From another thread:

>>> server minpoll 4 prefer

>> The statement I used looked something like this:
>> server prefer mode 1

> If you set 'minpoll 4' it will follow the GPS more closely.

Could someone elaborate on this?

If you are using a Garmin GPS 18 LVC as your time source and a Type 20
driver, my understanding is:

a) The PPS signal rate is hardware defined within the GPS. You are
going to get one pulse per second, on the second, from the GPS
hardware, no matter how much you twiddle with the minpoll parameter.

b) The Type 20 driver disciplines the system clock upon recieving the
PPS signal. How often this happens is determined by how often the GPS
hardware sends the PPS signal, not on what value you use for minpoll.

c) With the Type 20 driver, the $GPRMC sentence needs to decoded only
once at system startup to get the time. Thereafter, the time is known
by counting the PPS pulses from the GPS. Twiddling with the minpoll
parameter might cause the $GPRMC sentence to be decoded more often
but, considering that the $GPRMC sentence is always in sync with the
PPS signal, how is setting "minpoll 4" going to make nptd "follow the
GPS more closely" if, by selecting the Type 20 driver, you are already
following the GPS via the PPS signal? What am I missing?