I did some CPU usage testing for the CS80v standalone demo for Win2k, on a 2GHz P4. With all 8 voices sustained I found the following:
I start with "Basic" config:
HPF, LPF, SINE, PWM speed SLOW for both layers
Ring Modulator: all 0
Sub Oscillator: all 0
(features later turned on one at a time)
1.
"Basic"
CPU: 8%
2.
Enabled HPF on both layers:
CPU: 10%
3.
Enabled LPF on both layers:
CPU: 12%
4.
Enabled SINE on both layers:
CPU: 18%
5. Changed PWM speed to FAST:
one layer:
CPU: 28%
both layers:
CPU: 38 %
(Note: This is for sine and saw PWM waveforms. Square and Rand had no impact whatsoever. For Noise, CPU was always at 27%/38% regardless of PWM speed.)
6. Set Ring Modulator MOD > 0
CPU: 40%
(Note: Changes to Speed & Depth had no impact)
7. Set Sub Oscillator VCO > 0:
CPU: 58%
(Note: Changes to Speed, VCF, VCA had no impact)
From what I can see, these are the most CPU intensive features, in descending order:
1. Sub Oscillator VCO>0 (added 18%)
2. PWM Speed (added ~10% per layer)
3. Post-filter SINE wave (added ~3% per layer)
4. Ring Modulator MOD>0 (added 2%)
5. HPF/LPF (added 1% each, per layer)
I saw the same pattern of behavior on my PowerBook G4 867 running as AU plug-in in Digital Performer 4.11, except for much higher % due to the slower CPU. In fact, I can't get more than 7 voices out of it at any time...
So, Arturia, when you list "required configuration" as P2 500MHz or G3 500MHz, I don't think that is realistic. Not unless you include a max polyphony setting that would prevent the host application from running out of CPU.
Finally, I'm curious as to why the PWM Speed and Sub Oscillator VCO mod (especially) are so CPU intensive compared to everything else? Can this be remedied?
--Zotor