vblackrender., hi! I know this problem. It appears randomly and in different places. Basically some geometry small part is removed here and there. There is a problem to find solution because the bug changes positions and not always reproduced in the same conditions.
I hope that the problem will dissapear when the geometry engine (compilation + ray intersection) will be improved. Maybe some of geo compression scheme is the reason. But we need time to fix that.

Hi purush thaman, welcome here! I know it's a problem without stadard AOVs. They are important and will be implemented, not a big deal. But currently very busy with other tasks, like multi-instances, particles and bugfixes. Let's move to a new AOV titled topic without mixing with an old one?

Research has shown that HDR resoultion issues were not issues in that case. But! There are 4 drawn light sources in your HDR. You may see them if move the camera outside the window.
1 of lights reaches the room through the window. Another one partially reaches the room. But 2 light sources of HDR image don't reach the room at all because they are blocked by top wall. But they are bright and grab the samples computations + add 2x more noise because return no any result. So that's why so noisy. Octane is even more noisy in this scene (a lot).
I see 2 ways to fix this:
1) to erase not-needed lights from HDR image by user.
2) wait for some my adaptivity fix which may help and will cost probably +100Kb of memory consuption + tests/debugging or something like that

However HDR resolution issue appear in another case with very narrow bright light making very hard shadow. It's almost fixed. That problem gave 100x more noise than normal situation.

Hi vblackrender.,
At the moment I am working on improving HDR sampling with that your scene. Describe it in your topic.

Change Log: cntlc4d 0.592 alpha (2020 May 12) - currently for all R16-R21
- Fixed broken Absorption property
- Fixed bug for lost frames during animation rendering in PV
- Fixed bug with wrong geometry translation when IPR doesn’t account for some deformers (such as Surface Deformer and few others)
- Fixed few bugs which caused some crashes.

I have uploaded this fix. Sorry for some delay, there were a few annoying technical problems there. Now continue with other fixes, especially the one with HDR sampling.

One more tomorrow!

If you switch-off "Static noise" in render settings the noise seed pattern will be different from frame to frame. Such property can be needed in animation denoisers.
How did you make it working with black 1/2 frames?

There should be no flickering, it uses no caches. Flickering may appear in cases of caches (especially irradiance cache with some large gather radius).

Maybe I was a bit not clear deep night. It seems to be that not only the resolution of HDR is important (lowres or highres), but the area of light vs background and intensity of that light. If HDR is high res, has small area of light with very high intensity - there can be a problem. But will need to investigate more. Probably, some of my past HDR optimizations have failed and I need to spend more memory for them or just make them adaptive.

Btw, it lags on my PC too! Very slightly. The cooling system works too intensively and with the same rythm.

Quote
vblackrender. wrote: Here we go with the next test. I played with the "scratches" shader. I am amazed of this shader. So easy to give every material the little extra for more realism!

Personally, I don't like our scratches already. Need to have bend property and better distribution of their length (they are too similar). So, improvements are expected there once I know how to do them
At the same time I like Flakes a lot more. They can be used as masks for multi-material stacking and combine several reflection shaders. Such manipulations are much easier in Node Editor of course.

vblackrender., thanks for scene!
there are 2 problems with interior:
1) Post-process exposure 6.7 - it's too hard for rendering sample to operate on almost black lighting and then compensate it in post. Numerical imprecision and other things have more influence at this scale. It's better to operate on exposure at values close to 0 or 1, 2. Don't forget to make Expected Exposure in render settings to be similar to the one in Post.
What I did is to decrease Post Exposure, Increase HDR multiplier and increase the Clamp Direct/Indect values to 20. It's less noisy, but also has different contrast. So need to improve precision in future.
2) Some problems with sampling of your HDR. I don't know yet, but maybe the sampable resolution that we consruct is too low, need to see and not make +1 user setting. Actually I have tested another HDR (from the FB post scene). And it has much less noise. However, it's also bright, contrast (narrow sun), but it's lowres. Lowres/highres of original HDR shouldn't be influential, but the resolution of sampable HDR is.

