I have to say that I am not 100% sur anymore if max_path_depth relates to GPU use only. The images below all use a value of 99, just to be sure. In GPU mode max_path_depth is limited to 100. Not sure that is part of the documentation (could not find it in a quick search) , o if I red it in a forum post.
The scene has 2 light sources in it. One is the circus arena HDRI from Ployhaven.com.
The other is a very high intensity white laser like light source in the upper right hand corner, barely outside of the frame.
The background is a Velvet material from the online MatDB
The helix is the Lead Crystal (with Couchy-B and tabbed absorption) from the online material library.
Image number one was rendered 8 hours with GPU path tracing on an RTX 3070 Ti:
Nothing special, except the laser light source is completely missing from the image!
The following image was rendered for 10 minutes (not hours) BIDIR MLT on a Ryzen 5950X.
CoolColJ's test pics thread
- thesquirell
- Posts: 428
- Joined: Fri Aug 29, 2014 3:49 am
- Location: Novi Sad, Serbia
Re: CoolColJ's test pics thread
Awesome looking light effects! That's why I always loved BiDirMLT, it just has that WOW effect.
As for the limiting the depth to 99, are you trying to say that by capping the max_path_depth to 99 you got a clearer result in a much faster time with cpu? Like, you managed to cap the depth of the rays from infinite to 99? If so, that's awesome man! Sucks for the GPU, though.
I would love to test this scene out, or you could test it for us. Just render another image with the same settings for the BiDirMLT, minus the cap. Let it have infinite depth, and let it render for another 10 minutes. You should see much more noise, but some more reflection/refraction effects in the crystal, if the theory is correct (which is that max_path_depth affects both CPU and GPU algorithms).
As for the limiting the depth to 99, are you trying to say that by capping the max_path_depth to 99 you got a clearer result in a much faster time with cpu? Like, you managed to cap the depth of the rays from infinite to 99? If so, that's awesome man! Sucks for the GPU, though.
I would love to test this scene out, or you could test it for us. Just render another image with the same settings for the BiDirMLT, minus the cap. Let it have infinite depth, and let it render for another 10 minutes. You should see much more noise, but some more reflection/refraction effects in the crystal, if the theory is correct (which is that max_path_depth affects both CPU and GPU algorithms).
Re: CoolColJ's test pics thread
Simple BIDIR (without Metropolis Light Transport) also renders this effect.
That isn't surprising! Brute force path tracing almost never "finds" the very small laser light source.
The max path depth effects GPU only. When I open any of the test scenes that ship with Indigo with an XML editor, I can find this parameter names as "gpu_max_depth"
The BIDIR MLT or BIDIR images also take a long time to render to get reasonably noise free, but at least the caustic and dispersion effects are rendered. My final image rendered for 58 hours Not sure it really needed that long, but my computer really did not have anything else to do while I was at work
This shows that in some situations GPU path-tracing is simply the wrong choice.
I have more examples, albeit none this drastic, that highlight the importance of BIRID and BIDIR MLT. I'll make more posts if there is interest.
That isn't surprising! Brute force path tracing almost never "finds" the very small laser light source.
The max path depth effects GPU only. When I open any of the test scenes that ship with Indigo with an XML editor, I can find this parameter names as "gpu_max_depth"
The BIDIR MLT or BIDIR images also take a long time to render to get reasonably noise free, but at least the caustic and dispersion effects are rendered. My final image rendered for 58 hours Not sure it really needed that long, but my computer really did not have anything else to do while I was at work
This shows that in some situations GPU path-tracing is simply the wrong choice.
I have more examples, albeit none this drastic, that highlight the importance of BIRID and BIDIR MLT. I'll make more posts if there is interest.
Re: CoolColJ's test pics thread
del
Last edited by Zalevskiy on Sun Apr 10, 2022 5:29 pm, edited 2 times in total.
Re: CoolColJ's test pics thread
@Zalewsky
Can you elaborate on the intention of your post ?
I would believe that all or at least the vast majority of the models in this thread were created with mesh modelers. My models in this thread were created as Sub-D meshes with Blender.
Why would you say that the caustics in your images are fake?
Can you elaborate on the intention of your post ?
I would believe that all or at least the vast majority of the models in this thread were created with mesh modelers. My models in this thread were created as Sub-D meshes with Blender.
Why would you say that the caustics in your images are fake?
Who is online
Users browsing this forum: No registered users and 41 guests