We finally made CentiLeo 0.48 for Cinema 4D R16-R18! It took so long time, so many efforts. Sorry for very long waiting! We thought were close to release for the last 2.5 months but technical development issues emerged and forced us to do more complex solutions to get right result. And that cycle happened several times. This 0.48 version development was harder than integrating with C4D or 3ds Max .
CentiLeo 0.48 release is mostly about multi-GPU and other efficiency improvements for rendering and memory management. Graphics feature improvements (more shaders, AOVs, improved shadow catcher, motion blur and mesh lights) are expected in the nearest alpha releases which will come out few times per month.
Download here:
Download shaderball:
User manual from 0.44 (yes it needs update):
Change Log: cntlc4d 0.480 alpha (2017 May 2)
[core] Multi-GPU support (maximum 16) with adaptive render data and cache arrangement on GPUs with different memory size. It scales 1.9-2.75x for 2-3x GPUs
[core] Up to 50% improved GI sampling for scenes with harder lighting. Sampling becomes smarter at the cost of more complex computations. For scenes with more or less interesting lighting (interiors, exteriors with several CG lights and/or complex HDRs) it gives up to 50% faster rendering. But for scenes with simple lighting (e.g. outdoor with single area light or simple HDR) where new smart sampling is too smart the rendering is 15-25% slower due to more complex computations compared CentiLeo 0.47.
[core] Solved render initialization problem that decreased render speed in CentiLeo 0.47 by 2x until user manually plays with cache sizes. This solution converts to 2x faster render for production mode and for those who didn’t bother with cache size manipulation.
[core] 2.5x faster scene parsing for complex geometry (when render data is initialized and recompiled in IPR mode).
[core] A lot more fluent IPR experience when camera or post-process settings or other scene data are changed. 3D app viewport doesn’t lag as before. IPR responses real-time for simple data changes (like camera motion, post-effects, object motion or material changes) while user continues moving the UI slider params.
[core] Around 40% less memory needed for geometry storage. Now we pack 25M unique non-instanced triangles (with 1 UV channel, normal and tangent vectors) in 1GB. Wider range of highpoly scenes become less out-of-cory.
[core] 30% faster rendering for scenes with out-of-core geometry (and this accel adds on top of mentioned geometry memory savings).
[core] Up to 4 UV channels supported for each geometry mesh (selectable in mapping slots of shaders).
[core] Material editor preview: much faster and nicer preview render. [core] Each GPU is initialized with as much memory as needed by scene (unlike taking all free memory even for trivial scenes in CentiLeo 0.47).
[core] Render stop criterion is now driven by min/max number of image iterations and also noise level. If the pixel gets at least min iterations and the noise level matches the user defined noise level then the render will not revisit this pixel anymore. Noise detection becomes more conservative and robust now.
[core] Added roughness parameter for Diffuse channel of material (it also hides shadow terminator artifacts on low-poly objects if any).
[C4D plugin] Improved render time in native Picture Viewer mode by updating secondary AOV output less frequently.
[C4D plugin] Slight reorganization of Material editor GUI (all SSS are placed in one tab, Translucency is also placed in a separate tab).
Change Log: cntlc4d 0.481 alpha (2017 May 3)
[core] Fixed bug when one of several GPUs may randomly avoid participating in collective rendering
Change Log: cntlc4d 0.482 alpha (2017 May 4)
[core] Fixed bug when changing any Mix parameter of material (or any parameter which has no effect ) may cause render degradation and crash.
[core] Fixed bug when switching between materials (and/or removing them) may cause render stop.
[core] Fixed bug with artifacts related to wrong smooth normals of mesh.
[core] Fixed bug with artifacts on anisotropic reflections.
[core] Fixed bug when material tangent space doesn’t really change when user enables it.
[C4D plugin] Added Sun item to main CentiLeo menu which adds sun with proper tags understandable by CentiLeo.
Change Log: cntlc4d 0.483 alpha (2017 May 10)
[core] Fixed bug with missing polygons on small objects like plane.
[C4D plugin] Fixed bug with render stop upon deletion of material tag.
[C4D plugin] Fixed bug with IPR random wrong parsing of objects and render freezes.
