here's my take on this.
unless you KNOW otherwise, assume your sound system is NOT bit-perfect. meaning, if you play a cd that is at 44.1k and you send the spdif (coax or opto, both are the same) to a DAC that shows the samplerate on its display, the 'bad' sound chips will still show 48k. its a 'windows thing' (for some reason) to usually force 48k, at least for audio. so one way to tell is if your true 44.1k audio comes out 44.1k on a home stereo or DAC. for a long time, I used a m-audio 2496 (envy24 chipset) card. I -knew- for a fact it was bit-perfect and not a 'resampling' type card like almost all (or all) the creative brand ones are.
I also used to like the c-media 8738 (I think that was its number). it was a very standard and cheap part, sometimes found on mobos but also sourced from m-audio (midiman). it was bit-perfect, too, but only did 44.1 and 48k - no 96k support or 24bit audio, either.
then comes the cards that support 'exotic' spdif extensions, like dts and dolby-digital. my home stereo has 'lights' to show when those modes are on the cable, as well. with any of these modes, I think that windows will NOT resample or othewise monkey with your bits. I think (but not 100% sure) that they have to stay intact, or maybe volume controlled, but that's all.
there are so many BAD sound chips out there, that you can't tell how 'honest' they are with your bits. (I used to work on DAT tape machines and so I'm used to the idea of pure bit copying with no samplerate conversions needed).
anyway, spdif != spdif. just cause its digital does not mean that someone didn't fark with the data in some intermediate stage. you have to check the chipset and just ask around.