With its superior power to weight ratio, the hydrostatic dynamometer is an ideal candidate for transient engine or powertrain testing. Given its high bandwidth, the hydrostatic dynamometer can be further used as a virtual power source to emulate the dynamics of the automotive hybrid power sources. This will greatly expedite the investigation of various hybrid powertrain architectures and control methodologies without building the complete hybrid system. This paper presents the design, modeling, nonlinear tracking control and experimental investigation of a transient hydrostatic dynamometer. An electronically controlled load sensing mechanism is employed to facilitate the supply pressure control, and a two-stage high bandwidth valve is used as the primary actuator for the loading pressure control. To enable the model-based control, a 9^th order physics-based model is formulated and then, identified and validated with experimental data. On this basis, model-based nonlinear tracking controls are designed for this multivariable nonlinear system to realize the precise engine speed tracking. A nonlinear model-based inversion plus PID control is first implemented and then, a state feedback control via feedback linearization is designed for reference tracking. Experimental results demonstrate precise tracking performance with less than 5% tracking error for both transient and steady state operations.