Known issues to be solved
- If you have several GPUs in computer it’s not recommended to assign the display-GPU for rendering. Sometimes is causes strange behavior. Currently we are trying to fix it.
- Displacement mapping feature is deactivated due to large code changes overall. This feature will be recovered within the next few short CentiLeo updates.
- Sometimes GUI of IPR window may have old style controls.
- May not initialize if AMD GPU is in your computer together with NVIDIA GPU.
- Very limited scene converter tool from Cinema 4D native materials (only diffuse color and texture are converted and translated if assigned to the object)
- May not support non-latin texture paths.
Enjoy rendering with CentiLeo again! And please shader experience related to multi-GPU behavior.
Kirgman
Edited: - Apr 26, 2020 05:35(Added info on update 0.483 (May 10, 2017))
It's been sometime! How is the multicore going? Please give us an update on current status and keep up the good work!
Hi Konstantinos! Practically done, I hope to post it within few days . Very few small issues/bugs must be solved before that. Need to collect memory garbage when we switch on/off multiple times the threads for multi-GPU, also need to make Picture Viewer working (it was broken), and need to tune some params.
CentiLeo Chat:
CentiLeo for Cinema 4D 0.471 available (quality & materials improvement)
Hi Rodrigo! I hope to make releasable news soon Technically things are going well, but we had to rearrange things in preview of material editor, out-of-core texture engine and IPR. Currently finishishing some IPR bugfixes. We realised that it was needed to diverge some rendering resources (ME and IPR) and make them more compact just for the purpose of nice scalability in the presence of multi-GPU. That's why it took longer than promissed. But I hope very much to release it ASAP within a week, but let's keep fingers crossed because previous promises have failed Maybe to make release sooner we will postpone some planned things like GGX (which don't influence on multi-GPU) to 0.49.
The window look is really cool!
The other thing I like is dark cover on top of the bed. So, does it also use GGX emulation?
I think 32 min is ok for laptop GPU for this scene. Resolution 1600x860?
As for displacement:
Why using it as modifier is more convenient than through maps? Due to speed of tesselation?
In 0.48 displacement is broken with new tech and it seems to get quite a lot of time to fix it. Fixing it will be postponed to one of the next few versions. But the good thing is that there is very good chance to accelerate the tesselation speed by a lot! It should happen in seconds.
Hi Ali, thanks for sharing! The picture has many nice things, but I would work more on the walls. Is it bump or normal mapping? Maybe displacement geometry?
There is also some Photoshop here. I think that CentiLeo needs bloom and glare in one of the future versions.
The neareset CentiLeo 0.48 will get more tonemapping (post-pro) controls.
How have you made the windows? What material properties? That rain water on the windows - is it done with the texture of roughness or IOR for glass?
#1 We had the manual trick with cache manipulation which allows to increase the render speed. Now this works automatically. This is needed for production mode in 3ds Max and for Picture Viewer in C4D.
#2 Now the renderer consumes the GPU memory not larger than available scene geometry needs + some rays cache (up to 1GB) + texture cache (up to 768MB). This fixes the problem reported by Aimar . If there is more geometry than GPU memory then out-of-core comes in but now it is more efficient due to improved data transfer and data storage.
#3 In Multi-GPU system all the available memory of single GPU will be utilized. For example if you make the render of scene using one 4GB and one 8GB GPU then the first GPU will utilize up to 4GB of memory and the second GPU will utilize up to 8GB of memory. And the first GPU in the same render may use out-of-core operations while the second may use no OOC operations because of larger capacity.
#4 However there are still some details related to limiations that we need for robustness: if some GPU is very small like 1.5-2GB then it can accept up to 30M polygons and 100K mesh instances and if GPU has more than 4GB then it may accept up to 400M polygons and 1M mesh instances. All GPUs larger than 4GB have same limits.
If you combine 2 or more GPUs with different limitations in one scene render then the smallest limitation will apply for all other GPUs. The limits are quite high and will be higher one day.
But if all GPUs in the system have same limits then #3 will effectively work.
