by davidv » Fri Apr 23, 2010 8:54 am
Heres my test, and what I assume is right according to the precise audio calculations im doing in DSP for chipsounds:
Mac Book Pro (first generation) Core Duo 1.87Ghz
SampleRate:44100
Buffersize:256
platform OSX 10.4 and WIndows XP (relatively same results +- 1% cpu)
Bidule: Running 6 stereo instances of chipsounds each with one monophonic instrument loaded:
CPU usage (one core): 28%
Bidule: Running 1 stereo instance of chipsounds each with 6 monophonic instrument loaded:
CPU usage (one core): 21% (you see you do gain by not wasting cycles in sending different mixes to the host)
Now of course if I raise the polyphony usage to 16 in each instrumentand play lots of voices, then yes
I reach the 50%+ CPU usage (one core) in both cases, but that is to be expected.
Please dont compare chipsounds with a "simple synth". it might be "square wavs" but they are
non trivial to get right without aliasing, it requires juice:
Each voice is *oversampled*, and all their envelopes and LFOs are as well.
Now i dont know if you have a Live session that you can PM me for more analysis, but these are the figures
which i think are normal.