Max voices count VS performance

Is there any limit of voices that is good to follow, or actually forced by FMOD, for specific platforms?
I only saw some numbers for different platforms in one guide for Wwise, and its also 5 years old now.

It is written in the documentation that FMOD automatically virtualizes channels (are they what’s named Voices in FMOD Profiler?) when they exceed the limit.
I guess the limit is what you set in project (in my case Unity) settings - for Real and Virtual Channel Count.
Is there any way to track this in profiler, on what stuff is currently getting virtualized?
Is the Voices number the only thing in such case to look at, and if it exceeds the project settings, it’s just getting virtualized, but FMOD doesnt show which stuff is virtualized?

If with my gameplay generating many voices in cerain moment I am exceeding the Real Channel Count set in the project (but it sounds as intended), but profiler shows no huge impact for CPU: Mixer and Update, is it still for some reasons worth keeping the number down?
I know it’s generally a good and reasonable practice to limit the voices, both performance-wise, and to not end up with too much mess audible. But with such case if the experience is ok and CPU meters are quite low, does that still apply?

We do have documentation on per-platform performance in the platform details section of the FMOD API documentation, though these are less a recommendation and more a reference on what to expect on each platform. I’d encourage you to take a look at the white paper on CPU performance to work out what would fit your project and target platforms.

“Voices” in studio refer to channels, yes. You can monitor the current amount of real (active) and virtual (non-active) voices in the profiler, both globally and for all instances of a given event. You cannot specifically track which voices are being virtualized, but you can track when event instances are being virtualized. Note, however, that these are separate systems - an event instance can be “real” while some of its voices are virtualized. I’d recommend reading over the FMOD Studio API Voice Control section of the Virtual Voices white paper for a better understanding of how these two systems interact.

Besides CPU impact and audio mess, there’s no specific reason why you can’t increase the voice limit - it just depends on your game’s requirements and audio resource budget, and ultimately you’re the best person to make an accurate judgement on what an appropriate limit for your game is.