In recent years downsized turbocharged engines with gasoline direct injection are becoming more popular in light duty vehicle applications for better performance and fuel economy. However, engine knock is one of the major problems when developing and calibrating a turbocharged engine. Typically higher boost pressure induced by turbocharger causes higher tendency of engine knock, because unburned in-cylinder gas temperature increased as boost pressure is raised. To take full advantage of the turbocharged engine, understanding of knock characteristics and optimization of design and control parameters is important. Therefore, an engine simulation model that is capable of predicting knock can be a useful tool in developing improved engine designs. It is also important to pursue the balance between accuracy and computational expenses for practical engine design and control optimization. In this study, a knock prediction model is developed based on knock combustion modeling for SI engine combustion model. Without an appropriate knock combustion model, the severity of abnormal combustion due to auto-ignition and accordingly the knock intensity cannot be judged properly. In this proposed methodology, the auto-ignition of the end-gas is predicted with a two-stage ignition delay model. After the auto-ignition, combustion of the end-gas is calculated using a new concept of multi-zone model. One of essential parts of the methodology is a newly developed knock index model which processes calculated cylinder pressure from the knock combustion. The index model provides a measure of the knock intensity and it can be used to determine the knock limited MBT spark timing. The knock prediction model is implemented into GT-Power® as a user subroutine for a versatile simulation capability. Experimental data from various engine conditions are processed to obtain auto-ignition angle, auto-ignition probability, representative pressure and heat release rate data for knock occurring cases. The model is calibrated with the experimental data and validated with various knocking and non-knocking cases.