In "The g_m/I_D Methodology, a Sizing Tool for Low-Voltage Analog CMOS Circuits", we compare the semi-empirical to the compact model approach. Small numbers of parameters make the compact model attractive for the model paves the way towards analytic expressions unaffordable otherwise. The E.K.V model is a good candidate, but when it comes to short channel devices, compact models are either inaccurate or loose straightforwardness. Because sizing requires basically a reliable large signal representation of MOS transistors, we investigate the potential of the E.K.V model when its parameters are supposed to be bias dependent. The model-driven and semi-empirical methods are compared considering the Intrinsic Gain Stage and a few more complex circuits. A series of MATLAB files found on extras-springer.com  allow redoing the tests.