During the solidification of alloys, numerical simulation of forced convection eutectic growth is essential for accurate prediction and control of solidified microstructure. The phase field lattice Boltzmann model (PF-LBM) is established by coupling the phase field model of KKSM model with the LBM model of computational fluid. In order to solve the problems of PF-LBM model, which has a large amount of computation, the simulation region is small and can not be solved in a reasonable time. The parallel computing architecture based on OpenCL can give full play to the powerful computing power of NVIDIA GPU and accelerate the implementation of the algorithm to improve the efficiency of numerical simulation. The effect of forced convection on the eutectic growth of CBr4-C2Cl6 eutectic alloy is systematically studied. The results showed that liquid flow altered the eutectic evolution by affecting the distributions of the solute ahead of the eutectic solid/liquid interface. Compared with serial CPU code, a single GPU can achieve a maximum acceleration ratio of 136.X. Solved the problem of 3-D numerical simulation of forced convection eutectic growth due to computational limitations.