Not to get too technical here, but the latency wrt a DAW is not affected by CPU usage at a specific point in time. You have to basically choose a buffer size that is going to be large enough so that in cases where something does take a lot of CPU cycles,
it could still complete the processing before the buffer-size determined timeslice runs out (keeping in mind that smaller buffer sizes results in more overhead).
Basically, once you select a buffer size, your DAW latency is absolutely fixed, and is not affected by CPU usage. Unless you chose a buffer size that was too small, and the overhead of managing the smaller buffers more frequently results in a dropout.
That wasn't too technical at all, this IS Channel 9
You know, it's funny how everyone seems to understand all kinds of 3D graphics terminology because of the importance of games but audio is still considered some kind of "black art".
I've settled on a buffer size of 256 samples and it's worked well for me on this system.
All I know is that Wave-RT performs just as well for me as ASIO and if using it would mean that people with soundcards that don't have ASIO drivers can get better latencies it's about time software makers start using it more. I mean, why are we still relying
on Steinberg's technology (again, not that there's anything wrong with it and it has some advantages over Wave-RT) when we could be using something that's already built into the OS if we choose to? OSX users just use Core Audio (I think).