If you don't know what resampling means, I suggest you take a look at the Resampling entry in the Hydrogenaudio Wiki (I wrote it!).
Resampling can happen by design or by accident - the latter being often the case, with serious risks of audio quality degradation. Even when applied voluntarily to an audio sample, precautions are required for best results.
My first experience with the disastrous consequences of resampling occured when I tried to listen to some Vorbis ogg files stored on my Linux workstation, transmitted to my Pioneer 916 receiver using an S/PDIF connection.
This worked fine - I mean I could hear the music playing all right, but it sounded awful! Diagnostic: the ogg file had been ripped from a CD - sample rate of 44.1kHz - but was being played back using the S/PDIF interface at 48kHz. ALSA (the Linux sound card driver) was doing its best to resample on the fly, but clearly it was not doing a good job at all. I later learned that ALSA uses by default a fast but rather crude linear interpolation resampling algorithm.
But why did ALSA have to resample in the first place? Couldn't it just play back the music at 44.1kHz? No, unfortunately, not on the ALC850 codec. This codec has a fixed S/PDIF rate of 48kHz.
My solution was to buy an extra PCI sound card with an S/PDIF interface that could use both 44.1 and 48kHz sample rates. I ended up buying a Trust 5250 PCI sound card on eBay for 13 euros with optical output.
Now I get "bit-perfect" output from my PC to my Pioneer receiver. The Pioneer has 24-bit 96kHz DACs that provide their own high-quality conversion.