As I know you can obtain this frequency by simulating a simple amplifier (a folded cascode amplifier can achieve the maximum frequency or bandwidth of process)
I was asked this question by someone. So I had the same doubts as you too, while I found a paper which defines the cut-off freq. as 2*pi*gm/C. C includes intrinsic input capacitance, parasitic gate-bulk capacitance, and the gate-source and gate-drain overlap capacitances. I want to know the values.
the gm/C is adevice figure of merit from the days of valves/tubes. It is also called the 'gain bandwidth product'. It is only useful on devices with very little resistance from the controlling node (gate, base, grid) to the active part of the device. This is why it is not used for bipolar junction transistors.
If nothing specific is given to you then you can assume that 2*pi*gm/C = ft where ft is the transition frequency of your device (where the gain of the device in question is unity). The gross appx of C ~ Cgs (back of the envelope sort of thing). I hope it helps.
I think the ft is just an approximate value. It will change with the process variation or bias conditions. So 2*pi*gm/Cgs is a good approximation. Gray & Meyer's book talks about this.