decided to try a few things to better understand how audio depth works, for this test i used my Audigy 2 and the Realtek HD. ffdshow includes an option to enable 16/24/32 depths so i ticked 16bit and played an AC3 movie with 6 channels and check the directsound nBlockAlign the results were quite cool, according to this choosing 32bit would give the best results. choosing 16bit means you get only 12 real world bits the remaining bits are lost somewhere. choosing 24bit means you get 18 real world bits while choosing 32bit means you get 24 real world bits i can only conclude that once you enable 32bit in your audio player or ffdshow would let your card fall back to the maximum depth possible, when the actors where talking to each other and used the letter S a lot there was a clear difference between 16bit and 32bit but not enough between 24 and 32bit. in Foobar choosing 32bit gave me the best quality possible especially when the song featured a lot of cymbals while 24bit was good but not great, 16bit with Dither was similar to 24bit, without Dither the quality wasn't that good tbh, finally in 32bit i have noticed a couple of guitar leaks and background vocals i have never heard before in a few songs, the Realtek was better than the Audigy 2 especially when playing mp3s. WAV files got fixed depths and there won't be any difference, mp3s however got floating depth and since it is a compressed format a higher audio depth could improve quality by recovering some of the lost bits, without good speakers you're unlikely to notice any difference in both cases and finally the Creative 24bit Crystalizer thingy takes this a step further but while the Crystalizer is a hit n miss affair choosing a higher depth in your audio player is always better.