Vibro-impacts are common in various automotive engine and transmission gear applications. They are known to cause excessive noise levels, often called rattling or hammering. Input and output fluctuations acting on such systems cause tooth separations and sequences of impacts allowed by backlash at the gear mesh interfaces. The fluctuations leading gear rattling have often been studied for specific applications with the excitations produced typically by an internal combustion engine. As such, rattle evaluations have been often empirical and specific to the systems considered. In this study, an experimental test setup of a gear pair is developed to emulate the same torque fluctuations in a laboratory environment. This setup is used to establish an impact velocity-based rattle severity index defined by the measured torsional behavior of the drive train that is shown to correlate well with the measured sound pressure levels. With that, a validated dynamic model of the experimental setup is employed to predict the same index to allow estimation of rattle noise outcome solely from a torsional dynamic model of the drivetrain. Predicted rattle severity indexes are shown to agree well with the measured ones within wide ranges of torque fluctuations and backlash magnitudes, allowing an assessment of rattle performance of a drivetrain solely from a torsional model.