The purpose of the present paper is to evaluate the validity of CFD, especially Large Eddy Simulation(LES), to simulate the OECD/NEA benchmark experiment MATi S-H. The 5×5 rod bundle with a split-type spacer grid is simulated by both steady and transient simulation and the numerical results of time averaged and root-mean-square velocity field are compared to the experimental data. It is demonstrated that LES can better describe the mixing turbulent flow velocity field than the RANS method. The mesh sensitivity analysis further reveals that LES with fine mesh is more appropriate for analyzing the turbulent flow field in the rod bundle with spacer grids. LES results help understand the mixing turbulence characteristics with the mixing effect evaluation in transient. The simulated flow field data can also be of value for mechanical evaluations such as flow-induced vibration and grid-to-rod-fretting phenomenon.