My (perhaps misplaced) assumption has always been that the computer is 'doing less' to the audio data when it sends it out over USB as opposed to optical or coax. The two basic aspects I'd guess are: 1. changes of state, eg. to convert from electrical to optical 2. changes of format, eg. WAV to S/PDIF, S/PDIF to i2s etc. I'd also guess that both of the above can be done entirely transparently. Coming before both of the above is what the OS does before it gets sent out which perhaps would be identical irrespective of the output method, or not?