V-Ray 3 SP3 sneak peek

Testament to the amount of work going on behind the scenes, I thought it worth it to share a couple more tests pitching SP2 as it is currently released versus the coming SP3.

Render settings for SP2 are as posted (everything else is at defaults, with GI, and Reflective GI Caustics, just enabled and left alone).
Render settings are largely what will be presented as defaults for SP3.

The differences you see are due to two main new features: a new and improved pixel sampler, and an optimized method for the resolution of reflective caustics. A third, but not seen, feature is a new approach to progressive sampling, whereby the noise threshold is reduced with subsequent passes, which leads to seeing more of the image sooner, as it’s rendering.

Four things are worthy of notice, here:
A) The noise distribution is more uniform (see the noise Map), the alpha sampled better, sooner. (see the alpha crops)
B) Halving noise threshold, with the new sampler, doubles rendertime, quite linearly and predictably, for twice the quality in the final image. (see the Sample Rate REs)
C) The light distribution is markedly better, especially in the darker areas. (The beauty renders will show this.)
D) Shading through transparent surfaces is markedly better thanks to the new reflective caustics approach (see the light fixtures in the second set)

Given that the new sampler puts effort in different places of a given image, exact time and noise level comparison are best left to individual cases.
In other words, do NOT expect massive speed boosts across the board, but rather a more pleasant result within the same-ish time as before, with little to no setup effort from the part of the user.

The noise values shown in the contact sheet text are calculated through a nuke expression.
It calculates the absolute value difference (per channel) of the current pixel against the average of the eight pixels surrounding it plus itself, similar to what the currently released V-Ray sampler does.
The noise map contains the values resulting from such operation in each pixel.

Links to the high-res contact sheets and EXR Renders:
white roomset : https://drive.google.com/file/d/0B1P9eRh8EqXGQXEwb1NmeUN6cms/view?usp=sharing
AI37_007 : https://drive.google.com/file/d/0B1P9eRh8EqXGUGdXeXY0VDFQQ1E/view?usp=sharing
Pack of EXR Images: https://drive.google.com/file/d/0B1P9eRh8EqXGWTdXS2NOZC1rN2s/view?usp=sharing




Could you upload those images to imgur at full res? hard to see whats going on.

We did a few shots using bf/lc on a recent animation, i’m looking forward to doing a comparison because optimizing became the bane of our lives and took forever.

Yes! I’m keen to see what this new workflow will do to some images I’m working on. When updating an older scene do you simply plug in those render settings - and all the existing light/material subdivs will be ignored?

https://drive.google.com/file/d/0B1P9eRh8EqXGQXEwb1NmeUN6cms/view?usp=sharing
https://drive.google.com/file/d/0B1P9eRh8EqXGUGdXeXY0VDFQQ1E/view?usp=sharing

The SP3 by default in a NEW scene, or when the user resets the render settings (as per Vlado’s tutorial, switching to the scanline renderer and back), will ignore the local subdivisions in lights and shaders, yes.

So, in a new scene the user does nothing, in an old one all it’s needed is a render reset.

Nice!!

Thanks lele for the approach.

The builds are in nightlies?

Cheers

The renders were done with a nightly build from 2 days ago, yes.

YUHUUUUUUU.!:sunglasses:

Very interesting!

Is the noise threshold value comparable between SP2 and SP3? Or asked in another way: If I have been happy with a noise threshold value of 0.005 in SP2, will that value still be the go-to value in SP3 or should I except to have to thinker with it? Your render times imply to me that I should to consider lowering it further, but I guess that is down to the kind of noise levels I can accept (and where). :face_with_spiral_eyes:

BTW: Funny you should use the AI37_007 scene. I was tested it on the GPU RT yesterday. :sunglasses:

Oh, and another question: Are these goodies mainly for progressive sampling?

Maybe SP3 should be called Vray 3 Corona Edition? (aka super easy render-and-forget-settings) :wink:

This type of comment is pretty troll and not sound very friendly. I think the effort to always go further, Chaosgroup generates small companies that make the effort to get to where this Vray.

But a comment irritating.

Oh, come on. It was meant as a joke. I don’t use Corona. :sunglasses:

hahahahahhahahahahaha

I test the new build and the visualization of progressive looks much better in the early approximation.
Really a nice change ..
The most difficult areas for sampling were much improved.
Now you can make decisions much faster because the textures are better understood from the first minutes of rendering

Wow, I don’t know about that. I just downloaded your AI37_007 renders, and the SP3 looks much improved in some areas than SP2 and is faster. Awesome sauce! :o

Just to double check, the latest nightlies are for 2011-2014?

No, they affect the bucket renderer as well.

(aka super easy render-and-forget-settings) ;)<shrug> If that’s what people like and want, I’d better make sure that they have it, right?

Best regards,
Vlado

And you deliver :wink:

Stan

I should start to deliver some beer to the Chaos-Team as a big thanks for all the effort! ¯\_(ツ)_/¯

Buuu the new build not sympathize with itoo products.

:cry:

Will this SP be officially announced? If so, how?