[Bf-cycles] max bounces

Brecht Van Lommel brechtvanlommel at pandora.be
Thu Nov 6 22:45:25 CET 2014


If you render with 1 for min bounces that explains the timings. If the
min bounces is smaller than the max bounces, that enables Russian
roulette path termination. This means most paths will be terminated
early and not use the max bounces, when Cycles detects that they
contribute little to the image.

For the same reason 200 bounces is not that slow, since only very few
paths will actually use that many bounces. Even if min bounces is set
to 200, the path will probably escape to the sky or hit enough dark
surfaces that the contribution becomes too small to matter.

On Thu, Nov 6, 2014 at 10:12 PM, Daniel Salazar - patazstudio.com
<zanqdo at gmail.com> wrote:
> I've also found strange how you can make cycles calculate 200 bounces
> with not much apparent slowdown, wonder why..
> Daniel Salazar
> patazstudio.com
>
>
> On Thu, Nov 6, 2014 at 2:42 PM, David Fenner <d4vidfenner at gmail.com> wrote:
>> Hi Thomas, thanks for the reply. I'll try to find time todar or tomorrow at
>> most to do a more clear test, with pics and blend file. I'm pretty sure I
>> tested it with 1 min bounce all the time though, but I'll run them again and
>> upload in a more organized way.
>>
>> Thanks again,
>> David.
>>
>> 2014-11-06 17:37 GMT-03:00 E. David Ferriman <dferriman at gmail.com>:
>>
>>> Awesome! Thanks!
>>>
>>> On 11/6/2014 3:33 PM, Thomas Dinges wrote:
>>>
>>> Hi David,
>>> that is weird, I cannot think of a reason for this at the moment.
>>>
>>> The way the feature works, is that it checks if the Light Max Bounce
>>> setting is greater than the current bounce, if that's the case it skips
>>> evaluating the lamp.
>>> Did you use the Pathtrace or Branched Pathtrace integrator? Also what were
>>> your global min bounces? If they differed from the global max bounces, it
>>> might be worth to check again with these 2 values being the same.
>>>
>>> Thomas
>>>
>>> Am 06.11.2014 um 14:00 schrieb David Fenner:
>>>
>>> Hi Thomas, thanks for the commit yesterday, to control bounces per light,
>>> very nice!!
>>>
>>> Been doing some tests here, and I ran into a very strange scenario:
>>>
>>> When I have 3 or more max diffuse bounces in global render settings,
>>> reducing per lamp bounces have little to no impact in render times, even if
>>> set to 0 bounces. However, when I have only 2 max diffuse bounces in render
>>> settings, lowering per lamp bounces have a very big impact on render times,
>>> reducing them almost as much as if I lowered global diffuse bounces on the
>>> same quantity. It's like there was some sort of threshold at 2 global
>>> bounces. For example:
>>>
>>> (on a simple diff scene with one lamp)
>>>
>>>
>>> Global Max Diff Bounces = 0
>>> Lamp bounces                 = 0
>>> Render Time: 4.7 sec
>>>
>>>
>>> Global Max Diff Bounces = 3
>>> Lamp bounces                 = 1024
>>> Render Time: 10 sec
>>>
>>>
>>> Global Max Diff Bounces = 3
>>> Lamp bounces                 = 0
>>> Render Time: 9.7 sec   (almost the same as before)
>>>
>>>
>>> Global Max Diff Bounces = 2
>>> Lamp bounces                 = 2
>>> Render Time: 8.5 sec      (lower rendertime thanks to one less bounce)
>>>
>>>
>>> Global Max Diff Bounces = 2
>>> Lamp bounces                 = 0
>>> Render Time: 4.9 sec       (before it didn't even care, yet for 2 bounces
>>> and below its almost as there where no global bounces)
>>>
>>>
>>> It's great, with 2 global diff bounces and below the gains of lowering per
>>> lamp bounces are almost as big as if lowering global diffuse bounces, so
>>> having only a few lights with GI can increase your rendertimes amazingly,
>>> however,  with 3 global diff bounces and over there is strangely almost no
>>> gain, even when all lamps have 0 bounces.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> Bf-cycles mailing list
>>> Bf-cycles at blender.org
>>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> Bf-cycles mailing list
>>> Bf-cycles at blender.org
>>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>>
>>>
>>> --
>>> E. David Ferriman
>>> Web Designer
>>> dferriman,com
>>> dferriman at gmail.com | (614) 260-7238
>>>
>>>
>>> _______________________________________________
>>> Bf-cycles mailing list
>>> Bf-cycles at blender.org
>>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>>
>>
>>
>> _______________________________________________
>> Bf-cycles mailing list
>> Bf-cycles at blender.org
>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>
> _______________________________________________
> Bf-cycles mailing list
> Bf-cycles at blender.org
> http://lists.blender.org/mailman/listinfo/bf-cycles


More information about the Bf-cycles mailing list