The estimation accuracy of the stator flux and speed using open-loop integrators drops as the mechanical speed reduces in direct-torque-controlled (DTC) interior-permanent-magnet (IPM) synchronous motors. At lower speed, inaccurately estimated stator flux and stator resistance mismatch have significant influence on the steady-state and transient performance of the drive system. The robustness against parameter detuning and signal noises can be improved by deploying closed-loop observers with estimated stator flux and rotor speed. This paper presents a comparison of two stator-flux and speed observers for variable-structure DTC IPM motor drives: an adaptive sliding observer and an extended Kalman filter (EKF). The experimental comparison of the adaptive sliding observer and the EKF regards steady-state and dynamic performance, robustness against parameter uncertainties, stability, and computational complexity. Recommendations for observer selection and the effectiveness of the various schemes are concluded.