Nowadays Spark Ignition (SI) engine developments focus on downsizing, in order to increase the engine load level and consequently its efficiency. As a side effect, knock occurrence is strongly increased. The current strategy to avoid knock is to reduce the spark advance which limits the potential of downsizing in terms of consumption reduction. Reducing the engine propensity to knock is therefore a first order subject for car manufacturers. Engineers need competitive tools to tackle such a complex phenomenon. In this paper the 3D RANS simulations ability to satisfactorily represent knock tendencies is demonstrated. ECFM (Extended Coherent Flame Model) has been recently implemented by IFPEN in CONVERGE and coupled with TKI (Tabulated Kinetics Ignition) to represent Auto-Ignition in SI engine. These models have been applied on a single cylinder engine configuration dedicated to abnormal combustion study. This configuration includes all the characteristics of an up to date SI engine (turbocharged, direct injection). Based on the experimental data, combustion models are calibrated and validated. Unfortunately, direct comparison of the RANS and experimental pressure curves cannot be performed regarding knock. Therefore a specific methodology is developed in order to demonstrate the model ability to recover experimental trends. Over five operating points, spark advance sweeps are computed and lead to the following conclusions: (1) A computational knock index can be easily defined with ECFM-TKI (2) This approach is able to represent the increase of knock intensity with load (3) Its decrease with the engine speed (4) This computational knock index defining the knock limit can be used for iso-engine speed comparison (5) The computational knock limit is engine speed. Finally this methodology is applied on an application example. Two spark plug indexing are compared and the combustion settings deterioration due to knock management methodology is estimated.