Have you tried V-Ray (RT) GPU ?

[LudvikKoutny]

I guess I am a bit late to the party, but every time I decided it's time to give V-Ray GPU a try, I get immediately turned off by all the render settings being just thrown into a single tab, with even more RT/GPU related rollouts there. It makes any kind of setup very frustrating experience. Also, consider that most V-Ray users have already layout of V-Ray (Adv) render settings in their muscle memory. I mean, since V-Ray Adv and V-Ray RT/GPU share 90% of the render settings, why not just get rid of RT completely, and add that single RT rollout to "Settings" tab of V-Ray ADV, and then, when mode is switched to RT or RT-GPU, freeze other render settings that apply only to Adv.

I requested it already somewhere, but it's been probably already a year, and still nothing happened. If it's difficult, then it would at the very least be good to divide all the rollouts in V-Ray RT into same tabs and same order as V-Ray Adv has, when set to production renderer mode. This would at least let us use the muscle memory we already have from using V-Ray Adv.

[savage309]

I get immediately turned off by all the render settings

We will have update on the progress for the problems that we discussed here, including (rather major) UI changes that we are planning in a few weeks.

Best,
Blago.

[kosso_olli]

We will have update on the progress for the problems that we discussed here, including (rather major) UI changes that we are planning in a few weeks.

Best,
Blago.

Looking forward to these changes. I hope the last nasty bugs might be ironed out as well.

[eyepiz]

I use GPU RT on a daily basis for dev look. My system runs a Quadro P4000 for the viewport only and have Titan's XP's for the cuda accelerators and lately been having driver issues mixing the GTX and Quadro's while working, lots of crashing.

So I decided to put all the Titans in a 10gbe render node. It's runs like a champ locally but when I try to use RT DR on my main machine it runs slow as hell. Looking into it closely it won't break the 1Gbe barrier. It hovers around the max 1gbe speeds and GPU's run about at 25% load.

The network runs all other apps and transfer at 10gbe speeds with ease.

I'm probably doing something wrong but the bottleneck defeated the entire purpose of the RT GPU DR render node.

Maybe someone can point me in the right direction. I really don't want to scrap this project.

-E

[matanov]

I use GPU RT on a daily basis for dev look. My system runs a Quadro P4000 for the viewport only and have Titan's XP's for the cuda accelerators and lately been having driver issues mixing the GTX and Quadro's while working, lots of crashing.

So I decided to put all the Titans in a 10gbe render node. It's runs like a champ locally but when I try to use RT DR on my main machine it runs slow as hell. Looking into it closely it won't break the 1Gbe barrier. It hovers around the max 1gbe speeds and GPU's run about at 25% load.

The network runs all other apps and transfer at 10gbe speeds with ease.

I'm probably doing something wrong but the bottleneck defeated the entire purpose of the RT GPU DR render node.

Maybe someone can point me in the right direction. I really don't want to scrap this project.

-E

Higher Ray Bundle Size?

[Nearl]

For work, I use it from time to time but, tend to switch back to Adv at some point in the project. So I generally just use it for early look development. I agree with some earlier posting from jironomo and BBB3. There are just a number of tools I'm use to using that aren't supported yet and my scene size/RAM (mostly from displacement) has a tendency to climb past what my 1080ti can hold.

One other note, unless I'm missing something, RT doesn't seem to support Body Object Rendering Approximation, it seems to just pull the Viewport Mesh Quality on render. We get a ton of .STP format models from clients for product rendering. RT would be perfect for this, if it was supported.

For personal work, mostly characters... I love it Just waiting for the AlShader to be supported but, I'm patient.

Also, it you are running active shade as a viewport in 3dsMax 2018, don't move any dialogues around...because if they auto-dock 3dsMax will crash. It's a good time O_O

[KhyeKading]

I've tried to migrate to GPU rendering (1080ti). our scenes are SSS intensive and it seems that the benefits of GPU aren't as apparent with SSS. Is this correct?

[matanov]

For work, I use it from time to time but, tend to switch back to Adv at some point in the project. So I generally just use it for early look development. I agree with some earlier posting from jironomo and BBB3. There are just a number of tools I'm use to using that aren't supported yet and my scene size/RAM (mostly from displacement) has a tendency to climb past what my 1080ti can hold.

One other note, unless I'm missing something, RT doesn't seem to support Body Object Rendering Approximation, it seems to just pull the Viewport Mesh Quality on render. We get a ton of .STP format models from clients for product rendering. RT would be perfect for this, if it was supported.

