Im trying to get a acceptable rendertime for this:
Currently it needs about 40 minutes on my 5960x. The rendertime goes down to about 10 minutes when I hide the bottles.
Its an animation and most computers here are 2-3 times slower than my workstation which is the reason why I’m asking, 1.5-2 hours are simply too long.
I’d be happy if I could get a rendertime of about 20 minutes.
Its clearly the bottle material which is letting me down here.
Its not too important to get an super real image.
The material itself is nothing fancy just a refractive material with colored refraction map. Both glossiness parameters are at 1.
I tried to play with the refraction depth but surprisingly it didn’t help too much. I also tried different things like: full refraction with fog, fast sss2 (I must admit that I don’t know too much about that shader). And even opacity instead of refraction. I didn’t get very far with that.
There were a thread recently where someone tried to render a bunch of stacked translucent plastic bowls, which I couldn’t find anymore. I believe there was a trick.. I’m using BF+LC and the scene is lit with vray lights + vray softboxes only. Most geometry including the bottles are vrmeshes.
I’d be very happy if someone could give me a pointer. Its eating my nights already.
Would it be possible to send us that scene or a part of it for investigation?
Did you try to profile the scene further, its clear the slowdown comes from the bottles but have you noticed what specifically influence it the most?
For example did you try to isolate reflections/refractions?
Did you try to render the scene without GI?
What’s the difference in render times between depth 1 and 5?
What are the V-Ray Statistics(Number of raycasts,Camera rays etc) between scene render with and without bottles?
I managed to get to my target rendertime.
Here is what I did sorted from big difference in rendertime to small difference.
I deleted the back faces (shell) of the bottles and turned the IOR down to 1.05, to make them look empty again. (+1 for thin walled option)
Since I’m not planning to make much stuff in post, I turned the max ray intensity down to 3.
I also deleted many lights and replaced them with a few bigger ones.
And I simplified the materials.
The refraction depth is 5, since one didn’t really make anything better or much faster.
eh.. now I’ve got a big brightness difference from my workstation to the other computers.
Same vray versions… very strange… a test without the clipped static geometry part showed no difference.
What could cause this? All machines have the same vray version and acess to all assets. Im not using any other plugins, not even an osl.
This is probably going to be a rebus farm job… I’m running out of time.
almost like gamma isn`t being enabled for some reason ? if file gamma output is set to auto and the slave default gamma settings are different would that account for it ?
just a guess..
Unfortunately no. I can’t match the bright one if I gamma correct the dark. It just get washed out.
Tried to render with 1000 trans levels with no luck. Also, tried to render with a depth override of 20, also no luck.
It seems to be kind of random.. My workstation is always producing bright images while some clients sometimes render a bright one and sometimes not.
Most of them always render dark images.
Does this indicate a network issues? But I see no “map xyz couldn’t be loaded” kind of messages in the clients logs..
I’ve got the feeling that it’s something related to the clipper since its the only scene where this problem occurs.
swing and a miss, sorry. as for network if there`s a map that slaves can`t see/read it might change the reflection/brightness but you`d have seen the issue before i`m sure.
as svetlozar said above try with/without gi/reflections try hiding the bottles and vice versa and see if you can isolate the issue ? maybe a bit of damaged geometry or a co-planar issue could cause it.
have you tried it in 3.40.03 ? i assume you`re using latest.
maybe merge into new file.
The only thing which helped so far is hiding the big blue static vrmesh. And it shows no difference when rendered on one machine. Which is what I doing. I’m assuming the bright image is the right one and render as much as I can. I’ll give the rest to the rebus farm on Tuesday when I’m in the office again. And hopefully it was our crappy network or the ram of the render nodes.
Seems that it wasn’t the network. It still flickers even though I rendered on only one machine. I got the same problems without GI. I have no clue what to do. Will send the scene to the support.
Seems like I was right actually. The clipper is to blame. The slice modifier instead showed no difference between the frames.
If I only listened to my self
The clipper should be used with caution in animation jobs.
I thought I’d show the result.
I modelled the clipping by hand, fortunately it wasn’t very complex.
The quality suffers from the rendertime I had to reach finaly, since it took so long to get it under control. I didn’t even had the time to denoise everything.
I’m a bit annoyed about the bottles getting shot away above the rolls. This is because of bad collisions. It was four’o clock in the morning and I just thought: f**k it, nobody will care.(didn’t took me into account ;))
But what ever, I’m done and the client is happy
I’ll sleep a few days now I think
Its a MassFx rigid body simulation. That was the fun part of the project.
I used a simple collision shape for the bottles and build a simplified scene for the simulation.
Afterwards I replaced the bottles with vray proxy objects. And xRefed them in. I rendered the part above and below the rolls separately with different meshes for the proxy objects.
Then I just comped them together. It was pretty straight forward.
The client actually wanted them to get squashed by the press since the rolls are just perforating them. But he didn’t wanted to pay the extra effort and decided to be fine with what I suggested.
I have a project with heavy refraction, and render times are very manageable. This is with a rig with 7x 980 ti’s, and a render farm with more 980 ti’s though…not sure if that’s much help, but maybe something to consider for the future.
I think the biggest problem was the geometry which was too much. I didn’t know about the one voxel per mesh trick vlado mentioned here:- Chaos Forums
Which probably would have helped alot.
Deleting the backfaces of the bottles and turn the ior down to 1.05 also helped though.
