The only value in EXS in regards to performance is the fact that when the VI aka the VSL server is full on memory, you can start loading up EXS, which uses Logic's memory address, so in effect you can load more samples into EXS where your VI is full (or visa versa for that matter). However, I've never had to do this, and at the end of the day yes, it's a pretty good workaround in terms of really getting the most out of your machine - but my machine's CPU isn't fast enough to run all the optimised and freezed tracks that VI can deliver ALONE let alone thinking about starting up EXS. I always max out the cpu before the RAM this is a dual 2.0ghz G5, and I'm talking about 40 stereo 32bit freeze tracks of VI and a couple of altiverbs at low buffer settings. Leopard is coming in the next few months, so we are told, and that will be fully 64 bit. Give it 6 months to transition, by Christmas time, I'd say this memory ceiling thing will be history and we will be able to load 16Gigs of optimised (and that means 200+ tracks) of VI into a Mac Pro 8 core machine. It's not as far away as it sounds. In the mean time, only a Quad G5 or Mac Pro will max out the RAM using just VI, as it is just so efficient (if you take the time to optimise).
A quick tip by the way maybe everybody already knows this, but in logic if you want to optimise in VI, just click "Learn", then freeze the track, - the learn mode still works! - unfreeze the track, click "optimise" and you're done. Much quicker than waiting for the VI to "learn" in realtime. Never failed for me. An example - a 80 Mb string legato patch will be reduced to 2mb of RAM usage! However it is worth noting that each instance does use some RAM just for being there so you still should work to keep your instances down by using keyswitching to change patches of the same instrument rather than opening new instances for new patches of the same instrument (where possible). But all of this combined means you can load a vastly larger orchestration than you EVER could with EXS (unless you printed to audio, and unloaded the instrument - which you don't need to do with VI).
Miklos.
A quick tip by the way maybe everybody already knows this, but in logic if you want to optimise in VI, just click "Learn", then freeze the track, - the learn mode still works! - unfreeze the track, click "optimise" and you're done. Much quicker than waiting for the VI to "learn" in realtime. Never failed for me. An example - a 80 Mb string legato patch will be reduced to 2mb of RAM usage! However it is worth noting that each instance does use some RAM just for being there so you still should work to keep your instances down by using keyswitching to change patches of the same instrument rather than opening new instances for new patches of the same instrument (where possible). But all of this combined means you can load a vastly larger orchestration than you EVER could with EXS (unless you printed to audio, and unloaded the instrument - which you don't need to do with VI).
Miklos.