Hey folks,
Just wanted to check how many of you have tried V-Ray (RT) GPU in V-Ray 3.5 or later?
If so, what hardware have you used, do you have dedicated GPU for monitor and has it worked for you in general?
If not - why? Was it startup time, render speed, UI, lack of features, interactivity, stability or something else?
Thanks,
Best,
Blago.
Hi
I gave it a try a while ago (before 3.5 update though) and I must say it was slower than CPU rendering on the workstation I had (i7-5930K cpu, GTX 970). Looks like its not for me. Probably it will be faster if I put second GPU in, but I’d rather use render farm that we already have.
Not sure if that helps, just my 2 cents.
I tried it when the Hybrid renderer was introduced. Since then, I sometimes try it at the start of a project, but always end up going back to the regular V-Ray Adv. due to lack of features. For example, If I use EdgeTex to get lines in my renders, the lines are visible in the light cache pre-pass, but as soon as it starts rendering, the lines are gone.
I’ve also come across exposure inconsistencies between V-Ray Adv. and RT, where RT seemed to ignore the Exposure control. Could’ve just been a bug, but there are too many of these small things that stops me from using it.
I must also say that it’s a bit unclear what name you use… In this post you call it “V-Ray GPU”, but when switching to it in 3ds Max, it’s called “V-Ray RT”. And it has also been called “V-Ray Hybrid”. So… which one is it? =)
My specs:
2 x Intel Xeon E5-1650 v4 @ 3.6 GHz
64 Gb ram
Nvidia Quadro M4000
Windows 7
3ds Max 2018.2 (as of today, earlier tests was with both 2017 and first iteration of 2018)
V-Ray 3.60.01
I tried it a number of times, and I always come to the conclusion that our projects are way too big to be handled by the GPU.
To this day, I wasn’t able to render an interior shot of a car, because it exceeds the RAM limit by a very wide margin. Also, I would be missing the VrScans, and I had problems using the Displacement on GPU. I had strange triangle artifacts, there is a thread about this somewhere. However, this was two releases ago, so things might have changed since then.
Just an idea: Wouldn’t it make sense to send this out as a survey to all customers? I remember something similar being done with the VrScans shortly after release.
Yes, but we need to know what kind of choices (if any) to put in the survey.
Best regards,
Vlado
Since you helped me debug my slow RT GPU scene the other day, I’ve been testing it in a number of older projects and I’m getting really good performance on one lonely 1080Ti. I can see myself using it more and more on simpler scenes and for animations.
Right now though, I’m hitting two big limitations:
1. Displacement. I can fit my entire garden scene on my GPU’s 11GB, which actually was a huge surprise given the sheer amount of geometry and 4K maps all over the place, and it renders fast. But the scene just won’t fit with displacement. With some serious optimization (replacing some displaced surfaced with retopoed meshes with normal maps for instance), I’m sure I could make it fit and get about the same final quality. But it would need starting the project in a different mindset, much like people might be going about building a scene intended for Unreal.
2. No glossy fresnel. This was one of the main leaps for me recently in terms of realism. Not having it on the GPU feels like taking a big step back.
Hi, i have tested RT in many versions till now and im really willing to make it my main engine.
Unfortunately, the lack of features in our latest project couldnt let me actually see how it performs in some heavier scenes (missing Alshader which was key component in every shot). So one thing is that i d like 1:1 the features i use in VrayAdv + caustics.
Another thing i want to address is displacement. I experienced many crashes especially when i fiddle with displacement parameters real time, even in simple scenes. Same goes for hair (ornatrix tests). Missing stability, although renders really fast.
The most important thing that leads me back to Octane everytime when it comes to gpu, is the interactivity. I think Octane’s interactivity in that matter should be an example of how real time should work. It feels so light and i can see stuff as i am moving em, while vray delays to update. Try it out and u ll see the difference 
Besides these, overall i am really happy about the support i get and the updates from u guys and im sure our voices will be heard
Thanks!!
Though Octane only supports changes to materials and lights in real time, right? Transforming geometry in any way, even a simple translate, forces you to manually refresh the viewport.
Not sure about deformations, but if u set objects as movable proxies (in 3ds max) translations are real time. In c4d that i mostly utilise it, all is real time.
Octane has A LOT of bad stuff, but real time aint one of em
Oh, ok. I’m referring to the Max plugin, which I guess isn’t as well integrated as in C4D. There, anything that involves geometry (moving, deforming, hiding, revealing…) doesn’t get reflected in the Octane viewport until you flush it.
Have u tried using the movable proxy option? Its in octane properties and thats what it does. Havent used it in a while in max, but im quite positive it does transformations at least In any case, the optimal behavior is everything, thats why i mentioned c4d.
edit: check this out https://www.youtube.com/watch?v=31FvXIOxEVk
Gotcha. Will check it out.
Try to do that with a) production shaders and b) deforming geometry
Best regards,
Vlado
Dont get me wrong Vlado, i dont intend to undermine vray rt hehe Also, I dont wanna be talking about other engines in general, was just replying to Bertrand.
