V-Ray 3.30.01 BETA! available for download

[^Lele^]

That is expected, yes.
As mentioned elsewhere, the noise threshold for the new method represents something else than it did before, and as a result one HAS to adjust it compared to previous scenes.
While doing this helps match a previous noise level in a specific area, or across an image, the noise DISTRIBUTION across the image will be different, regardless of user efforts (there's ofc always the option to use the old sampler.).

[thablanch]

Going with the 'match the noise level' game right now!
average 25% faster on my scenes for now (once the noise level is kinda the same).. interesting, I like that!

[joconnell]

Hi, I don't know if I am posting in the relevant thread, sorry if I'm wrong but I would really like to have a VrayUserBitmap.
The same way the VrayUserColor works, but with a string variable that points on a path\name to make the same shader point on different bitmaps...

The vray hdri map can do that - https://vimeo.com/100399778

[mienda]

Hi,

As i understand, the new sampling method from vray 3.20.03 or beta does work well only with color mapping ticked on "none" or "color mapping only(no gamma)".
Compared to vray 3.20.01, if we choose to use "color mapping and gamma" (into the color mapping menu) the results are weird and false.
Is this option intended to disappear ( sometimes i used it when i wanted to speed my renders et work only in 8 bits) ?

Thanks

mienda

[kosso_olli]

Some more images. Left side is 3.2, right side is 3.3
Rendertime and noise threshold are identical, around 29mins for the full image, threshold 0.01.
The scene has sss for the fluid inside the bottle, it is much cleaner in the new version.

[Morne]

Some more images. Left side is 3.2, right side is 3.3
Rendertime and noise threshold are identical, around 29mins for the full image, threshold 0.01.
The scene has sss for the fluid inside the bottle, it is much cleaner in the new version.

EDIT: nevermind, had to refresh my browser

[kosso_olli]

Fixed, sorry

[vlado]

As i understand, the new sampling method from vray 3.20.03 or beta does work well only with color mapping ticked on "none" or "color mapping only(no gamma)".
Compared to vray 3.20.01, if we choose to use "color mapping and gamma" (into the color mapping menu) the results are weird and false.

I'm not aware of any issues; it would be good if you can post a scene.

Best regards,
Vlado

[^Lele^]

no need, Mienda, I did already.

As a workaround, for now, you can try and lower the noise threshold A LOT (10 or 100 times lower than what you had before... you'll have to specify a higher spinner precision in the max preferences, probably.), as if there is (soft) clamping at sampling time, the variance lowers dramatically, and the sampler misses out.
IT's just a case of making sure it takes notice, so to speak, which can be done by drastically raising the sensitivity through the noise threshold control.

Especially now that we are all getting used to the new way of working of the sampler, the SampleRate render element is our best friend.
There will always be a direct relationship between the look of that RE for a given scene, and its cleanliness/render speed, and looking at it will most often give us clues as to what is happening in the scene.

Standby for a set of samples illustrating the behaviour.

[rikou]

Hi,

As i understand, the new sampling method from vray 3.20.03 or beta does work well only with color mapping ticked on "none" or "color mapping only(no gamma)".
Compared to vray 3.20.01, if we choose to use "color mapping and gamma" (into the color mapping menu) the results are weird and false.
Is this option intended to disappear ( sometimes i used it when i wanted to speed my renders et work only in 8 bits) ?

Thanks

mienda

I'm really interested in that because I am in the same case.
It would be good to have more information on this subject...

[mienda]

Thank's Lele !
indeed before my post, I tried to render with a very very low noise treshold ( 0,0001 or 0,005) it works very well but i was surprised to get at this kind of extreme!
But i understand the internal logic of the process...great update, vray becomes manageable very effectively without these "old shadow's areas"!

regard's

mienda

[^Lele^]

From top to bottom:
clamped, with n.t. @0.01, the sampler clearly misses.
unclamped, same n.t.: the sampler finds the data it needs
clamped, with 10 times smaller n.t., the sampler finds more data.
The next trio is the same, but with a Reinhard Burn value of 0.25 (something i saw a lot in evermotion scenes, but which i hope no one actually uses...)

[WilleViz]

...
The next trio is the same, but with a Reinhard Burn value of 0.25 (something i saw a lot in evermotion scenes, but which i hope no one actually uses...)

Personally, I like my steak rare, so I normally leave the Reinhard Burn value at 1.0 (fully linear), save out in EXR, and do the tone mapping in 32 bit post, thank-you-very-much.

But if people just want to save out a jpeg from the VFB, I can see why it might make sense do that on a scene with a ton of contrast? The LearnVray course even advocate using Exponential at those times, so I'm not totally surprised that Evermotion did that. Having said that, a Reinhard Burn value of 0.25 seems a bit extreme.

Do you advise against using the Reinhard Burn value, ^Lele^?

BTW: I might use a Reinhard Burn value of 0.8.

[thablanch]

The next trio is the same, but with a Reinhard Burn value of 0.25 (something i saw a lot in evermotion scenes, but which i hope no one actually uses...)

Yep, always

[^Lele^]

Well, I advise leaving the radiometrically linear light transport equation alone.
As you can already see from the samples above, the colormapping done this way skews the light distribution in ways which won't be recoverable in post.
Also, I like the fact that if i double a light intensity, it actually doubles in my VFB, instead of going up by maybe, umh, yeh much, kinda.

Way too many times, during my time at a certain company, i was asked to double a certain light's intensity ("stop it up, will ya?"), but given we were under LOG instead of Linear (ie. closer to what lowering burn does to your highlights), doubling a light intensity in the UI meant absolutely nothing could be expected from the render, as the resulting intensity would more than double in the faintest areas, and less than double in the brighter ones, often less than perceptibly so.
Cue endless to-and-fro, borked daily sessions, babysitting at the UI to make sure we actually did double the light (forreal.), and so on.

How anyone could wish to work in a non-linear environment escapes me, especially when actually given the option. ^^
Then again, as long as it's an option that can be used, it needs full support, does it not?

edit: speaking of colormapping and endless looping, that's roughly the name of my past week trying to get a viable interior scene to be set up for proper comparison.
Anything coming from the internet (commercially so!) seems to be obstinately set up with all sorts of colormapping weirdness, to compensate for lighting and shaders odd setups.
I'd be all ears for a collection of scenes, or even just the ONE scene set up with proper LWF, so that i could then compare apples to apples, instead of turning a pear into an apple to compare it with an apple.
Which is as convoluted as it sounds. ^^