If your fet is showing 1.7 ohms @ your available Vgs, you probably want a different fet if it is to handle 2.5A, 6W or so is a lot of waste heat.
The threshold voltage is where things start to happen, but really you usually want at least twice the threshold voltage to get the thing to switch properly into saturation.
Note that for reasonably fast switching you actually need to supply quite some current to the gate, it does not last long, but there is a reason mosfet drivers can often manage to supply several amps for a few hundred ns at each switching edge.
Keyword is 'logic level mosfet', you can easily get parts that will fully saturate on a 3.3V output.
If you look at the IRF640N datasheet you will see (in the electrical characteristics table) that the threshold will occur somewhere between 2 and 4V, but if you look at at the conditions, you will see that threshold is considered to correspond to 250uA of current....
Further, if you look at the Rds(on) resistance you will see that it is specified at Vds = 10V, so this is not a logic level part, it is probably intended to be driven from a gate driver with a 12V rail.
Compare with something like an IRF7331, much lower working voltage, 20V, but Rds(on) is now specified to 2.5V (0.045 ohms max), so that thing can be driven (slowly) directly from 3.3V logic.
A general observation about mosfets is that as you increase rated voltage or rated current, everything not directly related to that parameter gets worse, you seldom want to use a 200V mosfet in a 12V circuit for example because the 200V rating will hurt you for gate charge, RDS(on) and just about everything else, same way, massively over sizing the thing current wise will hurt you for the same reasons (Not note that maximum steady state Id values become increasingly fictional above 10A or so).
Regards, Dan.