The reason for gain is to maximize the signal resolution. Say I have signal that is from -10 to 10 mV. Now if I have a DAQ board with a resolution of 12 bits from -10 to 10 V, then that means I have 4096 points over 10 V. That means I have one point for every (10 V/4096 points) 2.44 mV. For my +/- 10mV signal I have a total number of (20 mV / 2.44 mV/point) data points equal to about 8 points. That means I'm only using 8/4096=0.2% of my total resolution. I might as well have a 3-bit accuracy board.
However, by applying a gain I can make it to where it maximizes signal over the range of the A/D converter. For the above signal I would want a gain of 1000 to get my full resolution.
Here's where it gets easy. In DAQmx, if you set your voltage channel to read a certain range for the voltage it finds the most appropriate pre-amplifier setting for you (on our M-Series boards) and automatically adjust your signal so that you get the best possible resolution.
In your case it sounds like you have a signal that goes 5Vp-p then you might want a gain of 2. If the signal was 1V then you might want a gain of 10.
If you make the gain too big it will clip values beyond what the A/D can handle, or if it has a voltage above the overvoltage protection, then it can damage your board. The easiest way is to let the pre-amplifiers handle everything.