Understanding DMC Sampler

[joconnell]

I'm working on a small interior at the moment and am using this to experiment with. Here are a few of the images:-

[ATTACH=CONFIG]16406[/ATTACH][ATTACH=CONFIG]16407[/ATTACH][ATTACH=CONFIG]16408[/ATTACH][ATTACH=CONFIG]16409[/ATTACH]

This took around 15 minutes when using DR across 6 high-spec machines (i7, 32GB). It is 1000x1000 pixels for my testing. I have applied a mid-grey material to everything.

Adaptive DMC image sampler Min=1, Max=16, Clr thresh=0.01
Primary GI = BF with 256 subdivs
Secondary = LC with 1000 subdivs

Does this seem about right? The elements look clean, but I am a little confused by the vraymesh lights I have running along the ceiling. In the RawLighting element they are showing as black. Is this normal?

It's likely the jpeg compression adding some artifacts but there's some noise there. In this scene you're getting a little strip of higher sampling around whatever vray is seeing as an edge / area of contrast so you're seeing it on the geometry edges, the crisp line of the sun beaming in the windows and lastly along the edge where your bright ceiling lights are meeting your less bright ceiling.. Looking at your sample rate and how it's sampling the edges on your geometry, then max 16 is possibly too much in this scene - there's nothing too complex just yet except for just around the lights which is a big jump in dynamic range / exposure values so that's the only place you're getting red pixels / using your max AA of 16. I'd be inclined to reduce your max AA to 12 or to 8 and then see if all of the edges are still nice and smooth. Note by doing this that the AA value is going to be dividing your mesh light and BF GI subdivs less, so they might now be overkill. 256 divided by your original 16 is 16 samples per pixel potentially, 256 divided by 8 means 32 samples potentially so if you go down to 8 max aa and your GI passes are still really clean, then you might be able to lower your BF GI from 256 and still get a nice clean result. Same with your raw light pass and your mesh light samples.

As regards the Raw lighting and the lights being black, the raw lighting pass only shows light and shadow that falls on to objects, since the light surfaces are emitting light / self illuminated then there's no possibility of light being able to fall on them. Add in your self illumination element and you should get the data that's in these holes.

[ARTECONI-CGI]

Hi John,

Thanks for the reply, I wil try again as soon as I can.
Thanks to Richard as well for posting a step by step explanation of the workflow. All clear now.

My last question is: does it make sense to use that method with irradiance map as primary bounces?
Do the Hsps subdivision have to be proportionally multiplied to meet the AA needs?

And the LC?

Thanks,

Giacomo.

[joconnell]

The only GI that'll get affected by this is brute force - the irradiance map and light cache don't care about your aa settings at all.

[Pixelcon]

Without going thru the entire thread again and at the risk of asking a stupid question, are we keeping the clr threshold and noise threshold linked like the old Universal method or unlinking them?

[Morne]

I like the way I'm getting super clean renders. But 8 hours for an 1600X900 super simple exterior is way excessive on a single i7 3930k with 32GB ram

[tricky]

It's likely the jpeg compression adding some artifacts but there's some noise there. In this scene you're getting a little strip of higher sampling around whatever vray is seeing as an edge / area of contrast so you're seeing it on the geometry edges, the crisp line of the sun beaming in the windows and lastly along the edge where your bright ceiling lights are meeting your less bright ceiling.. Looking at your sample rate and how it's sampling the edges on your geometry, then max 16 is possibly too much in this scene - there's nothing too complex just yet except for just around the lights which is a big jump in dynamic range / exposure values so that's the only place you're getting red pixels / using your max AA of 16. I'd be inclined to reduce your max AA to 12 or to 8 and then see if all of the edges are still nice and smooth. Note by doing this that the AA value is going to be dividing your mesh light and BF GI subdivs less, so they might now be overkill. 256 divided by your original 16 is 16 samples per pixel potentially, 256 divided by 8 means 32 samples potentially so if you go down to 8 max aa and your GI passes are still really clean, then you might be able to lower your BF GI from 256 and still get a nice clean result. Same with your raw light pass and your mesh light samples.

