Author Topic: Corona 11 and DECALS / long render time  (Read 1728 times)

2023-12-20, 11:15:29

MUNT007

  • Active Users
  • **
  • Posts: 12
    • View Profile
Hi all

i don t know if it has been reported yet but the render time in a scene with decals is far longer in Corona11 compared to Corona10.

I tested in c4d R20 and in C4d 2024 and I have the same pb. If there s a lot of decals ( for example on a road for the signs) it becomes really really slow. If i desactivate the decals the render is very quick .

Note: i reported it in my post Corona11 bug

Thanks

2023-12-20, 11:21:51
Reply #1

Beanzvision

  • Corona Team
  • Active Users
  • ****
  • Posts: 3873
  • Bengamin
    • View Profile
    • Cormats
Hi there, can you share your results at all or perhaps a test scene?
Bengamin Jerrems l
Portfolio l Click me!

2023-12-20, 11:50:05
Reply #2

MUNT007

  • Active Users
  • **
  • Posts: 12
    • View Profile
Yes ok I prepare that

2023-12-20, 14:14:59
Reply #3

MUNT007

  • Active Users
  • **
  • Posts: 12
    • View Profile
Here s a file. you can compare the render time with and without decals.

2023-12-26, 18:30:23
Reply #4

BigAl3D

  • Active Users
  • **
  • Posts: 881
    • View Profile
I think you really need to collect the scene properly. This scene is missing 3 textures, no light and no camera. That would not be a fair comparison to what you are seeing. You might have settings in those areas that would be different than the devs.

2024-01-04, 14:12:18
Reply #5

MUNT007

  • Active Users
  • **
  • Posts: 12
    • View Profile
ok i can do that but the lighting doesn t change the pb. You can  put a corona sky+sun and remove missing textures . then test with corona 10 and compare with corona 11. Corona 11 is far longer.

Without the decals the render time is the same with corona 10 and corona 11

2024-01-04, 14:54:37
Reply #6

Beanzvision

  • Corona Team
  • Active Users
  • ****
  • Posts: 3873
  • Bengamin
    • View Profile
    • Cormats
As it is projecting the texture instead of uvmapped it is somewhat expected to be slightly slower. The amount of this I do not know so we will need to look into it further.
Bengamin Jerrems l
Portfolio l Click me!

2024-01-04, 14:59:59
Reply #7

TomG

  • Administrator
  • Active Users
  • *****
  • Posts: 5468
    • View Profile
Haven't tested in 10, so just a quick look in 11, and a lot going on in the scene. Some of the decals don't render for me (you can see what I get in the image attached).

That is a HUGE list of multipass render elements there, turning those all off left a render time of 24 seconds for 15 passes compared to 4 minutes. Are you sure it's not those that are causing the difference between 10 and 11 rather than the decals? If you turn those off, are 10 and 11 then the same even with decals on?

Also it has passes AND noise target set, meaning perhaps one end-condition is triggered differently with decals on/off/10/11. Usually best if "speed testing and comparing" to either used fixed passes, or fixed time, and to disable noise (it was definitely stopping on the noise condition for me in my first test).

Some of the geometry of the road looks a bit unhappy, as I am seeing some triangulation there in the rendering.

PS as suggested I just added a default sun sky and lowered the exposure.

Tom Grimes | chaos-corona.com
Product Manager | contact us

2024-01-04, 15:41:33
Reply #8

MUNT007

  • Active Users
  • **
  • Posts: 12
    • View Profile
Thanks Tom for your Ansewer, I LL make other tests without multipass to check. It s a template i use for complex scene that s why there s a lot of multipass and object ID

Otherwise i prefer not to lock the time in the settings since I have a render farm at office and all the computer don t have the same power.  put the noise that is acceptable for me but i also put a limit for the passes t o avoid too long rendering. Maybe it s not the best way to use the settings.

The scene was just an example ( and i had to reduce the size of the package ) I ll try on somthing more simple

2024-01-04, 16:14:21
Reply #9

TomG

  • Administrator
  • Active Users
  • *****
  • Posts: 5468
    • View Profile
Hi! Ah, pass plus noise combo is good for doing final renders and workflow, just not good for doing speed test comparisons - so I don't mean to sound like I am suggesting a change to your workflow there, just that when it comes to speed testing versions, it is best to swap to set passes or set time (but not both ;) ).

Understood on the render elements, again not an issue for workflow (though you may be spending longer rendering than needed in some cases if you don't use all of the template in a given scene), but also wondering if maybe it isn't decals alone that is the issue here. TY for any extra testing without them to see how it goes! Of course this might just mean the issue is "Corona 11 is slower than 10 when render element X is used" so there might still be an issue here, just aiming to isolate where it may be at :)

Cheers!
Tom Grimes | chaos-corona.com
Product Manager | contact us

2024-01-05, 10:47:47
Reply #10

Beanzvision

  • Corona Team
  • Active Users
  • ****
  • Posts: 3873
  • Bengamin
    • View Profile
    • Cormats
I also did some little tests of 20 passes and found V10 HF2 to be slightly quicker than V11. No doubt we will have to dive deeper to see exactly where this is coming from.
(Internal ID=1277067142)

« Last Edit: 2024-01-05, 10:57:29 by Beanzvision »
Bengamin Jerrems l
Portfolio l Click me!

2024-01-05, 11:29:52
Reply #11

MUNT007

  • Active Users
  • **
  • Posts: 12
    • View Profile
thanks a lot for your answers . I try to do some other tests today.


2024-01-05, 13:57:36
Reply #12

Beanzvision

  • Corona Team
  • Active Users
  • ****
  • Posts: 3873
  • Bengamin
    • View Profile
    • Cormats
Here's another basic test.

V10HF2: 3:27


V11: 3:15
Bengamin Jerrems l
Portfolio l Click me!

2024-01-31, 09:47:50
Reply #13

MUNT007

  • Active Users
  • **
  • Posts: 12
    • View Profile
Hi
I installed the corona 11 hotfix1 this morning and the pb seems to be solved. Render Time with my scenes with decals are faster

2024-01-31, 10:12:24
Reply #14

Beanzvision

  • Corona Team
  • Active Users
  • ****
  • Posts: 3873
  • Bengamin
    • View Profile
    • Cormats
Hi there, thanks for the update. We're glad to hear this :)
Bengamin Jerrems l
Portfolio l Click me!