Techniques exist to precisely measure the electromechanical efficiency in an electric drive unit (EDU, defined here as the combined motor and geartrain). While total drive unit efficiency is an important metric, it does not quantify the different loss sources in the EDU. Due to the tight packaging in production EDUs, it’s impractical to mount a torque transducer between the electric machine and gearbox to separate the losses from these components. Additionally, this approach does not separate the winding, core, and mechanical losses present in the electric machine. In this work, we developed a noninvasive approach to distinguish winding, core, and mechanical losses for EDUs containing synchronous electric machines. To separate the effects of these losses, mechanical output power was held constant at several test points while varying the stator’s current amplitude and angle. By maintaining the mechanical power at a constant value and varying the stator current, the core and winding losses were excited independently of the mechanical loss. We combined the resulting data with an equivalent-circuit model (ECM), including separate resistors representing the winding and core losses, as well as a flux-linkage map (providing the voltage for the ECM) to separate and quantify the individual losses. As validation, we implemented this approach on a Ford F-150 Lightning rear EDU. The Lightning EDU showed lower total efficiency during regenerative, rather than tractive, operation, which is unusual, since core losses are typically lower when generating rather than motoring. The loss breakdown revealed that the machine’s core losses exhibited the usual behavior and that its winding losses were symmetric about zero torque, as expected. The EDU’s greater overall loss when generating was solely due to the asymmetrical mechanical losses.