PDA

View Full Version : diffuse bounce problem



vonpietro
11-04-2018, 01:54 PM
Hi all,

SOOOOO... lw 2018 render times are through the roof - i'm still having a really hard time getting the render times under control.

Radiosity is the culprit - anything with radiosity seems to take 1 hour or more for me for interiors its awful.

The culprit seems to be diffuse bounces in the render tab. its default is set to 2.

FOr example i have a castle hallway - render time was 3 hours (depending on the location of the camera it ranges from 30 minutes to 3 hours) which is really weird as i can't pin point what makes the render time jump so high.

now if i just turn diffuse bounces to 1 instead of 2 - the render time went from 3 hours to a doable 9 minutes. 9 MINUTES. What the heck was it doing when it was set to 2 -
when i compared the two images the 3 hour and the 9 minutes - there was virtually no difference in the render -

I think it's a bug - Something is not right here - it should not be taking 3 hours

someone told me that turning diffuse bounces to 1 was essentially turning something off but i'm not sure what.

TO Newtek - what the heck is going on with diffuse bounces - a value of 2 is practically unusable as the render times consistently go into hours when at 1 its minutes.

a diffuse bounces value of 3 takes so long to render its like why even have it as a value you can adjust - Please look into diffuse bounces as something is very wrong with it

pixym
11-04-2018, 03:17 PM
Try Octane Render my friend…

Tobian
11-04-2018, 03:26 PM
Naturally you filed a bug report... Or you expect the engineers to debug your scene from imagination?

Norka
11-04-2018, 03:38 PM
Try Octane Render my friendů

Tru dat. No way could LW engine ever compete with Octane. See my many posts where I suggested (for quite some time) that LWG make Octane the official renderer of LW.

Ma3rk
11-04-2018, 05:20 PM
Hi all,

SOOOOO... lw 2018 render times are through the roof - i'm still having a really hard time getting the render times under control.

Radiosity is the culprit - anything with radiosity seems to take 1 hour or more for me for interiors its awful.

The culprit seems to be diffuse bounces in the render tab. its default is set to 2.

FOr example i have a castle hallway - render time was 3 hours (depending on the location of the camera it ranges from 30 minutes to 3 hours) which is really weird as i can't pin point what makes the render time jump so high.

now if i just turn diffuse bounces to 1 instead of 2 - the render time went from 3 hours to a doable 9 minutes. 9 MINUTES. What the heck was it doing when it was set to 2 -
when i compared the two images the 3 hour and the 9 minutes - there was virtually no difference in the render -


Would really need to see your scene, in particular your surfaces. If your coming from an FBX file, you might have discovered that the surfaces only come in as standard, Glossy Reflections ON, Spec & Gloss usually quite high and often with thier color space & smoothing option set wrong. Much of that really chews up rendering time w/o much return. Other than the infamous "Fireflies" usually.

Here are a couple recent renders that initially were in the 60-75 min. range, now running 15-20 mins.

143289

143290

These actually aren't using any radiosity nor benefits much if toggled on due to the way I lit the scene. That's actually part of my workflow. I light w/o GI but toggle it on to see where some add'l fill is needed. The difference is there but so subtle that the 1/3 rendering time is suddenly very attractive.

I've some exteriors where only GI is being used other than for props and those are in the 15-20 min range too, but the same surfacing notes apply.

I've also been resurfacing with textures generated with Bitma2Material & that might have made some surfaces more efficient but couldnt swear to that.

It's really sad that Craig Monins (RebelHill) has vanished & that so few managed to get his last series, RH_LSR. I did manage to get a copy (still a WIP), and watching his workflow & logic using the new surfacing tools had a lot of ah-HA moments for me anyway. For example, how to use the buffer views to determine which sample settings can get by with less & which need higher.

jboudreau
11-04-2018, 05:35 PM
Hi all,

SOOOOO... lw 2018 render times are through the roof - i'm still having a really hard time getting the render times under control.

What version of 2018 are you using?

Thanks,
Jason

jwiede
11-04-2018, 07:15 PM
It's really sad that Craig Monins (RebelHill) has vanished & that so few managed to get his last series, RH_LSR. I did manage to get a copy (still a WIP), and watching his workflow & logic using the new surfacing tools had a lot of ah-HA moments for me anyway. For example, how to use the buffer views to determine which sample settings can get by with less & which need higher.

Agreed. His series is quite excellent for explaining good tuning workflows for different types of LW2018 materials and scenes.

However, I do think there's a valid cause for concern regarding the "diffuse bounces" setting, as well. Even at very low settings raising it can cause drastic changes in rendering time, and while I could (perhaps) see anything w.r.t. specular/reflection having a large-magnitude effect, esp. will poorly-tuned materials, "diffuse bounces" having such a severe impact doesn't entirely make sense in that context. Diffuse path just shouldn't have anywhere near same potential for runaway exponential sample growth as in specular/reflective path, so why are the changes so extreme?

wolfiboy
11-05-2018, 01:07 AM
Excuse me of asking this question off the topic, but what do you mean by 'Craig has vanished'? Has he died? Has he left the Lightwave-community? I'm concerned!:oye:

adk
11-05-2018, 04:21 AM
Hopefully Craig is all well and just trekking through some remote part of the world.
I've emailed him a few times over the years and always found him to be fairly prompt, polite and a genuinely nice chap.
(Apologies for the slight derailment)

Dan Ritchie
11-05-2018, 08:57 AM
I was thinking the other day, when using global illumination, let it do the lighting work for you. Don't use any regular lights. This cuts down the complexity of shading a lot, and your radiosity is already bouncing light around the scene, so using illuminated surfaces instead of lights doesn't cost anything extra. Just a thought.

