From the original SynthMaker forum (2004–2014), with its members’ names and dates.
stw ·
infuzion wrote:
aliasant wrote: I'm not sure what the problem is but after I replaced the normal Matrix with a selector based ( electrogears) my synth runs much faster. The normal Matrix is simply unusable. At least to me.
"normal Matrix" = The one in the SM toolbox that uses the buss system? I've never been a fan of the buss system, & avoid it at all costs. However, I was considering it, since electrogear's uses alot of CPU in the one we're working on it seems.
Hmm, i guess we're mixing up different behaviours and "bugs". If i understand right the initial trigger for the multiplex exchange was a significant delay or audio stuttering with large routings. I didn't verify that but i also got my woes with the matrix thing.
My synth became a CPU killer step by step and i didn't sort that out until i deleted the bus-matrix which has eaten up 10% on my 4Core 🙁 I started trying different setups including the slect/multiplex routing. At first it seemed to run much faster until i reached the same amount of inputs/outputs like before. It ended up nearly the same cpu-load.
I hooked up a small cpu test setup for the different alternatives. Surprisingly the bus-matrix seems to be the fastest.
The selector/multiplex exchange is the slowest by far as Infuzion already mentioned.
However, in any case i think the cpu-load is way to high. Maybe someone got an idea how to reduce that?
infuzion ·
stw wrote: I hooked up a small cpu test setup for the different alternatives. Surprisingly the bus-matrix seems to be the fastest. The selector/multiplex exchange is the slowest by far as Infuzion already mentioned. However, in any case I think the cpu-load is way to high. Maybe someone got an idea how to reduce that?
Thanks for the handy-work stw!
?? I think I said the Selector/Selector seemed to be very slow, & now with tests on both machines Selector/Selector proves to be very much the slowest.
Cycles on infuzion's Core2 , AMD64:
select/multiplex: ~370 , 696
select/select: ~611 , 1230
Routing Matrix (bus): 422 , 780
Also, it seems that you were missing the mults that were in the Routing Matrix, but not the other 2. If more could test these out please, that would help!
{file deleted, use v04 below}
BTW, it still does not mean there are not audio issues with Selector/Multiplexer, but I still need proof of the audio stuttering please.
stw ·
infuzion wrote: ?? I think I said the Selector/Selector seemed to be very slow, & now with tests on both machines Selector/Selector proves to be very much the slowest.
Yeah, that's what i menat with the multiplex "Exchange". Sorry was a bit mistakeable...
infuzion wrote: Also, it seems that you were missing the mults that were in the Routing Matrix, but not the other 2. If more could test these out please, that would help!
😊 ... oops - wrong schematic. I deleted these in another version which still resides on my HD. Hmmm... it was late
EDIT: Another detail which heavily affects CPU usage with the matrix on my system is the amount of "Matrix source" modules. In my latest setup -which hopefully is attached NOW - i took onkly on of these modules with ten inputs instead of ten modules with only one input. I think we should rather take this schematic for testing.
Sorry for the inconvenience.
infuzion ·
stw wrote: Another detail which heavily affects CPU usage with the matrix on my system is the amount of "Matrix source" modules. In my latest setup -which hopefully is attached NOW - I took only on of these modules with ten inputs instead of ten modules with only one input. I think we should rather take this schematic for testing.
Ah yes, much better. I did add the Stream Mults back in, since that better reflects what I intend to do. I added a test to see just how much CPU the Mults+Adds themselves eat up. Reflecting upon your comments in prayer I found that just using the Buss in a Routing Picker seems to be the best bet for me:
Cycles on infuzion's Core2:
select/multiplex: ~361
select/select: ~600
Routing Matrix (bus, 10x10): 226
Routing Picker (buss 10x1): 54
Mult+Adds: 39
If the Buss system actually works, you'll know which version I'll take!
Could someone test this out in a real VST please? It is 2am, I need sleep, drive in snow, work, work... TIA!
BTW stw: those of us whose only connection to the internet is a weak WiFi source from across the street would prefer you save OSMs with the file compresson on.
😉
Wrote something here? To have a post removed or your name hidden, write to [email protected].
SynthMaker