Strange banding in Clouds

Started by cyphyr, December 17, 2011, 11:22:55 am

Previous topic - Next topic

cyphyr

Hmm ok I've upped the atmo samples to 56 (with RTA ON) and the noise is pretty much gone but not entirely. I'll let the render complete and upload the result.
I'm wondering if this may be a case for having RTA OFF.
Richard
www.richardfraser.co.uk
https://www.facebook.com/RichardFraserVFX/
/|\

Ryzen 9 3900X @3.79Ghz, 64Gb (TG4 benchmark 6:20)
i7 5930K @3.5Ghz, 32Gb (TG4 benchmark 13.44)

Tangled-Universe

I'd say yes...

In my experience (quite extensive in the meantime) I figured that RTA is fast when using AA levels of ~4.
Above that it gets lots slower than without RTA.

In my latest work, still a WIP, I had a quadruple increase in rendertime when rendering RTA with AA8 and 8 atmo samples (the bottom limit the renderer accepts) versus RTA OFF and 16 atmo samples.

Matt

Hi Richard,

The banding and blobs are caused by an incorrect setting for sample jitter in the atmosphere or in the clouds. You should usually set sample jitter to 1. There may be some cases where you can reduce the jitter if nothing 'interesting' is happening in the atmosphere but I don't think they are common.

Matt
Just because milk is white doesn't mean that clouds are made of milk.

Oshyan

RTA enabled is only going to be slower for the desired results *if* noise is acceptable at lower sample levels. RTA can generally give you higher quality at equivalent or lower render times. I would guess the main reason for the render time issue in your case T-U is that you can't set the sampling for atmo lower than 8, and it's probably a daytime scene where you could get away with a lot less. Disabling RTA when using higher AA can definitely make sense, but again only if your clouds and atmosphere stay noise free with equivalent or lesser render time overall. In Richard's case where noise is still a problem *and* RTA is being used, even with high atmo settings, I suspect using higher AA will not be a problem - turning off RTA would likely just result in even more noise.

- Oshyan

Tangled-Universe

Quote from: Oshyan on December 20, 2011, 01:20:50 am
RTA enabled is only going to be slower for the desired results *if* noise is acceptable at lower sample levels. RTA can generally give you higher quality at equivalent or lower render times. I would guess the main reason for the render time issue in your case T-U is that you can't set the sampling for atmo lower than 8, and it's probably a daytime scene where you could get away with a lot less. Disabling RTA when using higher AA can definitely make sense, but again only if your clouds and atmosphere stay noise free with equivalent or lesser render time overall. In Richard's case where noise is still a problem *and* RTA is being used, even with high atmo settings, I suspect using higher AA will not be a problem - turning off RTA would likely just result in even more noise.

- Oshyan


I don't wanna be a bitch Oshyan but I think I don't agree with that :)

You have seen my Dutch Landscape WIP which has a plain sky at dusk with clouds in the far background.
With that WIP I had 4x longer rendertimes when using RTA with AA8 versus RTA OFF and 16 atmo samples.
As you said this has a rather simple sky and it may not need many samples with AA8, likely lower than 8, which is technically not possible atm.
So in that case I can see your point.

However, the scene Frank and I are working on, crater lake, has the same. We use 4 or 5 cloud layers there and it's ~2-3 times faster to render it without RTA since we use AA8 here as well.

I think there's more to it, perhaps need to make a thorough test some time.

cyphyr

I think the main difference between our scenes is the "thickness" of the atmosphere. In other words the haze density.
In order to see the volumetric light rays I have had to raise the density to 5 (I could play about with this but there are just so many permutations I'd like to keep one part constant!).
I have not seen "Crater Lake" wip but from memory your Dutch landscape had quite a clear atmosphere, ie a lower haze density. I think it is this density that causes my longer reder times due to the increased noise. I tried with RTA OFF and the frame was about half way through after 2.5 hours, the quality was not great either. I don't mind a "little" noise as Nuke has a node to remove noise and flicker that is surprisingly effective.
Cheers
Richard
www.richardfraser.co.uk
https://www.facebook.com/RichardFraserVFX/
/|\

Ryzen 9 3900X @3.79Ghz, 64Gb (TG4 benchmark 6:20)
i7 5930K @3.5Ghz, 32Gb (TG4 benchmark 13.44)

Tangled-Universe


Matt

Quote from: Tangled-Universe on December 19, 2011, 09:45:44 am
8 atmo samples (the bottom limit the renderer accepts)


The lowest number the renderer accepts here is 1.
Just because milk is white doesn't mean that clouds are made of milk.

Oshyan

I seem to recall values lower than 8 in the atmo node resulted in some render issues though. Have to test again I guess.

