Author Topic: Physical Lights vs. Render Times  (Read 2722 times)

0 Members and 1 Guest are viewing this topic.

Offline cjwidd

  • Sr. Member
  • ****
  • Posts: 269
    • ArtStation
Physical Lights vs. Render Times
« on: October 03, 2017, 04:15:36 pm »
Scene I'm working with is lit completely by physical lights: 8 area lights (!) and an IES profile and contains a number of transmissive materials. For this scene, the following settings produce the cleanest (denoised) image:

Product mode
- Shadow quality = 4
- Ray bounces = 16
- GI bounces = 1
- Caustics = N/A, produces the cleanest images (less noise), but we are talking 7 hours for ~375 samples.

What I'd like to know is what is the relationship between render times and physical lights? Is it linear, exponential, or what? I assumed the render times would increase by some factor as a result of using physical lights, but the render times I'm experiencing are even beyond my less conservative estimates.

Win 10
i7-7700 CPU, 4.2 GHz (8 CPUs)

Offline MWo

  • Hero Member
  • *****
  • Posts: 836
  • Senior Technical Consultant @ INNEO (Germany)
    • german KeyShot page from INNEO Solutions
Re: Physical Lights vs. Render Times
« Reply #1 on: October 03, 2017, 09:33:59 pm »
Have you ever tested product mode vs. interior?
How fast is this scene without area light?

This would speed up the render time.

Offline cjwidd

  • Sr. Member
  • ****
  • Posts: 269
    • ArtStation
Re: Physical Lights vs. Render Times
« Reply #2 on: October 03, 2017, 09:38:28 pm »
Well, there are several area lights, but I did do test renders using basically every permutation of render option, including the new experimental mode. I found that product mode and interior mode converged on similar results with GI bounce = 1; additional bounces produce too much noise, in this case, for all render options.

EDIT: After doing some more region render tests, I'm fairly certain that about 90% of the issues I've faced in Keyshot, so far, are ultimately the result of bad topology. I'm just going to retopo the mesh and move on. If there are still issues after that, I'll address them later.
« Last Edit: October 03, 2017, 10:34:34 pm by cjwidd »

Online mattjgerard

  • Hero Member
  • *****
  • Posts: 1474
Re: Physical Lights vs. Render Times
« Reply #3 on: October 04, 2017, 06:43:01 am »
I do a lot of work with HDRI lighting, then having physical lights behind cloudy plastics (LED indicators and buttons and such) and my renders times skyrocket.

One way to figure out what is causing slowdowns is to hit H and have the HUD up so I can see the realtime FPS, and start turning things off in the scene tree. I can usually find that one to three pieces of metal or some weird setting in a matierals attached to an object that drops the FPS by more than its worth. I typically have a  lot of metal brackets, screw heads, connectors and stuff in my scenes, but for some of those I can swap out the materials for a diffuse color and not have the computer worry about and calculate all the reflections and stuff for a small insignificant part that is in the shadows. That can help, but that troubleshooting can take a lot of time too.

Offline cjwidd

  • Sr. Member
  • ****
  • Posts: 269
    • ArtStation
Re: Physical Lights vs. Render Times
« Reply #4 on: October 04, 2017, 11:47:03 am »
I'm glad you said this; at least I'm not alone. I've been taking full advantage of the render region option and the HUD when using Keyshot. I spent some time remodeling the part of the object and brought it into Keyshot last night. So far, doing so has reduced or eliminated most (if not all) of the artifacts I was experiencing with that part of the mesh.

I was experimenting with a different workflow before and really abusing Keyshot, so it's not altogether a surprise that I was getting artifacts and so on; super heavy mesh, horrible topology, no mesh clean-up; using this mesh in conjunction with transparent, transmissive materials - it was really a recipe for disaster from the beginning.

Online mattjgerard

  • Hero Member
  • *****
  • Posts: 1474
Re: Physical Lights vs. Render Times
« Reply #5 on: October 04, 2017, 12:39:22 pm »
Same situation here. All my stuff is CAD conversions, and those ain't pretty, no matter how they are converted or imported. Render region and the HUD are my best buds.

Offline feher

  • Hero Member
  • *****
  • Posts: 1554
  • www.feherfactor.com
    • Feher Factor
Re: Physical Lights vs. Render Times
« Reply #6 on: October 16, 2017, 03:19:42 pm »
9 times out of 10 it's the material that is created. Bumps too high, crazy samples...etc Next will be not enough light or the contrast is off on the HDRI.
If any of you have a scene that is giving you fits. I would be more then happy to go through it and find or look around and see what you can do to get better/ quicker results. Just let me know.
My door is always open.
Tim

Offline cjwidd

  • Sr. Member
  • ****
  • Posts: 269
    • ArtStation
Re: Physical Lights vs. Render Times
« Reply #7 on: October 16, 2017, 07:44:17 pm »
^Thanks for the offer, perhaps in the future. I ended up remodeling the asset which made a huge difference; in fact, with GI bounce = 1 (no caustics), the final render is usually artifact free. A major issue was that I was trying to cause light to shine through transparent sections of a closed surface (from the inside), without first modeling a shell; when you're dealing with a static renderer, geometry really needs to be modeled true to form, there is much less room for hacky shortcuts. I also had at least a couple materials which were layered incorrectly.

Addressing these issues improved the render times and eliminated artifacts, but once you start adding physical lights to the scene, regardless, render times will increase.