For system design, analysis of global performance and energy management of electric vehicles, it is common to use the efficiency map of electric traction drive. The characterization of the efficiency map with high accuracy is then an important issue. In this paper an onroad method and an off-road method are compared experimentally to determine the efficiency map of electric drive of electric vehicles. The off-road method requires a dedicated experimental test bed, which is expensive and time-consuming. The on-road method is achieved directly in-vehicle. Experimental data, recorded during an on-road driving cycle, are used to determine the efficiency map using non-intrusive measurements from GPS antenna, voltage and current sensors. A versatile experimental setup is used to compare both methods on the same platform. A maximal efficiency difference of 6% is achieved in most of the torque-speed plane. It is shown that, in an energetic point of view, both methods yield similar results.