Chaos Corona Forum
Chaos Corona for 3ds Max => [Max] General Discussion => Topic started by: Marchwiany on 2017-04-27, 14:56:27
-
Hi ,
got a problem with the noise and a kind of artifacts. It looks like some white spots near the lit places.
There was no problem before upgradeing Corona.
Does anyone have a similar issue and if yes, can you explain what to do to solve it?
Kuba
-
hi, can you show us an image of the problem?
-
Hi, this is 75 pass raw render. I have to wait till ~1000th pass to reduce noise to the acceptable level (+denoising)
-
Hello,
I notice weird behavior of lights, artifacts (much more noise) as well, I haven't had time to check everything properly and test it via 1.5 but now that I see this posted around, I guess my eyes weren't deceiving me. I will post shortly very similar issues, I'm in the middle of a tight project at the moment.
I've noticed that putting a portal light, gives me exactly those kind of artifacts (show in the image posted by Marchwiany) in the ceiling, (direct light) while darker areas area clearer than with NO portal light. (this happens on a square room with one wall being completely opened.
-
I have the same feeling, it looks like 1.6 noise cleanup is longer. Maybe it's just a feeling. I just reverted back to 1.5.2 to make some tests. I'll let you know.
-
Hi, this is 75 pass raw render. I have to wait till ~1000th pass to reduce noise to the acceptable level (+denoising)
can you share the scene privately with us (see my signature)? It looks like a perfect candidate to test new algorithms we are just now creating that shall help a LOT in this case
-
Here is the VR scene i'm currently working on (attachments).
Corona 1.5.2 : 73 passes - 4.5% noise - 7Hr
Corona 1.6 : 73 passes - 4.5% noise - 8Hr
Corona sun + Corona sky.
Edit : ceiling ligths looks strange in 1.5.2 because i was using this in 1.6 :
“Prevent black appearance” for Corona Lights and CoronaLightMtl is a new option that will avoid them showing as black when light constrained by Directionality or an IES profile does not reach the camera.
-
I wonder why rays per sample is about 12% higher in 1.6 in your example? Render time is higher by roughly the same 12% because of that.
-
I don't know, it's the exact same scene. Saved rigth after the 1.6 render, rolled back to 1.5.2 and redered again.
-
I just ran a quick test on 3 scenes and all of them behaves differently. One shows roughly the same rays/sample in both 1.5.2 and 1.6 and render slightly faster in 1.6, the other shows about 25% more rays/sample in 1.6, but renders in about the same time and the third one shows about 20% more rays/sample and renders about 10% slower in 1.6 All scenes uses PT+UHD. However those were very short tests, 2-4 minutes each, so results may be not very accurate.
-
we changed the light sampling algorithm, so more samples generated are not immediately rejected and get traced. This usually lowers noise a lot, but if the scene has heavy occlusion (windows without portals), most of those traced rays might end up being occluded. We are working on including occlusion into the mix for 1.7
-
That makes sense - the scene that renders faster is fairly open and has portal in it, two others are more closed and hasn't portals :]
-
As my scene is fairly open, i didn't tryed to place some portals. I'm gonna give it a try. Thanks for the feedback.
-
We are working on including occlusion into the mix for 1.7
What do you mean ? Some sort of auto-portal ?
-
and what is temporary workaround for this problem? in 1.6?
-
We are working on including occlusion into the mix for 1.7
What do you mean ? Some sort of auto-portal ?
well... if I was nvidia, I would say that we are unleashing the power of machine learning onto rendering! :D
-
We are working on including occlusion into the mix for 1.7
What do you mean ? Some sort of auto-portal ?
well... if I was nvidia, I would say that we are unleashing the power of machine learning onto rendering! :D
So does that mean that corona will eventually know what I need to render, even before I know it myself, and make it for me? Finally a "make art" button.
-
Hi, this is 75 pass raw render. I have to wait till ~1000th pass to reduce noise to the acceptable level (+denoising)
Is there a chance you have an object / some objects in this scene which are complex geometry with Corona Light mtl applied? The reason I am asking is that I have seen a very similar result in a scene with many lights (light objects) plus some complex geometry with light mtl applied.
-
I'm having the same problem. Much more noise and more time to clean it.
I have to increase gi to aa to 64 to make it better, but still taking more time than 1.52.
-
Please, post ta simplified *.max files. It will be much easier, for us, to help you. =)
-
Here is a simplified MAX file. https://we.tl/bU7JkLS35D
The main issue is cleaning the Indirect light. Even the direct light is taking longer to clean.
And this is not happening only in this file, all our files we have to increase GI to AA level and sometimes the Light Samples Multiplier.
-
Interesting, so is it confirmed that the new sampling is slower more often than not or how should we go about this? maybe default GI vs AA should be 32 instead of 16?