As regards the Raw lighting and the lights being black, the raw lighting pass only shows light and shadow that falls on to objects, since the light surfaces are emitting light / self illuminated then there's no possibility of light being able to fall on them. Add in your self illumination element and you should get the data that's in these holes.

Thanks for your responses.

I dropped the resolution to 800x800 for testing and changed the DMC image sampler. Results as follows:

Min=1 Max=8
Render time 8min32s
Lots of red

Min=1 Max=50
Render time 1min58s
No red

[tricky]

As regards the Raw lighting and the lights being black, the raw lighting pass only shows light and shadow that falls on to objects, since the light surfaces are emitting light / self illuminated then there's no possibility of light being able to fall on them. Add in your self illumination element and you should get the data that's in these holes.

That's what I thought, but if you look at my rawlighting render, there doesn't seem to be any bright light even coming out of those ceiling strip lights. However, when you look at the floor-level spot lights shooting up the wall on the right hand side, you can see all the light. I would have expected the rawlighting element to show black for the actual light source, but brightness all around it.

[Morne]

Thanks for your responses.

I dropped the resolution to 800x800 for testing and changed the DMC image sampler. Results as follows:

Min=1 Max=8
Render time 8min32s
Lots of red

Min=1 Max=50
Render time 1min58s
No red

Is this still with the grey override mat? And are you using VRaySun for this or an HDRI sky dome? Whats your subs for your lights at the moment?
Would like to see what you get once you switch off the override and see what subdivs you're using for the glossy mats.
Are you working with linear color mapping at 2.2 or reinhard?

[tricky]

Is this still with the grey override mat? And are you using VRaySun for this or an HDRI sky dome? Whats your subs for your lights at the moment?
Would like to see what you get once you switch off the override and see what subdivs you're using for the glossy mats.
Are you working with linear color mapping at 2.2 or reinhard?

Still with grey override. Yes, we are using vray sun/sky. Subdivs for lights as follows:-
spot ies uplighters = 32
ceiling strip = 256
sun = 64

Color (coloUr!) mapping:-
Linear
Dark = 1.0
Bright = 1.0
Gamma = 2.2
Sub-pixel off
Clamp output off
Affect BG on
Linear workflow off

This is all reasonably acceptible for greyscale, but as soon as I switch my textures back on the times go through the roof.

[Morne]

This is all reasonably acceptible for greyscale, but as soon as I switch my textures back on the times go through the roof.

Yes same here

[tricky]

Hopefully the thread will continue and shed some light upon this.

[joconnell]

Thanks for your responses.

I dropped the resolution to 800x800 for testing and changed the DMC image sampler. Results as follows:

Min=1 Max=8
Render time 8min32s
Lots of red

Min=1 Max=50
Render time 1min58s
No red

Is this still with the grey shader only?

[tricky]

Yes joconnell - still grey.

[Preisler]

wow! This is a great thread!

... but so close to Christmas one might wish for the special joconnell-script-that-starts-when-you-press-the-render-button-and-pops-up-with-suggestions-for-what-should-have-been-set-differently-for-faster-rendering!

[joconnell]

Yes joconnell - still grey.

So you're saying that 1/8 versus 1/50 on the grey only version is taking way more?

What I'd be thinking is this - If you've picked 1/8 with everything turned off and only a grey material - not even lights or GI, then that's probably the right amount that you need on your edges. If you turn on your lights and GI while keeping everything at 1/8 and you can get it clean by adjusting your light subdivs only, then this is probably still good. If at this point you find that changing your aa to 1/50 speeds things up, that's all well and good if you're delivering a shot with a grey material on everything so if you're happy with that then go for it. The problem is when you then turn on your materials, everything will go to crap. Stick with the 1/8 that you got at the very start since that's the right amount for your edges. When you go to the turn on your lights stage, you only care about getting your lights clean for that 1/8 value, then you go to your materials. You're optimising in predictable layers step by step from the AA up, don't worry about the middle grey stage. Red at the edges in the sample rate is good by the way - then it means you're using the right ranges of AA.