It is proposed a parallel Monte Carlo algorithm to simulate templated grain growth in sintering ceramics materials. The algorithm applies the general Potts model to treat the matrix as the discrete lattices for simulating the grain growth and there will be a number of lattices to be computed synchronously. The scheme is performed by CUDA GPU parallelization programming framework which is of much more feasibility and low cost comparing with the former conventional program. The most key point is that the parallel algorithm is of great temporal performance which means it takes less time to complete a simulation. The results of comparative experiments show that the algorithm is unquestionable effective while the other statistic numerical features of simulations are almost the same.