HD Rendertimes with new render settings

[olika]

Need some feedback from the community

Attached is a frame from current animation,
Full Interior, few lights on the ceiling and on the walls, moving camera and objects, BF-LC

I updated the scene and set it up like vlado suggest in his sample tutorial
Now i also have MSR 8, Image Sampler 1/24 0,008 Global Subdiv 0
Not using the new beta as i`m into a project.

Have to say Quality is alright with the settings but rendertime is about a hour on a dual xeon z800 workstation.

I ask myself is that fine? go for a walk and shut up like vlado suggest or is there something left to tweak for speed or is 1 hour totally off and you`ll say hey c`mon that scene will take 15 minutes with a correct setup?

...honestly still strugglin with just accepting rendertimes and enjoying the new freedom haha

In the past with tweaking i had somehow the feeling you find a lot of things in the scene which gives vray a hard time ( noise makers, tough strong reflektions of . i.e. a light with a high multiplier) and step by step solved it by working through the scene.
How to deal with that in the future? If you just build up the scene and rely on the new render settings and hit render, how do you know if there is a bad guy in the scene who slows down the whole thing cause vray for sure tries its best?
In other words i completly believe in vray and it`s abilities, what i have to fight the most time is my own dumbness and mistakes i make while building up a scene do you know what i mean??

For example just put in the new settings, everything looks fine except maybe the GI Pass, it`s a bit noisy, so what to do? Doesn`t make sense to change settings GI Settings because of the Globas Subdiv Multi = 0 right?
So whats left to tweak in that case? I feel like MSR don`t change much to the GI Noise as the color threshold does but with the cost of rendertime...

Does the new update (beta) something to this?

Thanks in advance!

[Raven]

What does the sample rate pass look like?

Just looking at the scene I think you could have a higher MSR, let the lights, mats and GI do a bit more of the work.
At the moment I think that the AA is doing far to much work on this scene with large glossy surfaces, not real tiny details.
You may still have to play around with it.

Pick a small region and do a couple test at 12, 16, 24 etc.

You will always be able to shave more time and get better quality with going the old route of trouble shooting element by element.
My understanding is the new system is to get "quick" and "clean" results with little input. It's not a magic button.

[olika]

Thanks for reply!
Sample Rate is attached,
i`m wondering in theory, not so much details (no hair, foilage etc.) means make sense to bring down Max Subdivs. But then Sample Rate turns mainly Green/Red. How to compensate to come back to a balanced SampleRate? Increase Color Treshold or MSR? Do i have to forget about the old balanced sample rate? I mean mainly blue on flat surfaces and good sampling near edges. I feel like a little lost between old and new methods

[olika]

hm as soon as i set the Global Subdivs Multi to 0 there is no more way to control a specific thing like i.e. a light, right? Upping Subdivs has no effect anymore. Did that often in the past with the lighting Domelight or GI BF Subdivs.

[olika]

Ok, have some interesting update here,

iwas just not satisfied with the noisy GI Pass so i stripped down the scene, turned off all lights and then realized there was still some grainy noise in the GI Pass.
I checked all materials especially those vray light mats from the ceiling spots (The spot has a vraylightmat but there is a invisible vraylight in front of it to lighten the scene)
So all mats have the "no direct light" and GI is alos unticked, but then there was still some noisy lighting in the GI pass?!
Never had this issue before, i was pretty sure the settings mentioned above takes care that vraylight mats just good looking and not intefering the lighting but it seems i was wrong?
Next step was to go into vrayproperties from those models (the light ring in the ceiling, light ring at the system and also spotlights, everything which is selfilluminating in other words)
After i did that suddenly the GI Pass was correct (with correct i mean completely black with all lights turned off)
and with lights turned on i have a way cleaner GI Pass and for sure the rendering speeds up a lot now!
The scene now takes only 10 minutes to render! and settings are now MSR 8, Image Sampler 1/16 0,008 Global Subdiv 0
Does somebody else had that thing ever before?

[Morne]

It sure is interesting, because in the past we used fake planes in shops set to invisible to camera, reflections, refractions, but then vraylightmtl applied to it. We did this to SPEED UP renders.
Sounds like you're getting the reverse

[joconnell]

Never had this issue before, i was pretty sure the settings mentioned above takes care that vraylight mats just good looking and not intefering the lighting but it seems i was wrong?
Next step was to go into vrayproperties from those models (the light ring in the ceiling, light ring at the system and also spotlights, everything which is selfilluminating in other words)
After i did that suddenly the GI Pass was correct (with correct i mean completely black with all lights turned off)

Could you clarify what you did here further? Did you use the vray properties to also turn off generate gi or something?

[olika]

Oh sorry deleted the second half of the sentence without realizing it...
Indeed i turned everything off related to GI (in vrayproperties: visible to GI / generate and receive GI)
also in the material self illu slot turned off GI
so then there was none of the lighting materials showing up in the GI Pass anymore.
Because the GI contribution they were creating was very noisy, so after turning off GI contribution in those objects, the pass was way cleaner / easier to get it clean, that is less work for the sampler and therefor faster render times. does that make sense?
I just thought self illu materials with turned off GI in material settings would be enough to keep them from showing off in the GI pass...

[Morne]

No this doesnt make ANY sense. Are you sure you didn't by accident tick "direct light" in the VRayLightMtl?
Any VRayLightMtl I use (without ticking direct light), gives me a pretty smooth and clean GI. As soon as I tick "direct light", that's when things go haywire and very noisy and extreme render times

[olika]

Hi Morne,
i`m totally with you, never had a issue like that before. will double check render pass in future projects for that behavior.
it comes to me now when i scrub through the scene again (checking accidental ticking hehe), I forgot to mention that the model and shader is coming from another agency. The Lamp has a metal case and a plane (the emitting lamp) which are intersecting. The material is a blend configuration, one glass shader blended with light mat.
Just guessing but maybe in that particular case (intersecting or blend mat) GI on offs somewhere in the shader then behaves wrong?
Seems very rar situation in my case but as i said well keep an eye on that in the future.
Maaybe all old stuff again with the coming update we`ll see but as always thanks a lot for helping!

[svetlozar.draganov]

Have you found where the noise come from eventually?

[joconnell]

Oh sorry deleted the second half of the sentence without realizing it...
Indeed i turned everything off related to GI (in vrayproperties: visible to GI / generate and receive GI)
also in the material self illu slot turned off GI
so then there was none of the lighting materials showing up in the GI Pass anymore.
Because the GI contribution they were creating was very noisy, so after turning off GI contribution in those objects, the pass was way cleaner / easier to get it clean, that is less work for the sampler and therefor faster render times. does that make sense?
I just thought self illu materials with turned off GI in material settings would be enough to keep them from showing off in the GI pass...

Yep, Cheers for the clarification!

[svetlozar.draganov]

I've missed that part - thanks for pointing it out