The more cycles you sample the better. The peaks of your output will be more resolved (smaller full-width-half-max). As for which peak is fundamental, it should be the largest amplitude and the lowest frequency.

Okay, that's a gross over-simplification. I'm assuming that your signal is essentially a sine wave, and that your algorithm limits the output to positive frequencies. But if your signal is a complex wave, like a note from a musical instrument, or worse yet a whole orchestra, there's no guarantee that any single peak will present as fundamental. What's the fundamental frequency of a chord?

The truth is that there are a number of variant FFT algorithms. Rather than risk giving you bad information I really recommend that you look up the details of how the algorithms work yourself. Try several until you find one you like, or one that gives you output that makes sense for your application.

Joe