OTOY Forums

Congratulations, great result!

The fireflies wasn't such a problem for me (easily removeables). I think the best is the improvement in difused caustics. It took a lot of time to get them right.

Just one question: It seems that dark areas are noisier. Maybe is it because less processing time is focused to these areas now? anyway the overall quality is a lot better.

Very nice to hear those good news

Kind regards
Alain

Wow, looks fantastic! What does PMC stand for?

It looks great, cant wait for it. Can you give at least approximate guess, when do you expect it all to be ready? In a month or 2 or 3?

Will multiple GPUs scale 100 percent? Right now they dont with CUDA 3.0, at least not on my gtx590.

Daniel wrote:Wow, looks fantastic! What does PMC stand for?

I suppose MC stands for Monte Carlo, but thats just wild guess No idea about the P though, maybe something with pathtracing.

I´d like to give thanks to the people at refractive software,
and thanks twice for the brain juice you spent on this new render kernel thing!
Goodbye fireflies, hello caustics.

Same little question thats bugs me; what does PMC stand for?
it´s mutating? Like something similar to the MLT thingy?
I just read something about it on wikipedia and understood less than a third of it. But it sounded promising for GPU approaches.
Please tease some more!

Thanks for the sneak peek and great work! I can't wait for the new kernel

this is WERY needed stuff

keep on good work

Impressive!

Will there be any explanations on how this algorithm is different?
I mean in the logic level not the algorithm itself of course

Btw - will the old kernel be still available?

I added another PMC kernel rendered image to the first post with a really tough scene, using area lights and IES distributions.


convergen and others -> I won't give any timeline/release date as we are not sure yet. It won't be more than a month, and won't be less than 2 weeks from now.
steveps3 -> The new kernel renders unbiased (just like pathtracing) and eliminates the firefly problem, renders complex lighting scenarios much more efficient and as a result explores and converges caustics much better. The sunlight through glass issue will also be fixed in it but I am not sure if this will make it into the 2.5 first release, might be 2.51 or so soon after.
areyal -> Yes, the new kernel is an importance sampled kernel, so it focuses more where needed and less where not. We are still working on it and optimizing it so it does so better, but it's already doing a great job.
Daniel -> PMC stands for Population Monte Carlo, it's a group of algorithms that are more advanced and efficient (but are similar) to MLT (Metropolis Light Transport) and are better suitable for a GPU.
Timmaigh -> the new multi-gpu framework should be near linear yes, but we are still working on it so I can't promise anything at this point.
kivig -> offcourse, it's a 3rd kernel, pathtracing and directlighting will remain available.


Radiance