T-U, tuning RTA settings can be tricky sometimes, but I've seldom seen it result in across-the-board longer render times unless it's in a situation where, as I said, you simply can't lower the sample/quality settings enough to compensate for the high AA level (i.e. a simple atmosphere situation). I'd be curious to see your scene and try a few things myself thoughg. Even if the render times were higher with RTA, 2-3 times faster without seems like a *lot*. Any chance you could share the scene? ;D

- Oshyan

Tangled-Universe

December 22, 2011, 03:56:33 am #24 Last Edit: December 22, 2011, 04:01:46 am by Tangled-Universe
Quote from: Matt on December 21, 2011, 02:35:00 pm
Quote from: Tangled-Universe on December 19, 2011, 09:45:44 am
8 atmo samples (the bottom limit the renderer accepts)


The lowest number the renderer accepts here is 1.


Since when is that? I tried that some times and below 8 the atmosphere turns entirely gray (I believe, at least something happened which indicated 8 is the limit).

By the way, if you replied to what I said literally then you're always right since TG2 doesn't prevent me from entering a value lower than 8 :)

Tangled-Universe

Quote from: Oshyan on December 21, 2011, 07:06:41 pm
I seem to recall values lower than 8 in the atmo node resulted in some render issues though. Have to test again I guess.

T-U, tuning RTA settings can be tricky sometimes, but I've seldom seen it result in across-the-board longer render times unless it's in a situation where, as I said, you simply can't lower the sample/quality settings enough to compensate for the high AA level (i.e. a simple atmosphere situation). I'd be curious to see your scene and try a few things myself thoughg. Even if the render times were higher with RTA, 2-3 times faster without seems like a *lot*. Any chance you could share the scene? ;D

- Oshyan


That may be the issue indeed. Frank created the cloudsystem in that work. I'm pretty sure he's fine with having you do tests with it, but I'll doublecheck and will send it out to you :) Interesting stuff :)

Cheers,
Martin

Matt

December 22, 2011, 07:07:17 pm #26 Last Edit: December 22, 2011, 07:08:53 pm by Matt
@T-U,

The main issue I'm aware of with low samples in the atmosphere is noticeable transitions between different levels of noise. The actual number of samples increases as you look through longer paths in the atmosphere, changing the amount of noise. If noise levels are low enough (e.g. with ray-trace atmosphere enabled and high anti-aliasing), you probably won't see these transitions. I haven't seen the grey atmosphere problem.
Just because milk is white doesn't mean that clouds are made of milk.

Tangled-Universe

Thanks Matt,

I can reproduce the behaviour you described when I'm not using RTA with <8 atmo samples.
I suspect somewhere in the meantime something has changed as I can't get the artefacts reproduced anymore.
It's good to know, and to see, that I can render with RTA and 1 or 2 atmo samples.

However, as I will soon show, this has very little impact on rendertime.

Oshyan, I have devised some tests which I need to put in a table-like sheet to present clearly.
I'm sure you'll find the results interesting :)

Tangled-Universe

December 27, 2011, 08:51:52 am #28 Last Edit: December 27, 2011, 08:56:40 am by Tangled-Universe
Here's test 1:

At the top centre is the benchmark which has been rendered without RTA and with 24 atmo and 220 cloud samples which is equivalent to a detail setting of 0.76.
I then tested with 4 different atmo sample settings and 3 different cloud sample settings, all with RTA enabled.
The AA setting was 4 for all tests.

My conclusion in regard to atmosphere:

In this particular test adding more cloud samples did increase rendertime just slightly for all 3 cloud sample settings.
So, lowering atmosphere samples to <8 with RTA enabled does not change rendertimes dramatically.
This fits perfectly with my observation I told here before where my image with RTA @ AA8 and 8 atmo samples rendered 4 times faster when without RTA and 24 atmo samples.

My conclusion in regard to clouds:

In this particular test the noise levels are visible at the bottom left and right of the clouds.
When comparing the RTA OFF vs RTA ON conditions you'll notice that @ roughly 100 samples the result is similar, regardless of atmosphere samples.
However, rendertimes are ~2x longer.
Rendering with 40 atmosphere samples gives worse results and not much improvement in regard to rendertime.

Combined conclusion:

Rendering with RTA at similarly resulting noise levels is ~2x slower.
Rendering with RTA with more resulting noise is a little bit faster.
As discussed before, reducing atmosphere sample levels below 8 is not the cause for slow rendertimes with RTA.

So I will not advocate for RTA here anymore ;) and especially not when using AA>4, since that gives even bigger discrepancies as I explained before.
Due to the exponential sampling nature of AA this makes sense to me.

One can of course argue about the test setup, but so far these findings fit with what I've recently been experiencing. (after some discussions with Frank who mentioned it first to me that RTA isn't faster.)

Cheers,
Martin

Oshyan

Interesting results indeed! I've benchmarked quite a lot of scenes showing a marked improvement in speed/quality with RTA (which I will now go and verify ;D). So it must be scene-dependent, which is certainly possible. This is why I asked for the scene you were working with so I can see if there's anything else going on. It would definitely be good to get a more comprehensive picture of how all this works in practical scenes.

- Oshyan