RF_2015 longer sim times
-
scooterabc@cox.net
- Posts: 1
- Joined: Thu Apr 07, 2016 11:42 am
RF_2015 longer sim times
What is the diff or reason that a .flw scene in RF2014.147 takes 33.15 mins to simulate (75 frames)..and the exact same .flw scene in RF2015.186 takes 47.25 mins..???? ..and there seems to be more "errant" particles in the 2015 sim. The scene is a simple emitter (water) w/normal gravity pouring into a container,w/-k daemon surrounding the container. I thought RF2015 was supposed to have "improved" render times..a nearly 50% increase in simulation is not what I would call an "improvement"? I have tested 4 other scenes with nearly the same results. Is this normal or am I missing something? CPU=i7 975x oc'd to 4.0MHz/water cooled...GTX Nvidia680 w/4G ram...24G 1600Mhz ram...Win7 pro 64bit.
Re: RF_2015 longer sim times
It's really hard to say from a distance what's happening there on your computer. I know it sounds like a standard answer, but here on my machines i see a speed boost around 40% compared to RF2014. Anyway, I've read few reports where users observed a similar increase in simulation times as you do.
Did you just take the RF2014 scene and loaded it into RF2015? If yes I recommend recreating the scene from scratch. There are always changes which might not be recognized correctly.
Did you try a simulation without OpenCL support? OpenCL's influence is very small when it works, but might increase sim times drastically when the GPU isn't supported (unfortunately this can be the case even if RF says the GPU fulfills all requirements).
Or are there any unintended scale, geometry, or substeps changes? Just an idea...
Are export resource active you actually don't need?
Did you simulate via command line on RF2014 and through the viewport in RF2015?
Another thing I can think of is the domain's "Creation mode". Try Grid > Creation mode > Dense.
Finally, try a simulation with disabled "Particle Sampling" options (under the domain's node params).
In RF2015 we introduced a lot of changes to the calculation of distance fields for fluid-object interaction. It could be that the scene's objects influence this process and make everything slower.
I hope one of these tips solves your problem.
Did you just take the RF2014 scene and loaded it into RF2015? If yes I recommend recreating the scene from scratch. There are always changes which might not be recognized correctly.
Did you try a simulation without OpenCL support? OpenCL's influence is very small when it works, but might increase sim times drastically when the GPU isn't supported (unfortunately this can be the case even if RF says the GPU fulfills all requirements).
Or are there any unintended scale, geometry, or substeps changes? Just an idea...
Are export resource active you actually don't need?
Did you simulate via command line on RF2014 and through the viewport in RF2015?
Another thing I can think of is the domain's "Creation mode". Try Grid > Creation mode > Dense.
Finally, try a simulation with disabled "Particle Sampling" options (under the domain's node params).
In RF2015 we introduced a lot of changes to the calculation of distance fields for fluid-object interaction. It could be that the scene's objects influence this process and make everything slower.
I hope one of these tips solves your problem.
Thomas Schlick | Next Limit Technologies
Re: RF_2015 longer sim times
I have the same problem between "RealFlow 2015 9.0.0.0144"and "RealFlow 2015 9.1.1.0186 " !!!!!
Amazing.....
Amazing.....
Re: RF_2015 longer sim times
Import your collision object in "MultiBodies"....it will fix your problem (it is a bug when import in .SD as for me)