Sound cards may specify a rang of sample rates, usually a multiple of 48 kHz. Why does this matter to you, if all of your audio is at the CD rate of 44.1 kHz, and FM stereo samples at 38 kHz?
Suppose the card really ran at 44.1 kHz. What does this mean?
Remember Nyquist's rule .. The sample rate must be at least twice the highest frequency to be reproduced. Any frequency higher than this will be aliased down to another frequency within the range. To prevent this, we need a low pass filter on the input, and on the output.
If the card's rate really is 44.1 kHz, it means it needs an analog filter. This is a very sharp cut off filter. On some of the cheap sound cards, the filter is not sharp enough so it causes a loss of high frequencies and allows frequencies that would cause aliasing to pass.
Now suppose we use a higher internal data rate, “oversampling”. If the card's converter frequency is 192 kHz, half of that is 96 kHz. We need an analog filter to eliminate anything above 96 kHz. For audio, if we care about only up to 20 kHz, this is a very simple filter. If the rate we really want is lower, we need to filter it again, but this is a digital filter. Usually, this would be a “finite impulse response” filter with many sections, so it is very close to ideal.
These filters are used both going in and out. There is also a digital filter as part of any sample rate conversion. In general, the bigger the difference in sample rate, the better the conversion filter works.
The conclusion here is that it really is worth the extra money for a higher sample rate, even if all of your audio is recorded at the lowest rate. The price difference isn't much.