Interesting discussions… :keenly: Just catching the thread up. Without reading the whole thing, I am curious the main reasons major vfx houses (MPC, Framestore, Cinesite, etc) choosing Arnold or PRMan RIS over V-Ray?
Has it to do with how much/deep they get to access to the renderer to bend features/shaders the way they want it?
It seems that except DD and ILM (used V-Ray mostly for env works), big vfx shots (talking about 200+ artists globally) didn’t choose V-Ray as their main renderer.
The more things change…the more they stay the same! Been forever since I’ve been up in here, and the debate continues on! 
There are some great insights in this thread. Thanks for sharing!
-Justin
The larger a company is the more important is working time compared to rendertime. If you have enough machines, it doesn’t matter if a sequence takes 1h per frame or 10h per frame, it is done the next day. If an artist spend 4h to speedup a shot it can be more expensive than simply render longer. Much more important is how fast they can work and how interactive a renderer can be.
Yeah that’s why I’m inclined to stay far far away from the big-name renderers that the huge studios use since we don’t have a 1000-2500 node render farm and don’t work on short 5-10 second shots @24fps.
We routinely do continuous shot 30-60 second animation sequences at 4k, often in stereo @ 60fps and they have to render in 1-2 weeks or less. Spending 1-4 hours if necessary is a no-brainer to optimize renders so they don't take 2 months on our small 30-node render farm.
Optimization time actually isn't even done per project for us anymore since we've developed tools/workflows to auto-optimize settings as objects are created or assigned shaders since we have different tiers of ready-made shaders depending on how fancy/fast we can get away with in the time we have to render.
This really only works for us since most of our animations are pretty much always about human anatomy so it's essentially the same assets/scene files which is naturally going to be optimized to over time.
We’re using mental ray. However if we didn’t have so many reusable assets/shaders built on MR or if we didn’t have much experience with it, it would be far more cost effective to switch to vray if we were starting from scratch - even with its new 3x higher price. At this point though, it’s not worth it for us to rebuild assets and trading a set of known render problems for a set of unknown render problems, not to mention the additional costs of buying licenses and training time investment.
But every studio is differentAnother point to take into account for larger companys is how easily you can write your own procedurals, shaders etc.
The first time I tried it with VRay I was amazed because you have control down to the ray intersection procedures for a user defined geometry. On the other side, due to the lack of documentation and the complexity of the API, I was completly unable to create a simple triangle procedurally. But of course this is some years ago, I suppose the current VRay API is a bit better documented.
Creating procedurals and shaders with mentalray is great because you have access to a large scene database, but on the other side you have to deal with a lot of low level data, e.g. for the creation of hair data. But you can do quite crazy things with shaders even if you don’t have classes in shaders, it is still very 'C’ish. Creating the same tools for Arnold and Renderman as really fun. Creating the hair geometry procedural took about a few days what took several weeks with mentalray.
I switched to Redshift3d and never looked back at vray, arnold, mray, etc. Its just so,… fast.
gpu based, so you need a decent card.
Of course it doesn’t do particles yet and a few other things, but they are currently being worked on, and the dev team is awesome.
I absolutely love Redshift for personal stuff. Do you think it would work in render farm tho? Most our render nodes are blades with no GPU : /
A render node with GPU I suppose would take up more space, for the card itself and for cooling?
Yeah OTOY (Octane) and Redshift are getting so close to finishing renderers.
I played with Redshift before it had displacement capability.
I’ve seen Octane do displacement. Now that the disp hurdle is over there may be networks setup to render out to computers filled with high powered video cards.
I will say this though. I think that EXR loading is why people use finishing renderers.
In Tron Legacy the rectifier had 16 sections with 9 EXRS per section at 8k resolution to get all the detail. These were mip mapped so that they didn’t have to be loaded into RAM (video card limitation) and could start rendering right off without having to load the EXRs before engine launch.
I am not certain if this is an issue with GPUs but I assume it is.
