try it and see ?

I've tried as I stated in the previous post, and it doesn't seem to cause any major increase/decrease... that's why I am asking this 😁

Don't expect huge CPU loads from a little mixing and forwarding data to the sound card...

On older computers the Soundblaster FM synth might hit your CPU a bit.

Yes, I supose FM synthesis emulation is more CPU-demanding as it requires realtime waveform calculation as opossed to pre-recorded PCM samples played by a GUS. That's for MIDI, but correct me if I'm wrong.

But I was talking about PCM channels only: let's say a game uses PCM for music & effects (music modules for example, souch as Jazz JackRabbit or SuperFrog): shouldn't it be A LOT more cpu-demanding with a GUS emulation instead of Sound Blaster emulation?
Or maybe all sample data is transferred to emulated GUS memory so it does, indeed, require less CPU for emulation...

Only a developer/advanced user can really tell how those things work!

The same mixing will have to be done either way. In one case (GUS) this was done in hardware on the original card, so it is now carried out in software on the host CPU. In the other case (SB) it always had to be done in software, which is now carried out on the emulated CPU. Simply put: anything that is carried out on the host CPU versus the emulated CPU is bound to be much more efficient. Even orders of magnitude, depending on the selected core.

That said, sound mixing doesn't take an awful lot of CPU power, so it's not really all that noticeable much of the time.

In the case of multiple sounds being mixed (like MOD music plus PCM sound effects), I can see how GUS emulation might be less CPU-intensive than SB emulation because the SB in DOS has only one hardware PCM stream (well, two if you are doing stereo I guess but that doesn't count) and all mixing is thus done by the game software via the emulated CPU.

The GUS, on the other hand, is built around hardware mixing of multiple sound samples. In the context of emulation, this means that sounds samples will be passed from the game through the emulation layer and be mixed via the real host OS. This is potentially more efficient, but not likely these days since the realm of PC audio is shifting towards doing everything in software due to the growing adoption of multiprocessor architectures (audio mixing and effects are not CPU intensive to begin with, so it only made sense to offload them to a sound card when CPUs were still slow and had only one core).