Automatic transmission torque converter is usually used as a power transmission element. As the reason, torque converter is the function of the torque matching and the torque amplification and the engine. This is referred as the fluid performance of torque converter. The fluid performance of torque converter is determined by the blade shape of the torque converter. Therefore it is necessary to predict the fluid performance of the torque converter at the design stage to determine the blade shape, Computational Fluid Dynamics (CFD) analysis is used predict the fluid performance. Generally, time-averaged turbulence models such as K-ε(Reynolds Averaged Navier Stokes (RANS model)) is often used in this CFD analysis. In this RANS model, since the analysis time exists shorter calculation examples is not limited to the torque converter , they are too used for industrial purpose. However RANS model is a single model such as pipes based model, so when using the complex 3D flow field, such as a torque converter and calculation object a reliability of its prediction accuracy is not enough yet. Therefore with respect to this issue the Large Eddy Simulation (LES) which can directly treat unsteady phenomena of turbulence in such complex flow fields is used which also seen in the industrial scene. However, LES turbulence model is from the abundance of the amount of calculation so mainly used in research utilization. In this research an open-licensed software "FrontFlow/red" specially developed for high-performance computing enables to perform the LES turbulence model for the torque converter performance prediction. Some numerical results and their validations are reported.