Ma3rk
11-05-2018, 06:03 PM
Excuse me of asking this question off the topic, but what do you mean by 'Craig has vanished'? Has he died? Has he left the Lightwave-community? I'm concerned!:oye:

Many of us are as well but unfortunately we simply don't know. That's sort of the "other edge of the sword" with the net. Don't even know who might know him personally, so unlesss someone who does comes forward, the rest of us are simply going to be in the dark.

I may have been one of the last to actually receive his LSR series and that was early April. Was hoping to bounce some questions off other users (as well as RH) in the forum, but that never happened. I've no idea how many were able to take advantage of his "Early Bird" special.

Ma3rk
11-05-2018, 06:09 PM
I was thinking the other day, when using global illumination, let it do the lighting work for you. Don't use any regular lights. This cuts down the complexity of shading a lot, and your radiosity is already bouncing light around the scene, so using illuminated surfaces instead of lights doesn't cost anything extra. Just a thought.

Largely depends on the image you're using but for the mostpart, but it is hard to improve on nature at the right time o' day.

jwiede
11-05-2018, 06:59 PM
I was thinking the other day, when using global illumination, let it do the lighting work for you. Don't use any regular lights. This cuts down the complexity of shading a lot, and your radiosity is already bouncing light around the scene, so using illuminated surfaces instead of lights doesn't cost anything extra. Just a thought.

Agreed, but not really salient to the OP's issue.

RPSchmidt
11-06-2018, 08:47 AM
Hi all,

SOOOOO... lw 2018 render times are through the roof - i'm still having a really hard time getting the render times under control.

Radiosity is the culprit - anything with radiosity seems to take 1 hour or more for me for interiors its awful.

The culprit seems to be diffuse bounces in the render tab. its default is set to 2.

FOr example i have a castle hallway - render time was 3 hours (depending on the location of the camera it ranges from 30 minutes to 3 hours) which is really weird as i can't pin point what makes the render time jump so high.

now if i just turn diffuse bounces to 1 instead of 2 - the render time went from 3 hours to a doable 9 minutes. 9 MINUTES. What the heck was it doing when it was set to 2 -
when i compared the two images the 3 hour and the 9 minutes - there was virtually no difference in the render -

I think it's a bug - Something is not right here - it should not be taking 3 hours

someone told me that turning diffuse bounces to 1 was essentially turning something off but i'm not sure what.

TO Newtek - what the heck is going on with diffuse bounces - a value of 2 is practically unusable as the render times consistently go into hours when at 1 its minutes.

a diffuse bounces value of 3 takes so long to render its like why even have it as a value you can adjust - Please look into diffuse bounces as something is very wrong with it

I gave these settings to a Lightwave user on Youtube who was making comparisons between LW2018 and Blender; they are pretty much my default scene settings and they usually provide very reasonable render times.

1. Turn off Camera > Adaptive Sampling and just set a sample (I usually use 8 to start).
2. Set Interpolated > Primary Rays to 175
3. Set Interpolated > Secondary Rays to 35
4. Set Maximum Pixel Spacing to 20.
5. Turn on "Use Bumps"
6. Turn on Enable Caching.
7. Set Diffuse Bounces, Relection Recursion, Refraction Recursion, Reflection Samples, and Refraction Samples to 5.
8. Enable Noise Filtering.
9. Set Render Tile Size to 8.

With these settings you are going to get noise.

So try adjusting your camera samples and number of rays evaluated to reduce the noise. Depending on the amount of specular you have in your scene, you may also need to increase your Reflection samples to clear hot spots.

In my experience, raising the camera samples with these settings doesn't cause a huge change in render times, unless you have a really outrageous sample setting.

Give them a try and see if you get a faster result that is usable. Then just tweak from there.

Good luck!

Dan Ritchie
11-06-2018, 09:54 AM
I was thinking this through in my head last night, and it sounded reasonable, so somebody check me on it.

Say you send out 2 rays, and that bounces off of something, and each of those rays split into 2 more rays, then each of those bound off of something, and they each split into 2 rays. In the end, you'd have 14 total rays instead of 2. etc.

jwiede
11-06-2018, 12:48 PM
I was thinking this through in my head last night, and it sounded reasonable, so somebody check me on it.

Say you send out 2 rays, and that bounces off of something, and each of those rays split into 2 more rays, then each of those bound off of something, and they each split into 2 rays. In the end, you'd have 14 total rays instead of 2. etc.

Correct, except that we're talking about the "diffuse bounces" limit, involving quite small default values and deltas, so the situation isn't very far "up the exponential growth curve". If we were talking about moderate values (8-16), the "exponential growth curve" argument would make a lot more sense. In many cases, though, the delta in render time between "diffuse bounce" values of 1 and 2, or 2 and 3, appears to be a LOT larger than easily explained, even with an exponential growth curve involved.