The problem of predicting the thermal performance of engine cooling systems is becoming increasingly difficult as more severe requirements are imposed on the cooling system. The overall capacity of the cooling system is adequate. However, local conditions cause excessive temperatures. To improve prediction accuracy, it is necessary to improve the ability to estimate local heat transfer coefficients. Although adequate correlations exist, the problem is the ability to predict local flow conditions. The presence of vapor bubbles will affect the pressure drop in the cooling channels and therefore the flow distribution. The direct approach by taking experimental data with the real engine is first attempted. To characterize effectively the convection and nucleate boiling heat transfer coefficient, the entire engine has been installed in the laboratory specially designed for engine testing at Western Michigan University. The experimental data with the mechanisms of nucleate boiling play a crucial role to provide a method how to implement the convective and nucleate boiling heat transfer coefficient developed to the real engine cooling systems as well as the mathematical model. The computations using a CFD software Fluent have been also performed in our laboratory at WMU. The three-dimensional numerical computations along with the coolant channel have provided the useful information such as the local pressures, velocities and wall-surface temperatures under an assumption of no nucleate boiling.