For personal work, mostly characters... I love it Just waiting for the AlShader to be supported but, I'm patient.

Also, it you are running active shade as a viewport in 3dsMax 2018, don't move any dialogues around...because if they auto-dock 3dsMax will crash. It's a good time O_O

RT will always use the Rendering Approximation of the Body Objects in the next release.

[maxwood]

I confess I haven't used it in a while, but definitely later than 3.5. I think last time I tried it I ended up wanting to the use the Al material that isn't compatible, so I swapped over to CPU. No big deal, it's not the most important shader but I like it.

I have trouble agreeing that CPU and GPU should be treated as entirely different renders, they really shouldn't be. Yes, I know they're completely different behind the scenes.

I work as a freelance artist and I have less and less free time outside of work to experiment with render engines. What I use is often dictated by what's being used in the studio I'm working at. I've not yet worked anywhere using V-Ray GPU for production, although I've heard one or two smaller studios using it. I think that's understandable though, studios aren't willing to invest in GPU's when they have CPU farms that do the job.

If V-Ray GPU is as good as other renderers like you say it is, then perhaps you have an image problem? Have you tried approaching studios and working with them to get GPU or even V-Ray in to their workflow? I spoke to a couple of studios in London and they all use Arnold, I know at least two are looking at moving to Redshift though. All it takes is one or two top studios to add GPU to their workflow it and people will want to use it. If studios are using it then do case studies, retweet their work, find out where V-Ray was used and make a point.

Lastly, I'm not sure who is in charge of your Twitter account but you need to take more care in what you retweet, the quality of some of the work shared reflects badly on V-Ray, especially compared to Twitter feeds for Corona, OTOY and Redshift.

[kosso_olli]

Lastly, I'm not sure who is in charge of your Twitter account but you need to take more care in what you retweet, the quality of some of the work shared reflects badly on V-Ray, especially compared to Twitter feeds for Corona, OTOY and Redshift.

Slightly off-topic, but I have to agree. It is the same on Facebook by the way.

[Nearl]

RT will always use the Rendering Approximation of the Body Objects in the next release.

SWEET! Love you guys

[savage309]

Hey guys,
As you might know, the V-Ray Next, Beta 2 for 3ds Max got released few days ago. We are very thankful for all the feedback here, and as a result of that feedback we have few changes in the V-Ray GPU Next.
Few things to mention:
* The V-Ray GPU has completely new (and hopefully nicer) UI. I want to thank everybody that helped with ideas and opinions for that.
* Active Shade and RT CPU are gone. Now you have V-Ray and V-Ray GPU. You can run them in Production or IPR. The results between production and IPR will always match.
* What is not supported in the V-Ray GPU is not in the UI anymore. So everything you see works (we might have missed something here and there, but we will fix those as we go).
* Displacement is much, much better. It will start up times faster and it will look many times better with much lower settings. This way it will also consume a lot less memory and will render faster.
* We have added support for glossy fresnel, environment fog, phoenix and openvdb and edge tex (so wire passes are possible now). Also there is support for the new hair shader, adaptive dome light, bercon noise, VRscans and many optimizations compared to 3.6. And many bugs should be fixed.

Just in case you missed it, here are the release notes for Beta 2. In the meantime, we will continue to work on the other notes from your feedback for Beta 3.

Thanks a lot,
Best,
Blago.

[maxwood]

Amazing work nice one Blago.

[thanulee]

Thanks Blago and team! Just found sometime to test vray Next today.
Didnt dive much into make final images. I wanted to see how interactive it feels -that is my main concern- cause im sure is fast for final image and due time all features will be there.
UI is finally making sense. Stuff just work and im glad u took this direction. Optix is a very good thing to have for specific type of scenes.

The downside is again for me the interactivity. Vray seems a lot heavier in same scene (simple interior no materials) than redshift and octane.
Its been a while im using octane as my main engine, sorry to say hehe, and i find it hard to go back to vray as much as i want to. Renders might be slower, but i work a lot faster and a lot more responsive, it just feels real time in octane especially.

Ofc, vray is a tool in our arsenal, as we ve invested years of working with it. I hope in the future, gpu improves in that area, so i can join the club again

Cheers,
Thanos.

[savage309]

The pipeline to get the pixels from the GPU and show them in the VFB can be better and the VFB2, beside all other features, should be optimised better for that. The GPU render engine/kernels, as you have noted, are already super fast. Overall the feedback should be not bad though. If you can send scene to Peter.matanov@chaosgroup.com and Blagovest.taskov@chaosgroup.com for us to check, that will be great.

Best,
Blago.