So 6 sine wave oscillators per voice.
I’ll try to load up something like that on my SSP and get back with it. It’s late here, so probably tomorrow.
Edit: was too curious so I set up a single voice like so:
As you can see from the CPU meter next to sample rate indicator, the CPU is barely pegged at ~3%.
Currently the LFO module can be used as a regular oscillator (tri, sine, saw, square) or an LFO (same forms) outputing a saw to the output module. The 6 additional LFO modules are outputing audio rate sine waves to the saw oscillators freq input. (Could have used them at LFO rates, but figured this would be a better stress test.) I have one midi module outputing pitch and gate to the saw oscillator, and the other MIDI module sending velocity on information to the freq inputs of the 6 sine modulators.
You could get way more complex than that, but for testing purposes, I thought it should be good. Someone else could chime in if I’m off base here as I’m not familiar with the DX7.
So based on what I’m seeing here, it looks like 16 voices is possible and would use 45-80% CPU. (also note that this is at a fidelity far beyond what the dx7 was able to produce back in 87.) As the voice count increases the overhead will increase, so my cpu estimations could be off. However, currently the MIDI module only supports 8 voices. So you could have 16 voices but would be controlling them through 8.