Still working on the other things, now on the part related to texture transfers to multiple GPUs and there is some fight between few features e.g.:
1) efficient texture cache utilization inside the bucket (needed for better speed increase due to more GPU)
vs
2) image-space noise-driven adaptivity at low sample rates
Hello everybody!
Previously, I was too optimistic on release date. We worked very hard but still have few tasks to make 0.48 update finished.
Development is advancing very well, local results are very good and the speed and interactivity of the renderer are improved noticeably. Hope to publish CentiLeo 0.48 soon, I think it's a matter of week or so.
CentiLeo Chat:
CentiLeo for Cinema 4D 0.471 available (quality & materials improvement)
There is no yet implemented "emission channel"... Will try to implement it in one of the near future versions.
However, one small trick is possible if you apply cntlConst shader with large value to diffuse channel. This way it will reflect a lot more light.
In bump channel it is possible to combine black&white bumps and normal maps.
However there is some small drawback is that when you apply colorCorrect on top of bump shaders then you need to extend the range of "clampA and clampB" values to negatives. Our bump channel also needs a shader for procedural flakes.
CentiLeo Chat:
CentiLeo for Cinema 4D 0.471 available (quality & materials improvement)
We are currently working on the multi-GPU and few other improvements for 0.48 such as better IPR, better look and better out-of-core.
This should be more interesting. However, many more things are planned for alpha version.
Please, don't hesitate to write what you want to see in the renderer! We listent to it. However right now the amount of accumulated requests is so large that we can't do them all immediatelly. But step by step these requested features will be performed. And of course we need to make the integration with C4D better, tighter and more native.
For example, I am desperately can't wait when will work on hair&fur and particles Because it is hard and very interesting. The other things are much easier.
Some small internal progress, we have improved geometry data storage by around 30%. Now more polygons will fit into geometry cache and it is also good for "out-of-core" case.
Still working on the rest of the TODO list.
CentiLeo Chat:
CentiLeo for Cinema 4D 0.471 available (quality & materials improvement)
Konstantinos Despotis wrote:
Hello!,
Any updates on the next version with multi GPU support? Can't wait to try it.. No rush or anything, just an update on where its at.. I've read also about the hack attacks, hope they didn't delay much of the progress.. keep up the good work, and keep us updated!
Konstantinos
Hi Konstantinos,
That attacks don't delay anything. This is just an unpleasant thing that my friend is working to fix.
The work on multi-GPU is going well. It takes some time because we have quite large pipeline of parallel sub-programs inside the renderer.
There are also a few things that should be done in a renderer that 1) should fix not-loaded procedural objects for Picture Viewer, 2) and would like to hide and automate that 2x manual renderer acceleration trick (see ) which is important for final renders and animation.
0.48 will be published when everything will be clarified with it. I think it may take at least 2 more weeks or something like that.
Maybe you have read that we made some GI acceleration (see )
And we also made support for up to 4 UV sets per mesh (in Cinema terms up to 4 UVW tags) that can be selected in any texture. After 0.48 will continue in this area to support more procedural texture mappings and transformations as TextureTag offers.
The goal is to make 0.48 closer to "production state" and then to continue implementing features requested by users. The path is very clear.
Currently we don't yet support native C4D shaders but have own set of shaders. However, we have IPR mode which understands everything (i.e. it responds to material, shader, lights, object motion and polygon editing inside object). This IPR is final frame quality and has button to transfer image or beauty passes to Picture Viewer. Since upcoming CentiLeo 0.48 the responsiveness of IPR will be further improved.
We plan to implement support for native C4D shaders as much as possible. But right now it is necessary to make more core features like motion blur, mesh emitters, hair, massive particles.
Gael LE GOLVAN wrote:
Congratulations...we can see very nice details in dark areas and the colors are nicely rendered.
Very nice work !
Thanks Gael! Yes, dark regions are rendered better now. This is also important for further post work based on outputted image layers.
Also it is possible to enable one "fake" feature that helps rendering darks even better. The light sources have the checkbox "contribute to GI only". If checked then the light source contributes to GI only (i.e. to the secondary diffuse bounce) as you can see in the recent post by Ali
However, pictures in the current thread have no enabled "contrib to GI only".
