I believe it should ignore channels it doesn't need. Is the plan to render out all these channels into one or more renders or just an rgb/rgba render?
Also just to be sure, as you hadn't mentioned it in your first message, but are you watching your bounding box and not letting it get out of control? Sent from my iPhone On Jun 19, 2013, at 2:40 PM, "Ryan O'Phelan" <[email protected]> wrote: > Yes, I'm merging lots of reads into one stream. > I assumed that Nuke only grabs the channels it needs at the time, and that > any channels merged in that aren't used, don't load into RAM. > Is this wrong? > > R > > > On Wed, Jun 19, 2013 at 5:14 PM, Elias Ericsson Rydberg > <[email protected]> wrote: >> I assume you're merging all the channels into one big stream? Does memory >> use go down if you don't? >> >> >> 2013/6/19 Ryan O'Phelan <[email protected]> >>> I'm running out of memory on the farm on one of my comps. >>> I don't remember running into this kind of thing with a comp of this size >>> on nuke 6.3x and I'm wondering if I have missed something. >>> >>> I've precomped all the branches that I can. >>> >>> I've also turned off all postage stamps. >>> >>> I've also tried turning down the cache memory usage to 10%, along with >>> turning all memory settings to very low. >>> >>> Renders consistently run out of ram on the farm. The farm nodes have 16GB >>> RAM. >>> On my comp station, this comp eats 22GB of RAM, until I clear the buffers. >>> >>> >>> It's a somewhat large comp. 781 nodes. 762 channels. About 75 read nodes >>> (reads are CG passes) >>> HD inputs are between 2MB/frame and 8MB/frame >>> Rendering zip1 16 half EXR >>> >>> Environment: >>> Windows 7 >>> Nuke 7.0v6 >>> same config on farm, except 16GB RAM >>> >>> >>> Looking at my buffer report, I'm noticing a large amount of ChannleMerge >>> nodes reporting memory usage of 300-400MB each. For example: >>> >>> ChannelMerge >>> total usage: 339MB >>> user data: >>> cacheRegion 1920x1080 >>> channels >>> [... lots of channels listed here ...] >>> usage 100% >>> lossage 19 >>> weight 0x0 >>> >>> >>> another example: >>> >>> ShuffleCopy >>> total usage: 247MB >>> user data: >>> cacheRegion 1928x1088 >>> channels >>> extraTex_vrayUserColor_restaurant01.g,nigel.body,props.dump,lloyd.body,linus,extraTex_vrayUserColor_set03.r,reflect.red,reflect.green,reflect.blue,refract.red,refract.green,refract.blue,specular.red,specular.green,specular.blue,GI.red,GI.green,GI.blue,SSS.r,SSS.g,SSS.b,diffuse.r,diffuse.g,diffuse.b,lighting.red,lighting.green,lighting.blue,extraTex_vrayUserColor_interiors.g,extraTex_vrayUserColor_restaurant02.r,extraTex_vrayUserColor_set01.g,extraTex_vrayUserColor_set01.b,extraTex_vrayUserColor_set02.r,extraTex_vrayUserColor_set02.g,extraTex_vrayUserColor_set03.g,gush.a,normWorld.r,normWorld.g,normWorld.b,gg.body,gg.hair,bldg.green,bldg.blue,bldg.alpha,lloyd.stripe >>> usage 100% >>> weight 0x0 >>> >>> Is this normal? >>> Why are there lots of channels listed in the ChannelMerge node, when it >>> should only be evaluating two input channels, and outputing one channel? >>> >>> Why is this comp so heavy? >>> What am I missing here? >>> >>> Thanks for your help, >>> Ryan >>> >>> >>> >>> >>> >>> _______________________________________________ >>> Nuke-users mailing list >>> [email protected], http://forums.thefoundry.co.uk/ >>> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users >> >> >> _______________________________________________ >> Nuke-users mailing list >> [email protected], http://forums.thefoundry.co.uk/ >> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users > > _______________________________________________ > Nuke-users mailing list > [email protected], http://forums.thefoundry.co.uk/ > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
_______________________________________________ Nuke-users mailing list [email protected], http://forums.thefoundry.co.uk/ http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
