Samples (material) vs samples (scene)

Started by zooropa, February 02, 2018, 05:36:55 AM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

zooropa

Hi I was wondering how related are one to the other.

For example will a 512 sample render with a 8 sample mat will look different with one material 32 ?

As I understand samples in a material controls the graininess of that specific material. Still trying to wrap my head around to understand where are the 512 samples of my scene going with that specific material.

Thanks

Will Gibbons

This is a very nuanced topic that can be tough to explain. I try to offer what these mean in terms of end results rather than the actual math/physics behind it (since I do not know that).

Increasing material samples will place more priority on that material during the rendering (often removing graininess like you said).

Increasing screen samples (ie. max samples) will increase overall accuracy and level of detail (removing overall image noise). Screen samples determine how long KeyShot spends refining all aspects of the image such as ground shadows, global illumination, caustics, direct lighting, depth of field etc.

So, think of screen samples as an overall quality value and material samples as a way to get a specific material to smooth out even if you don't increase screen samples.

Does that help?

mattjgerard

So would this help in this particular image where the rest of the image looks good but the cloudy plastics are still a bit rough? I could assign more samples to the cloudy plastic materials themselves, and allow the system to spend more time calculating on those rather then wasting the time on the cabinet and other stuff that looks fine at way lower samples?

For example, this image rendered at 2500x2500 300DPI at 500 samples, took about 15 hours.

I know the everything but the cloudy plastics look fine at 128 samples, so if would set my main max samples to 128, but then up the samples in the cloudy plastic material? And if I understand it right, the system will spend more CPU on the CP's than the rest of the scene resulting in the CP's being cleaner but the rest of the image looking fine at its 128 sample level?


Edit: And this is after I did a bit of Nik Denoising in PS on the LED's

Will Gibbons

#3
Quote from: mattjgerard on February 02, 2018, 12:21:42 PM
So would this help in this particular image where the rest of the image looks good but the cloudy plastics are still a bit rough? I could assign more samples to the cloudy plastic materials themselves, and allow the system to spend more time calculating on those rather then wasting the time on the cabinet and other stuff that looks fine at way lower samples?
Yes.

Quote from: mattjgerard on February 02, 2018, 12:21:42 PM
For example, this image rendered at 2500x2500 300DPI at 500 samples, took about 15 hours.
Not to nitpick, but DPI is irrelevant when setting render resolution by pixels.

Quote from: mattjgerard on February 02, 2018, 12:21:42 PM
I know the everything but the cloudy plastics look fine at 128 samples, so if would set my main max samples to 128, but then up the samples in the cloudy plastic material? And if I understand it right, the system will spend more CPU on the CP's than the rest of the scene resulting in the CP's being cleaner but the rest of the image looking fine at its 128 sample level?
Yes. Cloudy Plastic is a material that I've set samples as high as around 100... (which would be way overkill for most all other materials)

And now that I think of it... since the LEDs behind the Cloudy Plastic parts are not shining on anything else, if you really wanted to optimize this, just render your image at 128, then hide everything except your LEDs and Cloudy Plastic bits at a higher sample rate (maybe 1000 or so) and comp them in PS. Probably what I'd do in this case.

zooropa

Quote from: Will Gibbons on February 02, 2018, 10:07:14 AM
This is a very nuanced topic that can be tough to explain. I try to offer what these mean in terms of end results rather than the actual math/physics behind it (since I do not know that).

Increasing material samples will place more priority on that material during the rendering (often removing graininess like you said).

Increasing screen samples (ie. max samples) will increase overall accuracy and level of detail (removing overall image noise). Screen samples determine how long KeyShot spends refining all aspects of the image such as ground shadows, global illumination, caustics, direct lighting, depth of field etc.

So, think of screen samples as an overall quality value and material samples as a way to get a specific material to smooth out even if you don't increase screen samples.

Does that help?

Thanks Will.  I guess I was in the right direction. A scene with 512 samples and 32 samples material will look different from a 512 and a 8 samples material. It sounds coherent now, but just that I thought...ok 32 or 8 is always within 512. That was the confusing part.
Summing up...they are two different thing. Where material samples are in a way above the general samples...in the tree hierarchy.

KeyShot

You can think of material samples as a way to boost the samples for individual materials. This is useful if you have a complex cloudy plastic material surrounded by less complex materials and lighting. The material samples are applied at the material level, while the overall samples is applied to every pixel in the image.

zooropa

Quote from: KeyShot on February 04, 2018, 09:52:42 AM
You can think of material samples as a way to boost the samples for individual materials. This is useful if you have a complex cloudy plastic material surrounded by less complex materials and lighting. The material samples are applied at the material level, while the overall samples is applied to every pixel in the image.

THANKS