Thanks Darius for your help.
Well, in fact I was able to get some sound out of my digital speakers (pluged into optical output of the Mac Pro using TOSLINK), so it seems that at least a part of the passthrough process works. The sound is awful and is more like garbage than real playback, and after few seconds, it dies.
I have not undertood yet why the video playback, and the VM reactivity in general is highly crippled when the VM uses the sound controler in passthrough.