Multiple bit depths in sonar 5

[jamie_drum]

Chapter 20 in the manual does a pretty good job of explaining this issue, but I still have some questions.

It seems that Sonar can now support multiple bit depths in a project. Before, it was simple--you chose the bit depth and sample rate for your audio files before you started recording, and you got what you got.

Now, however, there is one bit depth for playback, one bit depth for recording, one bit depth for audio import, one for export, one for rendering, and (I believe) one for mixing. You control the playback bit depth from the options--audio dialog box, the record bit depth from the options--global--audiodata dialog, the import depth from options--global, the export depth at the time of export, and the render depth from options--global. The mixing it always done in 32 bit (I think) unless I click the "enable 64 bit double precision" button.

I see the virtue of this added flexiblity, even if it is confusing. My question is this, however: how does this all affect CPU usage? For example, if my audio files are all at 24 bit, but I am setting the playback bit depth (aka the audio drive bit depth) at 16 bits, surely the conversion uses up some CPU? Similarly, does the render bit depth affect how much CPU a real time effect uses, or only when I apply the affect offline or "freeze" it? On the other hand, what happens if I try to play 16 bit audio files when I have the playback bit depth set at 24 bits?

Maybe I am not understanding this issue properly. I should probably just be doing everything in 16/44.1 anyway, but I have some older projects, with just a few tracks, that are 24/96, and I am wondering how it all works.