Otherwise. yes. They are a viable solution for general CG work.
little hint… next week redshift 2 is arriving… :buttrock:
and with the new nvidia cards in 2016 this could be a turning point…
at least for smaller shops…
from Panos… " IOR nesting (think liquid in glass or droplets on bottle rendering problems), new shading models (like GGX), a new material layout (Redshift material or Redshift PBR), automatic VRAM management, baking, volumetrics, etc"
Next week would be a nice Christmas present!
Question for you guys. I have been tinkering in mental ray and can’t decide for materials if I should be using the mental ray materials vs the standard maya materials. I am working on a knight character and am trying to get the best result possible. I am downloading the trial for vray so will be tinkering with that as well, but the same question stands. Should I be using materials specific for the renderer? Or should I stick with standard Maya materials?
Oh missed this one…
I never have tried Clarisse and I am not sure I ever saw it deployed at Digital Domain. Does it have a linux build?
Try to learn the materials standardized by the engine. If you are using Mental Ray use the mia or mila materials. Though, I do wish it defaulted to no reflection turned on.
Mental Ray can use standard maya lights but you can really get nice realistic fall offs if you use the mental ray specific lights.
I have recently made a full switch to Redshift as well, and not looking back. Even in its current form, without all the features of Vray, it is just incredible. I am able to render my scenes faster with a single 980ti than using Vray on my small farm of 5 decent machines. Not only that, but it is much cheaper than Vray, and building a set of GPUs is also cheaper.
VrayRT doesn’t even come anywhere close by comparison, primarily due to its inherent design of full reliance on the GPU, making it an entirely separate renderer from its CPU counterpart.
For sprite or streak particles or fluids I can still render them with Maya Software or MR and then comp them on top of the Redshift renders, until Redshift is able to provide full support for them. Redshift supports the particle instancer very well though, and it’s fast rendering it.
just a note to the filtering talk…
The VRayLanczos filter in V-Ray is not a sharpening filter; it is similar to the area filter but with reduced moire effects.
Best regards,
Vlado
You should never use maya standard materials. Okay, there is one reason, if you want to reuse your scene with another renderer. The standard materials are simple not energy conserving what means that reflections are simply added to the diffuse value of the shader what can result in a unwanted bright and shiny material. On the other hand, most renderers offer a much better control in their own shaders where you can modify e.g. the contribution to the indirect lighting or simply a matte option which lets you use the shader as a matte without assigning a complete new shader. Arnold e.g. supports maya materials only partially.
I made an architecture specific forum in Renderman. and did test render there and comparing the result with Vray. I have been using vray for 7 years and when Renderman was made free to public , I had a chance to test it out.
both are capable of producing the same high quality result. though I am only familiar with Archviz, I am pretty sure both renderer are really fast. so in terms of render quality, I have no doubt Vray or renderman can do the job nicely.
it’s just… if u are working in ArchViz, then best weapon is to use Vray since there are TONS of 3D model u can buy or download are mostly available in Vray.
I like renderman so much but when it comes to downloadable asset on the net, Vray provide wayyy more availability. (see Archmodel from Volume 1 - 150 )
if you are working in ARchviz then its Vray obviously, if you are in Movie industry, then you would have to create every single asset one by one with your team, in which Renderman can save u a lot of time because the setting is simpler than Vray. (How I love Disney Shader). and btw, displacement in Renderman is super awesome.
Maxwell is extremely good too. the setting is so simple. layered material works really good and very easy to control. it has one disadvantage, super long render time. I use it in Rhino before gave up due to super long render time. its using light transit approach which simulate lights correctly as it is in real world (awesome reflection n chaustic effect. rendering Metal object is the best in Maxwell). it stores light data and u can change the light properties, its color, intensity, even turning on and off the light AFTER the rendering is done. honestly I think once the computer power is strong enough in the future, all renderer will go towards Light transit method.
I dont know much about arnold since I never use it but I heard its extremely good and capable of handling massive amount of geometry.
Thats all I can say,
happy renderinggggg