Its just my personal opinion though, i feel that Octane is behaving as i want in real time, even in complex scenes. Love to hear thoughts from other people as well on this
Yeah, I know, I just couldn’t help myself We are asking precisely because we want your feedback.
Best regards,
Vlado
We’ve been using it in production for about a year now. My workstation has 2 x titan X and we also have a small farm with 8 more and another workstation with 2. Most of the work is for tv infomercials (ie. technical animations) so it’s not high level film work. The projects generally have mid level CAD files (20 + million polys) so there aren’t any problems with the RAM (12 Gig). Startup time on some scenes with larger CAD files for RT GPU Activeshade is a bit long, but works. Although I’m generally happy with the updates and stability, the latest version was not a good version for our setup as there were fundamentally bad bugs which required us to downgrade to 3.5 (I’ve reported the bug in another thread).
Hi Blago
I tend to use the IPR more so because I find it:
1. Starts quicker particularly loading maps and geometry.
2. Has less issues with materials.
3. Better stability - I have one model that always crashes with V-Ray RT no issues with IPR I should send it to you for analysis.
4. I don’t have to change between renderers it’s just one click on the production renderer.
5. Works well with forest pack imho.
6. I feel it’s a truer representation as V-Ray RT is using limited rays for GI and Trace depth - I appreciate I can change the settings.
7. For large models I don’t have to worry about graphic card memory limitations.
I have a single Nvidia GTX 970 (3GB) in a I7 5960x @ 4.2GHz 64GB workstation 3dsmax 2016 with V-Ray 3.60.03
We have been playing with RT gpu for about 6 months now. There is so much potential but just a few small things are holding us back from making the switch permanently. We are a visualisation studio doing a lot of archviz and product viz and the following things are crucial for us:
1. Displacement. Would be great if this could somehow be handled better on gpu.
2. Forest pack tint map support. Need this for grass and some plant shaders
3. Forest pack edge mode. For nice clean grass edges.
4. Grow fx not fully supported in gpu. Need interactivity when adjusting grow fx models.
5. A few more render elements. Wirecolor in particular… but I think this is coming in the next release?
6. Alpha layer. Need to be able to assign alpha contribution to objects.
7. Translucency. 2 sided material translucency maps don’t work in gpu. Also translucency in vray mat would be nice.
8. Environment fog.
9. Stability. Get quite a lot of crashes when using RT active shade mode.
Basically I reckon Rt Gpu is the future and we are super keen to switch to it asap but just need support for a few more features. So I would love to see you guys just concentrate on getting all the features of vray adv onto Rt gpu.
And in answer to your actual questions Blago: We use 970’s for viewport and 1070’s for gpu rendering (but will soon update to1080ti’s for the greater vram and speed), render speeds are fantastic compared to cpu (external shots more so than internals) and even better now with hybrid rendering.
I have 2 x 1080ti and i didn’t intent to use GPU engine in out production pipeline yet since we have bigger scene for archviz so i knew it will be limited.
But i did some tests nevertheless and indeed pretty much whatever was mentioned in here i experienced also.
- displacement is clearing our slowly
- even on light scenes i couldn’t load many time because of displ
- still some features don’t work (i am comparing to CPU where everything works as expected, i know it is different engine but still i end up going back to CPU) some gradients, vray mesh clipper, caustics, env fog and such
- unable to render scenes and animations just by switching from ADV to RT. i am aware again that it requires different approach but it still gets me, that even simpler and smaller scene usually require a lots of optimizing
- one bug or issue (that i will report) is that i am loosing g-cards connection randomly, all of the sudden it stops reading them for what ever reason. I will look into it and when i learn more i will also report it if it is not HW issue.
- i had 960ti for viewport and 1080ti for render and all worked great, now i have 2x1080ti and running on low priority and i notice lower interactivity (like noticeably lower)
I would love to continue on Vray RT, cause it really is more fun and all, but unfortunately i cannot really use it still in pipeline.
Dont know if this helps but this are my thought so far,
cheers
I’ve written about my RT GPU experiences a few times now as it had a big positive impact on how I now work.
My last big project I had to be done very quickly, and the timescales didn’t allow anything else in my set up but to go for RT - and RT (GPU) was a God send.
I have 4 x 1080ti’s, and saw my render times reduce from 26mins (Adv CPU) down to 6mins (RT GPU).
The only thing I found annoying from the whole RT process, and this is a small thing was the VRay Log messages would zoom through very quickly with no ability to scroll back through them all afterwards - it was as though I had access to only the last 100 lines or something (so could not ever see what any warnings or error were about - I was having to take screen shots using my phone !).
Some of the warning messages too could have been more informative, giving more clues which materials (names) were an issue (i.e exactly where maps were being used - as 3DS Max doesn’t have terribly easy ways, or existent ways, of finding materials easily).
RT GPU has really changed the way I work so much that it is now my renderer of choice.
PS - I did try Hybrid, but found in my case (4 x 1080ti’s and 1 x 5960x) that adding the CPU slowed my renders - surely ‘more hands to the pump’ ought to be a speed improvement ?