Isotropix Forums

Test with the new PBR engine in upcoming 3.0 RC1

Post your WIP image/animations here

Re: Test with the new PBR engine in upcoming 3.0 RC1

Unread postby xilofoton » Mon May 30, 2016 1:36 pm

Hi!

Photogrammerty model and texture of cobbles:
Cobbles_BudaiVar_WIP_2layerSh_bump.jpg

This is a WIP status, no comp at all(but LUT of course yes), more will come: height textured glossiness, rained version etc.
VFX Artist, Clarisse Specialist
https://scivfx.wordpress.com/
xilofoton
 
Posts: 198
Joined: Tue Jul 21, 2015 2:29 pm

Re: Test with the new PBR engine in upcoming 3.0 RC1

Unread postby MartinO » Mon May 30, 2016 4:26 pm

Sup_a10 wrote:I don't understand why the blog wrote that new PBR engine is slower ( than old System ), but also wrote "This means that you should need less samples to get rid of the noise" ??? If the new PBR need less sample, It should be faster than old render system, why it wrote " slower " ???


Yeah, I'm a bit curious about this as well.
MartinO
 
Posts: 184
Joined: Wed Jan 13, 2016 1:13 pm

Re: Test with the new PBR engine in upcoming 3.0 RC1

Unread postby paq » Mon May 30, 2016 4:32 pm

The way I understand it, is that samples are slower to compute, but you need less of them to get rid of the noise.
User avatar
paq
 
Posts: 306
Joined: Sat Jan 26, 2013 12:33 am
Location: Montréal

Re: Test with the new PBR engine in upcoming 3.0 RC1

Unread postby sam » Mon May 30, 2016 4:41 pm

@Sup @MartinO

That's easy. It depends what's your criteria to define actual speed. Is it the number of samples per second (iteration time) or the variance level on your image? Iteration time is a very objective measure whereas variance level isn't.

For example you can launch 100 iteration in 5 mins. On the contrary you can get the same image with 5 iterations but still take 5 mins (or more)...

It's very difficult to bench both, but one thing is sure we make a lot more computations for each iteration (so it's slower) to reduce the level of noise.

Is this clear now?
Sam Assadian
Isotropix
CEO/Founder
User avatar
sam
 
Posts: 979
Joined: Sat Jan 26, 2013 12:33 am

Re: Test with the new PBR engine in upcoming 3.0 RC1

Unread postby sam » Mon May 30, 2016 4:47 pm

I have another example: BDPT is a lot slower per iteration than a regular forward Path Tracer. However, in some specific configurations the BDPT will converge faster to a noise free image. It will require less samples and the overall render time is faster. However, the interactivity will be slower as it will take a lot more time to display the first iteration.
Sam Assadian
Isotropix
CEO/Founder
User avatar
sam
 
Posts: 979
Joined: Sat Jan 26, 2013 12:33 am

Re: Test with the new PBR engine in upcoming 3.0 RC1

Unread postby daaims » Mon May 30, 2016 11:06 pm

Does that mean we will have a sort of slider to control the interactivity in Clarisse with PBR engine ?

I have to shade --> full interactive
I have to render --> no interactivity, faster render

Right ? :mrgreen:
User avatar
daaims
 
Posts: 431
Joined: Fri Mar 22, 2013 1:17 am

Re: Test with the new PBR engine in upcoming 3.0 RC1

Unread postby sam » Mon May 30, 2016 11:52 pm

No because the we aren't doing BDPT and the Clarisse's PBR should be in average around 20 to 30% slower per iteration than with the legacy one. Using layered material vs Legacy Blend the gain is way bigger (up to several times faster) in favor of the PBR btw. The speed difference can't be really measured with the naked eye. Can you tell the difference between a 200 ms render vs a 260 ms one? :)
Sam Assadian
Isotropix
CEO/Founder
User avatar
sam
 
Posts: 979
Joined: Sat Jan 26, 2013 12:33 am

Re: Test with the new PBR engine in upcoming 3.0 RC1

Unread postby MartinO » Tue May 31, 2016 8:42 am

sam wrote:No because the we aren't doing BDPT and the Clarisse's PBR should be in average around 20 to 30% slower per iteration than with the legacy one. Using layered material vs Legacy Blend the gain is way bigger (up to several times faster) in favor of the PBR btw. The speed difference can't be really measured with the naked eye. Can you tell the difference between a 200 ms render vs a 260 ms one? :)


Makes sense, thanks for clearing it up!
MartinO
 
Posts: 184
Joined: Wed Jan 13, 2016 1:13 pm

Re: Test with the new PBR engine in upcoming 3.0 RC1

Unread postby xilofoton » Tue May 31, 2016 4:26 pm

Hi!

More lookdev:
Cobbles_BudaiVar_WIP_advSh_advComp.jpg


Roughness/bump depends on the "wear exposure" (occlusion), Now it has comp but no HDR env, manual lights (non-sunlit areas are flat).
VFX Artist, Clarisse Specialist
https://scivfx.wordpress.com/
xilofoton
 
Posts: 198
Joined: Tue Jul 21, 2015 2:29 pm

Re: Test with the new PBR engine in upcoming 3.0 RC1

Unread postby Sup_a10 » Thu Jun 02, 2016 2:36 am

I got it. Thanks. Now I understand why old system of Clarisse does faster interact than Arnold interact but slower than Arnold at final Render. so expect the version 3 !!!
Sup_a10
 
Posts: 10
Joined: Sat May 02, 2015 5:02 pm

Previous

Return to Work In Progress