AreaLight type efficiency

[Andrew_Wilkins1]

Hello everyone,

I have a question regarding area light "type" computation efficiency, does anyone by chance have some results from testing and/or a way to explain or rate the efficiency?

For example: "In general lighting scenarios in both interior and exterior environments when additional lights are added to the environment and/or characters to creatively push the lighting (kick, rim, fill, bounce etc.)—which of the above options is more/less efficient on the computations of the scene"

1. Default AreaLight uniformly scaled. (Should be the most efficient)
2. Default AreaLight scaled on a certain axis only (rectangle shape, long strip)
3. AreaLight with a map plugged in—such as a "SoftBox"
4. AreaLight with "Roundness" set to 1
5. AreaLight with "Soft Edge" set to 1
6. Default DiskLight (More or less efficient than AreaLight with roundness?)
7. Default CylinderLight
8. Above combinations (AreaLight with a map, scaled on one axis and values on the roundness/soft edge)

I know there is a lot of factors to take into consideration, I've generally heard that Disc lights for example are less efficient than standard Area Lights but I'd like to make educated decisions with our light setups.

Thanks very much!

[lee_griggs]

Light portals are recommended for interiors.

[Andrew_Wilkins1]

Hi Lee, yes thank you. 🙂 That is indeed our workflow as the base setup for interior lighting.

Maybe let me re-phrase the question/circumstance slightly.

"In general lighting scenarios in both interior and exterior environments when additional lights are added to the environment and/or characters to creatively push the lighting.—which of the above options is more/less efficient on the computations of the scene"

Going to edit above post.

[Andrew_Wilkins1]

Hey all, just posting a render test I did regarding AreaLight portals vs Skydome + Portal.

I see LightPortals are mentioned quite a bit on and I wanted to accurately test the efficiency compared to arealights. (I've used both setups on multiple productions)

The test scenario is a cube with a hole for a window.

Pretty default lights. Arealight and Lightportal in exactly the same place. There is a slight visual difference which to match the "look" would require moving the arealight back and adding a near light decay—which for testing purposes I didn't want to do.

R961| 00:00:01   480MB         | rendering image at 2580 x 1080, 4 AA samples
R962| 00:00:01   480MB         |   AA samples max     5
R963| 00:00:01   480MB         |   AA sample clamp   <disabled>
R964| 00:00:01   480MB         |   diffuse           samples  6 / depth  3
R965| 00:00:01   480MB         |   specular          samples  2 / depth  2
R966| 00:00:01   480MB         |   transmission      samples  1 / depth  2
R967| 00:00:01   480MB         |   volume indirect   <disabled by samples>
R968| 00:00:01   480MB         |   total                          depth 10
R969| 00:00:01   480MB         |   bssrdf            samples  1
R970| 00:00:01   480MB         |   transparency      depth 10

Skydome:
An average of 7 mins "pixel_rendering" across 10 frames.

Arealight:
An average of 9 mins "pixel_rendering" across 10 frames.

The result of the Skydome is noisy, whereas the Arealight is much cleaner.

In order to match the results for the Arealight with the Skydome—I need to increase the Diffuse samples by 4 putting us on 10 Diffuse samples.

Skydome Extra Diffuse Samples
An average of 18 mins "pixel_rendering" across 10 frames.

This means that the Skydome+LightPortals (Set to "Interior Only") is essentially double the render time when compared with AreaLights.

I ported this concept over to a production environment and the findings hold true, in some cases being more than double the render time to clean up the indirect noise from the Skydome.

Regards,
Andrew