce the operations required. We propose to handle two buffers, one for operations and the other for rendering. While the rendering buffer increases as the number of spheres is required, the operational buffer maintains its size and the number of operations is hold. Experimental results demonstrate that the proposed method shows up 100x throughput improvement over the sequential version. We define a configuration by a 4-tuple as input to the algorithm, and we found a pattern to choose better configurations.