Author Topic: Bright RGB (high ALBEDO) and better rendertime??? Why? (with MAX file)  (Read 4747 times)

2016-07-26, 15:01:06

cecofuli

  • Active Users
  • **
  • Posts: 1577
    • View Profile
    • www.francescolegrenzi.com
Hello,
as in the title, if I use bright RGB value, I have a better rendertime, for the same noise level.
And, the difference is HUGE!!!

PT+UHD (noise lever= 2%)

RGB= 128, 128, 128 ---> 17 minutes (515 passes)
RGB= 230, 230, 230 ---> 2 minutes (45 passes)



PT+PT (noise lever= 8%)

RGB= 128, 128, 128 ---> 6 minutes (125 passes)
RGB= 230, 230, 230 ---> 11 minutes (40 passes)

Isn't strange?


« Last Edit: 2016-07-27, 22:44:31 by cecofuli »

2016-07-26, 17:01:35
Reply #1

maru

  • Corona Team
  • Active Users
  • ****
  • Posts: 12780
  • Marcin
    • View Profile
My guess is (but it's jus a guess!) that it's a matter of contrast. Lower albedo = more contrast = adaptivity putting more samples to make it clear, more difference between bright and dark areas so the noise is generally stronger. Higher albedo = less contrast = more smooth/solid areas = easier to sample.
Marcin Miodek | chaos-corona.com
3D Support Team Lead - Corona | contact us

2016-07-26, 17:13:36
Reply #2

cecofuli

  • Active Users
  • **
  • Posts: 1577
    • View Profile
    • www.francescolegrenzi.com
Yes, I thought the same.
But, with PT+PT I haven't this problem,
So, I think, something doesn't work in the UHD....
It's a bug, IMO...
If you look at on the outdood balcony, with PT+UHD with RGB= 230, there are a lot of noise.
With low RGB, the noise is less. Obviously... 16 minutes vs 2 minutes! And better AA ( 500 passes vs 50! )


Anyway, without adaptivity, we aren't able to see the noise level.
But, I already do it:

PT+UHD (Renderign time 5 minutes)

RGB= 128, 128, 128 ---> (150 passes)
RGB= 230, 230, 230 ---> (105 passes)

« Last Edit: 2016-07-26, 17:17:45 by cecofuli »

2016-07-26, 17:36:29
Reply #3

Juraj

  • Active Users
  • **
  • Posts: 4765
    • View Profile
    • studio website
Yup, that's something I struggle with in 1.4 . Scenes that are seemingly similar, can be up to 10 times slower, for no reason identifiable to me.

Can you try to equalize the exposure though ? So it's different albedos, but same visual result +/- .

I turn adaptivity off for now, but as you say, impossible to see noise ratio that way,
Please follow my new Instagram for latest projects, tips&tricks, short video tutorials and free models
Behance  Probably best updated portfolio of my work
lysfaere.com Please check the new stuff!

2016-07-26, 17:37:35
Reply #4

cecofuli

  • Active Users
  • **
  • Posts: 1577
    • View Profile
    • www.francescolegrenzi.com

2016-07-26, 17:38:45
Reply #5

maru

  • Corona Team
  • Active Users
  • ****
  • Posts: 12780
  • Marcin
    • View Profile
Scenes that are seemingly similar, can be up to 10 times slower, for no reason identifiable to me.
Any chance it's bloom or glare set to more than 0? Or set to more than 0, and then to 0 again?
Marcin Miodek | chaos-corona.com
3D Support Team Lead - Corona | contact us

2016-07-26, 17:41:55
Reply #6

romullus

  • Global Moderator
  • Active Users
  • ****
  • Posts: 8868
  • Let's move this topic, shall we?
    • View Profile
    • My Models
Scenes that are seemingly similar, can be up to 10 times slower, for no reason identifiable to me.
Any chance it's bloom or glare set to more than 0? Or set to more than 0, and then to 0 again?

