Sampling problem

[fAEkE]

Take a look at the print. Same settings for light and materials e.t.c, the only difference is samples and noise threshold as seen to the right. Look at the artifacts. Looks like something between noise and splotches. How come this is produced by using higher sample settings?

[fAEkE]

I changed the HDRI dome subdivs from 32 to 64 and the artifacts dissapeared. Why :/ doesn't make sense that they came at all when changing to higher sampling subdivs when Max subdivs was lower than the light subdivs in both renders.. >.<

Vlado, I need you for my three sampling threads. I get more confused by each one

[king_max]

I think you can get the answer here: http://www.interstation3d.com/tutori...yfing_dmc.html
There was a thread around here by Toni Bratincevic linking to his site.

[fAEkE]

I've read it 2 times before, and if I understood it correctly, my artifacts shouldn't be there. I hope that someone can explain.

[king_max]

From what I understood, it makes perfect sense. Since you rose the DMC max samples to 12, you had to increase the light subdivisions to get the same result.

[fAEkE]

hmm, I read it again.. and I still don't quite understand it.

The total amount of samples from the light is the same (32*32=1024). In the first render they are divided by 5, so 6 subdivs per ray. In the second render they are divided by 12, so 3 subdivs per ray. The total amount of light sample available would be 1024, without adaptiveness. I guess this is what I don't get, right now atleast

First render= 2/5 DMC subdivs and 32 light subdivs. Adaptiveness 0.85.
Vray sends 15% of the samples and then get adaptive. So it sends 32*32*0,15 = 153 samples? One ray contains a max number of light samples which is (32/5)^2 = 41. Now a question, when is the DMC adaptive. I guess it's at this stage? So one eye ray actually only splits into 41*0,15 = 6 samples. And this is when Min samples of 8 kicks in? 153/8 = 19 eye rays used. What happens if the pixel ain't good enough compared to the noise threshold? Vray keeps adding secondary rays, or does it add eye rays?

Is this correct? This is so.. backwards :>

[king_max]

Ah. So much math going on here. I must admit I didn't completely comprehend the article. It's kinda backwards thinking as you said. To simplify things, I think that by rising the DMC max subdivisions to 12, you actually gave the renderer more "freedom" (margin) in terms of adaptiveness. Since the noise is only present on the background plane, it's safe to assume the sampler "thought" that area needed less samples and subdivided to the max allowed (by 12).

[fAEkE]

Ye, it's probably it, more freedom. But still, I would like my math to be confirmed by someone and I also need to know why vray behaves like this.. it's obviously something wrong since the result is bad even tho I use alot of samples and pretty low noise threshold. I mean.. I even went lower with the threshold when the quality got worse..

[super gnu]

hehe ive just experienced a similar thing.. had my dmc aa set to 1/100 and noise threshold to 0.07 did a few tests with different settings, and found that aa set to 1/50 rendered faster - no surprise- but also the AO and reflections were -less- grainy..

very odd.

[fAEkE]

Ye, I'm watching a Gnomon matte painting DVD, thinking about switching.. they don't have a DMC sampler to worry about..

[super gnu]

gardening i believe doesnt have these DMC issues either.. plus the weather is nice outside... hmmm.

[fAEkE]

lets hope someone will shed light on our sample issues so that we can stick to 3d

[3DMK]

gardening i believe doesnt have these DMC issues either.. plus the weather is nice outside... hmmm.

Gardening has LOTS of BUGS though........

[olika]

Wow! Why i didn`t see this thread when it started a few weeks ago?!
The more suprised am i that this thread just stops without somebody who can prove the math of the calculation above. I also read the "Demystifying V-Ray DMC Sampler" thread many times now but i don`t get it completely in my brain.
So just to ressurect this thread, isn`t there someody who can explain it? For me it`s just sometimes like a "bermuda triangle" between material subdivs, thresholds (dmc and also sampler) and min max settings. I just can`t understand the relationsships between them and the weighting of the different settings.
I guess in the end it`s not unbelievable complicated but maybe more naming conflicts between rays, subdivs, samples etc.
Another explanation with few other words could bring the big Ah ha! effect i am waiting for so long now

Grüße
Oli

[fAEkE]

Good that you bumped this thread, I hope that somebody can explain because I gave up a while after this thread. The ones that seem to understand says that you can optimize your rendertimes when you figure it out, but I have tried just about everything and I find it impossible to understand and use the article to get better rendertimes.

I still use the same settings as before, the ones I have figured out on my own just by testing and working with vray for many years.