yeah. i think i got it.

the sound architecture in xp is much maligned. it's one of those examples of microsoft taking too much control over something that ought to have minimal operating system interference. they never *really* fixed this, but it changed a lot in vista.

i've known this the whole time but it was never really a problem before, so why would it be a problem now? well, i may have changed a few settings at some time in the past and forgotten...

the kmixer seeks to resample everything, in ways that often make little sense. it seems like the people that designed it didn't really know a lot about audio, they just figured that upsampling everything to the highest formats possible would be a best practice strategy across the board. that's not true, for reasons i won't get into.

both the m-audio and the alesis come with lock settings that prevent any modification of the sample rate from outside sources. locking the devices to 44.1 had a big effect on the wdm driver outs for both devices, presumably because it's stopping any sample rate conversion. an a/b with the asio outs through cubase actually suggest that this minor fix creates an inaudible difference through wdm and a larger difference through directsound. before i locked them, it was directsound that clearly sounded better. so, this is a big difference.

i want to draw attention to this, though. because i had both outs to 44.1. why would windows upsample a stream when the source and hardware are set to the same value? that is entirely absurd. and, who would think you need to lock the rate in place to prevent the operating system from interfering? it's utter madness, truly.

but, i think this is at least a partial fix. i'll need to do more testing over night to determine if it's fully fixed or only half fixed. but, the difference is quite positive and quite obvious, so i think this is the solution to the bulk of the problem.

the problem was less fixing the problem once i figured out what it was and more trying to figure out what it was.

and, the reason the soundblaster stayed stable was because 44.1 was the max - which was also the reason that switching from the soundblaster to the m-audio fixed the m-audio [it went back to 44.1, until it was re-opened].