While I can not guarantee bit-exact in all scenarios, by setting the volume control to 100%, and not playing other sound, you essentially have a direct path out.
... with which you say that it would come to this in the XP environment, or the other way around : in Vista this would work out the same as in XP, in non-exclusive mode.
I can't help myself ... FYI :
I have a soundcard with non-certified drivers. Lucky me, because they bypass kmixer.
I've written a dedicated DirectSound player for that (so this is no KS and no ASIO), with a DMA buffer in the soundcard of 112 samples. The souncard won't allow me to go lower ...
The audible result is amazing, mind you, in my system (which is not an el cheapo system).
When I apply all to have -as you say- "an essentially direct path" via the certified drivers of the same soundcard (which do not bypass kmixer), the sound is still amazingly good because of the way I process the data (with the same very low latency), but it is different. It is -if you like- more crispy. Is crispy better ?
No, it psychologically can't be better, because of the stupid reason I can measure it's not fed with a 1:1 bit perfect stream. The differences are marginal, but, all over the place. Looking at the waves (not the data therein) it looks like the least significant bit wobbling to the wrong direction.
Since the theory is -indeed- that when all is applied to let kmixer not interfere (no resampling "needed" etc.), it still touches the sound somehow. This is stupid and unnecessary, once the "rules" have been applied.
What I understand from you, Amir, is that around this phenomenon nothing will change in Vista, as long as non-exclusive mode is applied. Okay, depending on what causes the stream not being bit perfect, it can be that my "all over the place" turns into "here and there" because of the better processing in Vista. However, there's still no reason to touch the sound, again, once the rules are applied.
Please note what this is about : when all is done like it *can* be done, no player needs to be re-written (towards exclusive mode etc.) to achieve bit perfect playback. Also, I don't think it is that easy to have a player working in exclusive mode, because of all the environmental constraints that would apply, thinking of it causing deadlocks within itself, at e.g. capturing sound at the same time.
Just a thought.
Noise level during system activity <= -90 db FS A-weighting
The last one is key. If you hear your hard disk accessing media as you listen to your music, you want to ask for WLP 3.0 compliance .
I think this is a good example of how far people can be off track from eachother;
No matter how well you mean all, the time that I "allow" my harddisk to be accessed during playing has long gone. Don't think I'm alone ! Just think of by 112 sample buffer ... it couldn't be done ...
Oh well ...
Dynamic Range >= 90db FS A-weighting
I know, this is in "your" context (of the general audience ?). But with 192KHz sampling it would be quite nothing. In fact, it should be infinite, because of being in the digital domain without D/A conversion, yet. Again, it will be in your context, and I just place it in "our" context ...
Lastly, please note that I am (and not alone again) in the leage of eliminating the $$.$$$ boxed CD player, which should be an easy task because of, say, the PC principles. And no, it's not about saving the $$.$$$, but about achieving much better playback. That too was achieved long gone by many, but it's a struggle. Today it needs the (by MS) not advised DirectKS hence instable driver-systems, or the workaround with ASIO which someone like me wouldn't even begin thinking of writing my own. Add to that the way better results with DirectSound, once you own a driver that is not certified ...
And please don't say that there can't be a difference once there's bit perfect playback, because there's always the (very audible) jitter phenomenon, which in the end comes to latency ...