Correct me if i'm wrong, but B&G should have zero impact on rendering speed, as those are just post effects.
I'm not Corona Team member. Everything i say, is my personal opinion only.
My Models | My Videos | My Pictures

2016-07-26, 17:43:19
Reply #7

Juraj

  • Active Users
  • **
  • Posts: 4765
    • View Profile
    • studio website
Scenes that are seemingly similar, can be up to 10 times slower, for no reason identifiable to me.
Any chance it's bloom or glare set to more than 0? Or set to more than 0, and then to 0 again?

I don't use dailies (well I have them but go back to 1.4 for projects), I meant 1.4.

I sent you the scene though. In some other projects I run, some scenes clear fast, some super slow. The difference is far too big.
Please follow my new Instagram for latest projects, tips&tricks, short video tutorials and free models
Behance  Probably best updated portfolio of my work
lysfaere.com Please check the new stuff!

2016-07-26, 18:01:26
Reply #8

maru

  • Corona Team
  • Active Users
  • ****
  • Posts: 12780
  • Marcin
    • View Profile
Correct me if i'm wrong, but B&G should have zero impact on rendering speed, as those are just post effects.
It is, but it is also updated during rendering and it seems it is heavy. But it's acknowledged, and I am sure will be fixed.
</offtopic>
Marcin Miodek | chaos-corona.com
3D Support Team Lead - Corona | contact us

2016-08-22, 11:29:10
Reply #9

Ondra

  • Administrator
  • Active Users
  • *****
  • Posts: 9048
  • Turning coffee to features since 2009
    • View Profile
Ok, I think I know what is happening. First of all, PT+PT works as expected - higher albedo = slower rendering, no surprises here.

PT+UHD is different story. Maru is right - low albedo image has higher contrast, so it takes longer. But lets elaborate. Each rendered image is made by adding up DIRECT and INDIRECT components. In PT+UHD, the DIRECT pass is calculated with path tracing, and the INDIRECT pass is calculated with UHD cache. UHD cache is MUCH faster than path tracing and produces very low noise. But it cannot be used for direct pass because it could not handle its contrast.

First lets compare the result images:

Low albedo image is obviously much darker, but also computes longer. Why? Lets decompose the images to DIRECT and INDIRECT parts:

Direct:

Indirect:


Notice anything?
1) Direct passes have roughly the same brightness
2) Direct pass in high albedo image is MUCH noisier
3) Indirect pass with high albedo is MUCH brighter
4) Indirect passes have roughly the same noise

Explanation:
1) + 3) direct lighting is conceptually computed as emission*albedo, but indirect lighting is emission*albedo*albedo + emission*albedo*albedo*albedo + emission*albedo*albedo*albedo + .... - so the albedo plays much bigger role in indirect illumination calculation. This makes the result much brighter with high albedo, and also makes it compute longer when not using caching
2) This is because UHD cache does not help with direct light computation, and the high albedo image has fewer passes
4) This is because thanks to UHD cache it basically does not matter what the albedo or scene complexity is - it will always compute fast

Now why the high albedo image stopped calculating earlier? Because as I said earlier, the resulting image is DIRECT + INDIRECT. Indirect is always noise-free and cheap to compute thanks to UHD cache. Direct is noisy. So you add noisy DIRECT + clean INDIRECT and compute the whole image noise from it. In the high-albedo case, the INDIRECT pass is much brighter and influences the result more. Basically it masks the direct pass noise out from the image. In low-albedo case, the indirect pass is weak and the noise from direct pass can be seen over it clearly, so it needs to go away so the whole image looks noise-free.

Low albedo:
noisy direct + weak clean indirect = noisy result

High albedo:
noisy direct + strong clean indirect = clean result

Rendering is magic.How to get minidumps for crashed/frozen 3ds Max | Sorry for short replies, brief responses = more time to develop Corona ;)

2016-08-26, 14:17:45
Reply #10

cecofuli

  • Active Users
  • **
  • Posts: 1577
    • View Profile
    • www.francescolegrenzi.com