A little preview of our new PMC render kernel in beta 2.5

[Refracty]

PMC is really promising. Thank you for sweating for us.
But please tell us if the 'sunlight does not shine through transparent / specular materials' problem will be fixed for the next release (pathtracing)*.
Thanks
- Refracty


* I hope this will fixed in the direct lighting engine as well because I get headaches solving around.

[abstrax]

PMC is really promising. Thank you for sweating for us.
But please tell us if the 'sunlight does not shine through transparent / specular materials' problem will be fixed for the next release (pathtracing)*.
Thanks
- Refracty

* I hope this will fixed in the direct lighting engine as well because I get headaches solving around.

Sorry, I can't say. We are trying, but it's a hard problem for path tracing (and alike) algorithms. To give you an idea:

The sun has a visual diameter of 0.5 degrees. A quick calculation on a napkin shows, that the sun covers less than 1/100,000th of the hemisphere. This means of 100,000 rays you shoot randomly from a point only 1 (in average) hits the sun.

Cheers,
Marcus

[Refracty]

Thank you Marcus for the quick reply.
Now I understand the problem. I hope you will find a way. Maybe you could let the sun object send out (only) parallel rays towards the scene extension instead of loosing 99,999 % of the emmitted rays. I know that this is easier said then done and might bring other complications on the table - just an idea.
Cheers
Refracty

[abstrax]

Thank you Marcus for the quick reply.
Now I understand the problem. I hope you will find a way. Maybe you could let the sun object send out (only) parallel rays towards the scene extension instead of loosing 99,999 % of the emmitted rays. I know that this is easier said then done and might bring other complications on the table - just an idea.
Cheers
Refracty

What you described is called bidirectional path tracing. We are looking into this and other options, but we don't have a solution that runs fast and robustly on the GPU yet.

Cheers,
Marcus

[tyrot]

marcus

as far as i understand you address the high end side of octane ..

do you have any solution for improving directlight rendering as well. Something between directlighting and pathtracing.. For me it is virtually impossible to use PMC or PATHTRACING for archviz video...

thanks

[tungee]

Abtrax why i dont get any answer to my questions.

[Refracty]

do you have any solution for improving directlight rendering as well. Something between directlighting and pathtracing.. For me it is virtually impossible to use PMC or PATHTRACING for archviz video...

Radiance mentioned some month ago that he is thinking about a improved direct lighting kernel.
I am sure that RS will dip into that after completing 2.5.
We all need a 'complete' version 1 with a descent engine for animation and quick render jobs.
They know.

[abstrax]

Abtrax why i dont get any answer to my questions.

Sorry, missed it. At the moment our focus is on the new kernel and the new framework. Absorption will definitely added in the future, but I really can't say when. We haven't thought about nk-materials yet, but I don't see a reason why they could not be added.

Cheers,
Marcus

[abstrax]

marcus

as far as i understand you address the high end side of octane ..

do you have any solution for improving directlight rendering as well. Something between directlighting and pathtracing.. For me it is virtually impossible to use PMC or PATHTRACING for archviz video...

thanks

Well there is only so much you can do without slowing it down. The only obvious thing is transparent shadows, which we can't add at the moment, because the CUDA compiler/optimizer dies on us. Otherwise we would have added it to direct lighting a long time ago. To solve that problem we have to restructure the kernel, which takes time, we couldn't find yet.

Cheers,
Marcus

[tungee]

Thanx Abtrax for clarification