CentiLeo Chat:
CentiLeo for Cinema 4D 0.471 available (quality & materials improvement)
I aslo thought about it. Will probably swich off checkboxes of the layers "Reflection 2", "Transmission", "SSS" but make their respective Weighs = 1. When checkbox is enabled the weight is already there.
At some point I will make this and other toggles of IPR to be saved with the scene file.
Quote
Rodrigo Bitencourt Rodrigues wrote:
We need a bump intensity slider to positive and negative values(like cinema 4d standard material.
Ok, will extend the range of the main bump strength slider. However now it is possible to do it inside the bump texture in cntlBitmap when you play with out_rangeA and out_rangeB. E.g. when you make out_rangeA=1 and out_rangeB=0 the bump goes inverted. Similar effect happens when out_rangeA=0 and out_rangeB=-1. It's also possible to make it negative with cntlMath shader with mutliplication by -1
It has it actually. When you change the value of out_rangeB from 1 to 10 then displacement become higher 10x. When you make out_rangeB=-1 it goes under the surface and so on. When you change out_rangeA to the value other than 0 you can change the "sea level". It seems we need to make cntlDisplacement shader with respective naming convention.
I hope to add support for more UV tag mesh projections next version. Right now I think about making 3 UV tag mappings per mesh selectable in material shaders. Max 3 mappings is usually seen in other engines. Maybe we will be able to do more than 3 UV tag mappings. How many are needed?
Also need to add support of TextureTag params too.
We plan to do it in a couple of versions. There are many possible ways to do it. Need to figure out the most convenient one and do it. What good example reference would you recommend?
Maybe yes. In one of the few upcoming CentiLeo updates there should be version with displacement improvements (there are a lot do there). There is also a chance to move displacement settings to the Tag of the object (and Opacity which is found in Multi Material will also probably go to the Tag). There is aslo a chance the displacement will understand shader networks but don't promise yet.
Quote
Rodrigo Bitencourt Rodrigues wrote:
We need a alpha channel to our material.(not like the multi material, we need one to hide some parts of the model.)
Ok, will make separate Alpha channel (limited with "specular ray depth" param of the render settings). Right now such things are implementable using Transmission channel with IOR=1 (but it will be limited to maximum ray depth in render settings) and it can be done with Opacity slot (see in cntlMultiMaterial) which has unlumited depth (but limited to only cntlBitmap input and discrete nature: the surface is either visible 100% or not).
Hope to see comments from other guys on the subject too, especially on the number of required selectable UV tags in material textures.
Sure Ali, I also have noticed that after roughness = 0.6 or maybe 0.7 the values are useless also because of some energy loss.
One option for us is to re-scale the input roughness by 0.6. But the final decision will be after adding GGX, because the highlights should look different with it.
Need to post how things are going since there are now news from us for 3 weeks already
We are working on many improvements including further noise reduction, multi-GPU, nicer material reflection look and nicer tone mapping.
E.g. the open question is whethere GGX reflection will be added or not and when we will finish multi-GPU. Hope to do it ASAP.
Few pictures about development in progress:
These are 15 min renders with scenes complex illumination interior with HDR lighting coming through 3 windows. HDR map has sharp bright sun drawn in it.
Windows have lattice to make obstacles.
The other 3 light sources are covered by some geometry: 2 lights covered by glass vases (on the left) and another one is covered by cylinder on the right.
Comparison presents "Another GPU engine", old CentiLeo 0.47 and current 0.48. Everywhere portals for windows are used, 6 GI bounces, and other settings are tuned for better performance in both engines. One 970 GTX is used for these tests.
Materials and shaders have the same complexity stucture in both engines, same amount of textures but the colors of floor and furniture are slightly different.
Hi Ali!
It's already added. Here is the difference between 0.47 (top) and internal 0.48 version (bottom). Everything is diffuse. 0.48 has Oren Nayar diffuse with roughness=0.4
. Very few small issues/bugs must be solved before that. Need to collect memory garbage when we switch on/off multiple times the threads for multi-GPU, also need to make Picture Viewer working (it was broken), and need to tune some params.