EXS 24 Logic Sampler Users Group

Well, I don't know if this comment is worth it, but why don't you simply do
a performance test?
I did it once like this:
- Place a single looped mono string sample on C3.
- Set the EXS to 64 note polyphony.
- Start with one sustained note at C3.
- Keep adding notes. Once you reach 64, open the next EXS instance and
proceed on that one.

On my rather mediocre Athlon 1GHz I can get 180 voices out of the EXS
(sample conversion set to best, also I've set the internal sample rate to
32bit float).
Note that in this scenario the EXS has to do quite some pitch shifting
internally, so if I'd use a patch with proper keymaps I would most likely
get around 220 voices or so.
With filter activated (shouldn't be required for most natural sounds) I can
get around 90 voices in the above mentioned test.

A friend of mine did the very same test on an Athlon XP 1800+ and got 350
voices. A Dual 1GHz Mac should eventually even give slightly more voices. On
one of those 2.5GHz+ new machines with extremely fast Ram one should be able
to run 450-500 voices.

Now you should compare that to what you're maximally using in Gigasampler.
Of course each stereo sample takes up as much power as two mono ones (well,
maybe a slight bit less ?!? Should check that out...), each layer eats up
resources again.

IMO for a full symphony with lots of tutti parts we're not having enough
native power yet, especially when you plan on using native FX as well, but I
guess it's allready quite enough for a lot of things. But then, if you had
one voice per each orchestra member - a 200 person orchestra would be quite
a big one... ah well, maybe all the samples are stereo, but 100 persons is
still a lot and I doubt that there's too many occasions where you'd like
them to play all at once.
I would just do such a performance test and then have a careful look at how
much voices you're actually using in your largest Gigasampler project.

Regards,
Sascha