Author Topic: New Light Solver ...  (Read 5731 times)

2017-10-26, 18:37:46

dj_buckley

  • Active Users
  • **
  • Posts: 876
    • View Profile
So the tooltip for the light solver states 'can speed up rendering by a factor of 10'

I just tested it.  Results after 5 passes

Solver Off - Estimated Time Remaining = 1.5 hours
Solver Off - 9.5% Noise

Solver On - Estimated Time Remaining = 2 hours
Solver On - 12% Noise

So are there certain scenarios where it can slow down your scene?

2017-10-27, 01:19:47
Reply #1

Tanakov

  • Active Users
  • **
  • Posts: 831
  • Corona is faster than diarrhea
    • View Profile
    • https://www.behance.net/Gringott
Yes I experienced slowdown at some scenes, didnt experience speedup yet.
Using Corona since 2014-01-02
https://www.behance.net/Gringott

2017-10-27, 01:31:38
Reply #2

cecofuli

  • Active Users
  • **
  • Posts: 1577
    • View Profile
    • www.francescolegrenzi.com
My experience is in a small improvement in complex scene (+100 Coronalights+Fog+GI+Glossy+Occluded objects)
This is one of our test


2017-10-27, 11:20:33
Reply #3

maru

  • Corona Team
  • Active Users
  • ****
  • Posts: 12752
  • Marcin
    • View Profile
Rendering just 5 passes is not a good idea. Please check with some more "real production" job. Like an interior with many lights, rendered for ~50 passes.

Cecofuli, please disable the fog and test again.
Marcin Miodek | chaos-corona.com
3D Support Team Lead - Corona | contact us

2017-10-27, 22:56:05
Reply #4

Bormax

  • Active Users
  • **
  • Posts: 568
    • View Profile
Hello

Did some tests with my working scenes and found out that for exterior scenes lit with corona SUN/SKY or HDRi new solver didn't do any difference in performance,
My test with interior shown that new light solver slow down rendering a lot. I rendered scene lit with HDRi and some artificial interior lights until the noise level 4%. Without new solver it took 21.5 min (1200x900), with activated new solver after 26 min of rendering noise level was 8.4% and estimated left time was 58 min, I didn't have time to wait so I just stopped it.

Win 10, Max 2014, Dual Xeon E5-2660

2017-10-27, 23:44:54
Reply #5

denisgo22

  • Active Users
  • **
  • Posts: 700
    • View Profile
I Agree with this/
Unfortunately, I have not yet discovered any significant difference /etc. no 0.5 sec into 2 hours): in time with solver or without solver, or generally in 1.7 version, in all my works, than was in 1.6:(( -very sorry--
I understand-- just need to buy a more powerful hardware ,--and nothing can be done about it :))
I think, for now, the main thing is not to get worse and longer time  than was in 1.6//
as they say in those TV program "Born in Fire"---"it can kill, but...":))
I remember with Vray was something different...
If written in 8-X times faster, so it was valid  for 90 percent of all cases, without many kinds of limitations :))
« Last Edit: 2017-10-28, 14:09:34 by denisgo22 »

2017-10-28, 09:41:30
Reply #6

maru

  • Corona Team
  • Active Users
  • ****
  • Posts: 12752
  • Marcin
    • View Profile
Did some tests with my working scenes and found out that for exterior scenes lit with corona SUN/SKY or HDRi new solver didn't do any difference in performance,
My test with interior shown that new light solver slow down rendering a lot. I rendered scene lit with HDRi and some artificial interior lights until the noise level 4%. Without new solver it took 21.5 min (1200x900), with activated new solver after 26 min of rendering noise level was 8.4% and estimated left time was 58 min, I didn't have time to wait so I just stopped it.

AFAIK if there is only one light in the scene (e.g. sunlight), then the new solver will not work at all. (or should not be enabled manually)
Check out the release blog post - https://corona-renderer.com/blog/corona-renderer-1-7-for-3ds-max-released/
Quote
The new algorithm can give benefits whenever there are at least two lights in the scene, and it has no parameters that you need to set. The biggest benefits will be seen when there are many lights, or where the lights have complex emission profiles, or in scenes with high occlusion (where a lot of the lighting is blocked by obstacles).

Quote
Win 10, Max 2014, Dual Xeon E5-2660
We have some reports about issues with Xeons and the new solver. Can you contact us about this at support@corona-renderer.com?
Marcin Miodek | chaos-corona.com
3D Support Team Lead - Corona | contact us

2017-10-28, 12:10:34
Reply #7

denisgo22

  • Active Users
  • **
  • Posts: 700
    • View Profile
Did some tests with my working scenes and found out that for exterior scenes lit with corona SUN/SKY or HDRi new solver didn't do any difference in performance,
My test with interior shown that new light solver slow down rendering a lot. I rendered scene lit with HDRi and some artificial interior lights until the noise level 4%. Without new solver it took 21.5 min (1200x900), with activated new solver after 26 min of rendering noise level was 8.4% and estimated left time was 58 min, I didn't have time to wait so I just stopped it.

AFAIK if there is only one light in the scene (e.g. sunlight), then the new solver will not work at all. (or should not be enabled manually)
Check out the release blog post - https://corona-renderer.com/blog/corona-renderer-1-7-for-3ds-max-released/
Quote
The new algorithm can give benefits whenever there are at least two lights in the scene, and it has no parameters that you need to set. The biggest benefits will be seen when there are many lights, or where the lights have complex emission profiles, or in scenes with high occlusion (where a lot of the lighting is blocked by obstacles).

Quote
Win 10, Max 2014, Dual Xeon E5-2660
We have some reports about issues with Xeons and the new solver. Can you contact us about this at support@corona-renderer.com?

from all the above, if we take into account that in most scenes, try not to use more than 1 or 2 light sources,/ include environment background /sky,HDR//,to achieve a more realistic
scene lighting and mood, especially in interiors scenes, ----let alone the exterior's:)
---so we as not to feel any difference in the render speed:)))
« Last Edit: 2017-10-28, 12:19:16 by denisgo22 »

2017-10-29, 21:13:19
Reply #8

dj_buckley

  • Active Users
  • **
  • Posts: 876
    • View Profile
Rendering just 5 passes is not a good idea. Please check with some more "real production" job. Like an interior with many lights, rendered for ~50 passes.

It was a 'real production' job.  And it was effectively the same after 50 passes - and it was an interior with no environment light, and about 20 artificial lights.

2017-10-30, 21:42:59
Reply #9

cecofuli

  • Active Users
  • **
  • Posts: 1577
    • View Profile
    • www.francescolegrenzi.com
Maru, I used fog because without it, the difference in rendering time was less ;-)