arnold vs. PRMan for FX


#1

hi.

I was about to switch my FX-pipeline from Renderman to Arnold because of nice volume rendering in Arnold as well as great raytracing capabilites for liquids etc.

Now I saw renderman is getting A LOT cheaper …

So what’s your choice when it comes to rendering refractive water, 40mio point-particles and scattering volumes?

Thanks


#2

No idea how prman behaves with volumes. Arnold does it quite nicely, but with more lights your rendertime increases a lot, same with indirect lighting for volumes. Rendering a lot of particles works quite nice in Arnold, but transparencies are handled much better in Prman as much as I know. All our tests lead to the conclusion that we do not render a lot of semi transparent things like leaves or particles. We do full geometry and use non transparent particles if possible. If you have a lot of semi transparent particles, Arnold becomes quite slow. And of course you don’t have shadow maps. Sometimes it is quite useful if you can use shadow maps to create soft lighting without introduce noise.


#3

Volumelights and fog is very nice and fast in Arnold. Polygon count is not a problem as well, people would rather use million poly objects instead of displacement mapped lowpoly objects since it would be faster.

Volume rendering for maya fluids isn’t very fast AFAIK. refractive transparent materials like glass and water render much slower compared to other raytracers. Forget caustics if you need them. Also opacity maps, like those for tree leafs, render very slow.

If you can live with these restrictions, arnold provides a fast feedback for lookdev, but for clean rendered final frames you better have a render farm. Maybe that’s why it’s sold only as 5 license packs.

In a perfect world, Arnold would be a GPU based renderer and would render as fast as Octane. :slight_smile:


#4

In a perfect world, unbiased renderers wouldn’t have problems with noise :stuck_out_tongue:


#5

Arnold renders like hundred times slower for the same noisy results. also Nothing is noise free, even realworld film cameras. That’s why they added film noise effects in post for years on 3d renders that were too clean to be true. :wink:


#6

With other raytracers you don’t mean renderman, I guess? … I’m curious how that new RIS will compete to the other raytracers

Well, when it comes to semi-transparent volumes renderman always gets very slow… similar with particles…

I guess I have to make some more comparisons between MR, renderman and arnold for volumes and renderman and arnold for points…


#7

That renderer exists and it’s called Redshift. :slight_smile:


#8

Arnold tends to get really slow in some cases .
For Volumes it is preferable to use renderman remember no software gives you out of box solution and you need to work with it’s strengths and weaknesses.
The latest renderman is really fast so if you have renderman stick with it.
GPU renderers are really great too .


#9

I was under the impression Redshift doesn’t deal with volumes yet. Has it been added? (or particles, or over-network splits, or topology variant instances, or a number of other things needed or convenient for FX work).


#10

For raytraced surfaces like water as well? compared to Arnold? … thought Arnold as a real raytracer should be faster for tracing …
(I have renderman studio 18 in my pipeline but didn’t use it recently for refractive stuff in production…)


#11

Forget mentalray for particles and volumes. During the last years I desperatly tried to deal with multiple intersecting volumes (fluids) and it was a horror. Most of volume shading in mr is done in the shader e.g. the raymarching what means the shader itself has to know how to deal with intersecting volumes. And almost no volume shader is able to do it. In Arnold the raymarching is done in the renderengine and the shader only calculates the density at the given point. So intersections are no problem any more.

Same with particles. Even if mr now has some sort of particle primitives, they only work with raytracing, not with rasterizer, and they do not support motionblur. With other words, they are pretty useless. Where mr needed a huge amount of memory to render only a few partilces, Arnold did it in a fraction of time with a fraction of memory.


#12

How do you actually manage to get opacityPP to work with Arnold? Or do you mean setting the particles semitransparent with an AiStandard shader?

Also, rgbPP seems to only work with nparticles, not particles. Is that true?

EDIT: got all working in the meantime with aiUserDataFloat and aiUserDataColor


#13

It seems when AB-comparing maya fluids between renderman and arnold that arnold scatters the volumes a lot more (no indirect lighting) than renderman which looks nicer.

With renderman the volume shadows to black a lot quicker depending on density than with arnold. Of course I can raise the deep-shadow color but that produces different results.

Is there something that can be tweaked in renderman to get more scattering (kind of sss-effect) without introducing expensive volume-scattering (which is of course very slow)?


#14

Just a information here : VRay is very fast with volume and particles.
For volume render, if you have VRay 2, you have to install Phoenix (the demo). After on the fluid shape, you will have an option to choose what shaders do you want to use.
Use the Phoenix shader, and you render will be divise by 7-8 (tested currently on a production, we created a lot of clouds (fly on clouds). Render time HD1080 : 4-5min).
This shader is now by default in VRay 3.

For particles, VRay is very optimized for this. Recently, you can add particles in VRay proxy to render billions and billions of particles.
A page about this : http://help.chaosgroup.com/vray/images/stuff/vrmesh_particles/vrmesh_particles.html

My 2 cents


#15

what’s this ?


#16

Great, thanks. I always forget vray. I’ll have a look.