RF Transistor fmax Calculation Problem

I'm using this formula fmax=rad((fT.n.L)/(8.pi.Rs.W.Cgd)) to calculate the fmax of an rfnmos2v transistor in Cadence, where fT=gm/2.pi.Cgg, n=number of fingers, Rs=sheet resistance, and Cgd=0.4f.W(in mircometers). W and L are clearly the width and length of per finger, and I'm using 180nm technology. Problem is, when I use sp analysis to view the fmax with simulation, the results are not consistent with calculations. For example, with a transistor with w=5.34u and one finger, I expect fmax to equal 40G, while I see 80G in simulation. I didn't predict very exact results (because the equations had many estimations and simplifications) but the variation seems too much. Is the formula wrong or something?
Also, one very big issue is that this fmax formula and the other two that I have similar to this one predict fmax to decrease by increasing W, and increase by increasing N. Sweeping N in sp analysis, the latter doesn't seem to happen at all, and my transistor fails at lower frequencies with a higher number of fingers. Why does this happen?