Today turbochargers are used by car manufacturers on Diesel engines and on an increasing number of gasoline engines, especially in the scope of downsizing. This component has to be well understood and modeled as simulation is widely used at every step of the development. Indeed development cost and time have to be reduced to fulfill both customers’ wishes and more stringent emissions standards. Current turbocharger simulation codes are mostly based on look-up tables (air mass flow and efficiency) given by manufacturers. This raises two points. Firstly, the characteristics are known only in the same conditions as manufacturers ‘tests. Secondly, the current efficiency given by turbochargers manufacturer is the product of the isentropic efficiency by the turbocharger mechanical efficiency. This global efficiency is suitable for the calculation of the power transferred to the compressor. But the isentropic efficiency has to be determined to calculate the turbine outlet temperature, which implies to evaluate the mechanical efficiency. Most of the time, although tests show this is not so, users make the hypothesis of a constant value. This hypothesis has a strong impact on the turbine outlet temperature and, as a consequence, on the modeled after treatment devices’ light off . This article will present a study for characterization and modelling of turbocharger losses. First a specific experimental campaign is conducted on a test bench using a standard automotive turbocharger. To eliminate the influence of thermal transfers, an adiabatic measurement methodology was developed. Then additional turbochargers were tested on the bench to enhance the experimental database. Furthermore, the influence of oil viscosity was tested by using three different oil grades. Finally this study enables to quantify friction losses in a turbocharger and hence to calibrate new physical models. Additionally, it allows to calculate the isentropic efficiency from manufacturer global efficiency.