Page 4 of 24
Influence of light type on despersive Glass 4
Posted: Sun Oct 21, 2018 6:07 pm
by FarbigeWelt
- CPU BiDir, Metropolis 10,8192,2, depth 15, plane Light 0.1 90 deg., 2600 Samples, 540x540, 1h45
Note, planes emit light in the direction its face normal points.
Again cube light, this time with special emission settings: e.g. face up 25W prio 4, faces around 100W prio 1, face down 500W prio 4 (three materials).
- CPU BiDir, Metropolis 10,8192,2, depth 15, cube Light 0.1 90 deg, 4 500x1, 1 100x4, 4 25x1, 511 Samples, 540x540, 13m
There is a shadow on the ceiling that looks like as thrown by the cube light, although its opacity is 0.0.
- CPU BiDir, Metropolis 10,8192,2, depth 15, cube Light 0.1 90 deg, 1 100x1, 5 25x5., 292 Samples, 540x540, 7m
- CPU BiDir, Metropolis 10,8192,2, depth 15, cube Light 0.1 90 deg, 5 100x1, 1 0.25.x5, 360 Samples, 540x540, 10m
There is one post left, GPU path dispersive glass (maybe with caustics).
and now, something completely different
Posted: Sun Oct 21, 2018 6:12 pm
by FarbigeWelt
Dispersive Pattern, fancy denoised
Posted: Sun Oct 21, 2018 10:08 pm
by FarbigeWelt
GPU Path is a different world.
Whereas CPU BiDir requires objects emitting light, GPU Path wants area lights to show caustics. Finally I thought to have a half way solution setup up for brilliant diamonds. I liked to remove noise and got this
:
- Diamond Briliant, GPU Path Sobol, area light, 5135 S, 5m, denoised 0.75
This Image is a mix: 0.75 render, 0.25 denoiser (with 0.6,6,6,1).
Re: and now, something completely different
Posted: Mon Oct 22, 2018 10:41 am
by epilectrolytics
FarbigeWelt wrote: ↑Sun Oct 21, 2018 6:12 pm
stormy-unexpected
OMG a lashing tree!
You have a swiss/german blender localization ("Es wird Frühling und die Bäume schlagen aus") with wrong time setting??
Re: and now, something completely different
Posted: Mon Oct 22, 2018 9:10 pm
by FarbigeWelt
epilectrolytics wrote: ↑Mon Oct 22, 2018 10:41 am
FarbigeWelt wrote: ↑Sun Oct 21, 2018 6:12 pm
stormy-unexpected
OMG a lashing tree!
You have a swiss/german blender localization ("Es wird Frühling und die Bäume schlagen aus") with wrong time setting??
Very well observed!
("Be aware of budding trees down there!" Or, "Watch your step, it's spring!")
Leaves-Fall.gif
Close but yet not there, 'cause leaves behave strangely due to their connection to source animation.
And yes, sapling tree gen again it is not as parametrizable as Modular Trees, but hey, a lot more predictable and is simple to animate leaves and branches.
There would be a long and stony path ahead of me to animation based on inverse kinematics and skeletons with hundreds of bones. I have no idea how Blender natives manage armatures in 2.79b efficiently.
GPU Path, Dispersive Lenses
Posted: Mon Oct 22, 2018 10:33 pm
by FarbigeWelt
- Lenses GPU Path, area light, Metropolis, 10000 S, 2160x2160, 1h27m
After 500 samples focused light spot is good visible. Render time is quite long because of depth specular =24.
Lets see later how many samples per MS/s are possible with some improvement.
Solved: Asymetric CPU BiDir slows with every sample
Posted: Tue Oct 23, 2018 6:12 am
by FarbigeWelt
With the lense scene CPU BiDir depth 12,3 starts with a few hundreds kS/s decreasing with every Sample, after 7 Samples the speed is around 50 kS/s. May it is the asymetric depth setting. Focused spots require 3 light depth only, lense effects take at least 8 eye depth. Lens effects show up very slowly below an area light.
EDIT: Solution here is: Check Metropolis settings.
Metropolis seems to act depending on render engine. After increasing Image Mutation Rate from 0.1% to 5.0% render speed got stable after a short start peak at 245 kS/s.
Image follows:
- Lenses, CPU BiDir Metropolis, 48S, 8m, 1620
Wish List: Since I use luxcorerender quiten often, I wished there were a combi-render engine as path and bidir both have their weaks and strengths regarding speed, caustics, speculars, color depth and settings.
Conclusion: Path and BiDir are different and require their own appropriate setting.
Re: Solved: Asymetric CPU BiDir slows with every sample
Posted: Tue Oct 23, 2018 6:40 pm
by mikegold10
FarbigeWelt wrote: ↑Tue Oct 23, 2018 6:12 am
With the lense scene CPU BiDir depth 12,3 starts with a few hundreds kS/s decreasing with every Sample, after 7 Samples the speed is around 50 kS/s. May it is the asymetric depth setting. Focused spots require 3 light depth only, lense effects take at least 8 eye depth. Lens effects show up very slowly below an area light.
EDIT: Solution here is: Check Metropolis settings.
Metropolis seems to act depending on render engine. After increasing Image Mutation Rate from 0.1% to 5.0% render speed got stable after a short start peak at 245 kS/s.
Image follows:
Lenses, CPU BiDir Metropolis, 48S, 8m, 1620.png
Wish List: Since I use luxcorerender quiten often, I wished there were a combi-render engine as path and bidir both have their weaks and strengths regarding speed, caustics, speculars, color depth and settings.
Conclusion: Path and BiDir are different and require their own appropriate setting.
I wonder how this would look with greater depth numbers (20-30)?
Re: Solved: Asymetric CPU BiDir slows with every sample
Posted: Tue Oct 23, 2018 6:45 pm
by Dade
FarbigeWelt wrote: ↑Tue Oct 23, 2018 6:12 am
With the lense scene CPU BiDir depth 12,3
Is "12, 3" eye and light path length ? For rendering caustics, you should do exactly the opposite (and probably use larger values in general because of the high amount of internal reflections).
Lenses CPU BiDir and GPU Path comparison
Posted: Tue Oct 23, 2018 8:04 pm
by FarbigeWelt
Lenses, Metropolis, CPU BiDir 1000 S, depth e24,l5 11h27 and GPU Path 10000, depth s24 1h27, comparison and mix.jpg