Forum

Quad core, cpu usage and vvvv

vvvv beta question
#1

hi there

i am wondering if vvvv is completely using my processor ressources…
i explain: perfmeter says “CPU” when i am in my patch, which should means that the processor is kind of the bottleneck here right?
but when i open windows’ (7 64bits) ressources manager, i see the processor usage floating around 30%…
i have a quad core (intel q6600 oc to 3.2ghz) so could it be a core/thread problem?
i’m not really informed about how vvvv is using core/threads but if there is a way to optimize CPU usage here i would be glad to learn it… cause using less than half of the CPU power… :/

#2

if you have one core of a quadcore maxed out that amounts to 25% cpu usage in windows task manager. so your readings are correct as vvvv only utililizes 1 core per instance with the exception of some multithreaded plugins. implementing multithreading for more of vvvvs core functionality is a topic also touched https://discourse.vvvv.org/t/6849.

cheers

#3

thanks for the quick answer. boythreading, yumyum… :)

#4

hum see documentation of your proc or very specifics forums but quad core is usually 2 real core. if you take for example an i5 your are facing 3 core. i 7 is more a real quadcore than 7.

you can use Process.exe (attached) to attribute an instance of vvvv to a thread

actually vvvv is not multithreaded ( if i well understood)

Process.zip (25.4 kB)

#5

merci karistouf ;)

it’s for using tonfilm svvvvitcher… maybe it would be possible to assign each channel to a dedicated core but don’t know how to achieve this if it’s possible at all…

new intel quad core 2600k seems to handle better single core power http://guru3d.com/article/core-i5-2500k-and-core-i7-2600k-review/3
but nothing compared to real multicore-threading…

#6

hi aze, i would say say that with the frame system of vvvv there would be none trouble of synchro. perhaps delesting Fx dll or subpatches on another core ould be great, and using R and S ( Node) to send them to svvvvitcher ?

#7

aze, if I were you I would try to offload a piece of your patch to another core. Not the entire patch, just the part that’s using most of the CPU. I think this would be the simplest route…

#8

was lost in overclocking settings…
so, thing is i don’t know how to do guys… if i get it: maybe launching several instances of vvvv, let’s say 1 for each svvvitcher channel and 1 for svvvitcher’s engine + postFX and exchanging variables with S and R nodes between them (sending rendering textures, midi assignations etc…)? makes a lot of variables… and assuming what i’m saying is not absurd i don’t know how the GPU would deal with that…
how do i assign an instance of vvvv to a specific core? using your exe karistouf? when i launch it a command window is launching and nothing happens… may i drag it somewhere or something?

#9

that wouldn’t work for a number of reasons, only one instance of vvvv can have a renderer and as far as I know S and R nodes talk over instances. You need to use OSC or any network communication.

#10

ok i see, thanks. so it could be usefull in a single patch scenario for big mesh calculation or what… anyway here with svvvvitcher: too much complicated for me!

#11

Multiple instances can indeed have multiple renderers, maybe it’s just that only one can go full screen.

Aze, find out what part of your patch uses all the CPU and put it in a separate instance on a separate core. If you don’t know how to do that look at the help file for SetProcessAffinityMask. Use SharedMemory to communicate between instances.

#12

not so much time to do this kind of stuff right now but i will try later thanks.

© vvvv group