Dark corners in 2 fixed versions have less noise as I was looking for. So I am calm for that CentiLeo special GI sampling algo. Just need to improve HDR initial procession

Bad... Too much noise! It should be less! Something like here on a similar setup https://www.facebook.com/photo.php?fbid=949066952138604&set=pcb.2439178389694860&type=3&theater&ifg=1 (less noise with 4 iters).

Please, can you send me a scene file and HDR for experiments to info@centileo.com (via some kind of dropbox or similar). Non-distribution is guaranteed! Maxon trusts us
Maybe I can find some quick improvement, maybe not.

Hi vblackrender., the issue is known.
And it's solved in 0.59 (http://centileo.com/news/126/centileo-for-cinema4d-0591/)
For earlier versions I just need to recompile earlier plugins (there were some issues there). I hope to do it tomorrow

Ok, HDRI with portal is less noisy than without them. I am more calm.
Please, show another HDRI portal vs no portal but without any interior area lights. With equal amount of iterations: 1, 2 or 4, and show the image and time for both cases (with portal it should be around 15-20% more). That would be interesting to analize some internal technical solutions.

Quote
vblackrender. wrote: I havent any bump activated on the material on the wall. I have denoised in post the walls a a little bit, but the noise on the dark corners where hard to get rid off.

Light Shift is not only for Bump (but it sits there) - it's for shifting light alone, for some low poly areas. However, for Bump it's more valuable in some cases that's why it's there. Maybe I will move it to Diffuse.

Quote
vblackrender. wrote: Yes, I am using always Portals. But in this case, when I use a HDR the rendertime is logically going up. Just with a white sky, would render as least twice as fast.

Sure the HDR + Portals add more computation (shadow rays, etc), but the Portals add more inteligence to sampling and converge cleaner. What's the difference with and without them and without internal area lights? I am asking all of this because I feel the day scene can have less noise.

Quote
vblackrender. wrote: Here we go with the night version of the interior. I have used a blue sky here. The render was much faster compared to the HDR lighted version.

The last image is cool too! And iterations are faster because the HDR is dim

It seems the 3rd party JUCE library that is used for IPR window is going buggy in your system. Potentially it's conflicting with something else present there. I don't have such problem not on Win7 nor on 10. I think that transitioning to Cinema native window GUI will fix the issue and will be docable (as C4D people love so much).

Just an experiment: what if you remove Corona from plugins folder, test the subject issue whether there is a bug. And then move Corona back? There were some problems with co-existence with Corona in the past because of IPR (their's and our's).

Thanks for clear comparsion! Pls send me your settings, I will have a look. Not right now, but will make a note on this.
Btw, have you used the a light portal in window to accelerate HDR lighting? There is a setting in light tag (Is Portal). A corresponding light source should be area rectangular light.

Also there is a little fake feature called "Light Shift" which sits inside Bump tab of Material settings. Higher values narrows the incoming lighting angles and hide bump mapping artifacts (if any) on low poly parts (if they are there). In the front left corner of a bed there is probably one such issue or just a hard shadow.

Hi vblackrender.,
It's quite serious lag, I would like to solve it. There will still be some optimizations there with image delivery, IPR itselft and render time, I hope they will help more.

Have you tested when Photoshop is not running at the same time? Just Cinema + CentiLeo. How is that? What is your RAM and free Disc space amount, OS paging file size?

Yes, absolutelly OptiX denoiser will be integrated and all it's resource needs will be accomodated inside CentiLeo. Actually in R21 there is Intel Denoiser in Cinema itself and CentiLeo uses it from there for all AOVs. It's not realtime, it takes around 5 sec to denoise a 1-2 Mpix image if I recall correctly. Quality is not super, but anyway for interior walls it's fine. I think that OptiX is close to quality and much faster.

Hi vblackrender., It's great! I like it a lot!
The fabric on a bed is very-very good! However, to my taste the carpet still needs some work. Btw, what's the raw output of CentiLeo image? Just to make a visual comparsion