Page 3 of 3

Re: LuxCoreRender v2.0beta2 released

Posted: Sat Apr 07, 2018 2:20 pm
by Racleborg
This is so coooooool! :D

Re: LuxCoreRender v2.0beta2 released

Posted: Sat Apr 14, 2018 3:08 pm
by zukazuka
It's working great

Re: LuxCoreRender v2.0beta2 released

Posted: Mon Apr 16, 2018 10:07 am
by Sharlybg
The solution for light cache I have in mind is a bit less sophisticate than Corona one but it is less sophisticate to make possible a GPU implementation. So the outcome should be in favor of LuxCore. Especially if you have a Titan V :mrgreen:

Corona's UHDCache is really a good solution with on-line training, small memory footprint, etc. but if you run 5-10 times faster in term of brute force, you have a very good head start ;)
What are the difference ?

Luxcore will not be online ?

Re: LuxCoreRender v2.0beta2 released

Posted: Mon Apr 16, 2018 1:57 pm
by Dade
Sharlybg wrote: Mon Apr 16, 2018 10:07 am
The solution for light cache I have in mind is a bit less sophisticate than Corona one but it is less sophisticate to make possible a GPU implementation. So the outcome should be in favor of LuxCore. Especially if you have a Titan V :mrgreen:

Corona's UHDCache is really a good solution with on-line training, small memory footprint, etc. but if you run 5-10 times faster in term of brute force, you have a very good head start ;)
What are the difference ?

Luxcore will not be online ?
No because of GPUs. It will be built during a pre-processing stage so it can be transferred to the GPUs.

Re: LuxCoreRender v2.0beta2 released

Posted: Mon Apr 16, 2018 4:42 pm
by Sharlybg
No because of GPUs. It will be built during a pre-processing stage so it can be transferred to the GPUs.
excuse me if i'm insisting like this but i've more question ;)

If we take CPU only as reference (no GPU acceleration).

1) Do we lose a lot of performance compared to online method ?

Re: LuxCoreRender v2.0beta2 released

Posted: Mon Apr 16, 2018 5:05 pm
by Dade
Sharlybg wrote: Mon Apr 16, 2018 4:42 pm
No because of GPUs. It will be built during a pre-processing stage so it can be transferred to the GPUs.
excuse me if i'm insisting like this but i've more question ;)

If we take CPU only as reference (no GPU acceleration).

1) Do we lose a lot of performance compared to online method ?
On-line training is better because the quality of the light cache improve over the rendering time. The quality of light cache built during a pre-processing phase is constant over the rendering time. CPU Vs. CPU, a light cache with on-line training is better than one without.

Re: LuxCoreRender v2.0beta2 released

Posted: Mon Apr 16, 2018 5:17 pm
by Sharlybg
On-line training is better because the quality of the light cache improve over the rendering time. The quality of light cache built during a pre-processing phase is constant over the rendering time. CPU Vs. CPU, a light cache with on-line training is better than one without.
Oh this is what i fear. It seem to be the most efficient way. If i've to choose i go for online. The current state of luxcore bidir engine is already good even on mid range i7. All my interior and caustics renders goes througt bidir cpu. I prefere to have online light cache on a Threadripper (3x faster than my i7 6700k). i think it is more efficient to have Bidir + online LC for indirect and caustic and GPU + path tracer + denoise for simple case and exterior.

P.S

Or maybe i'm wrong :

(offline LC + Bidir / offline LC + GPU) is better than (Online LC on CPU only) :?

Re: LuxCoreRender v2.0beta2 released

Posted: Mon Apr 16, 2018 6:08 pm
by Sharlybg
Finally i bet on pure online on cpu + bidir for now. And we can wait after to add simpler LC on GPU. A solid and efficient method can do wonder like we already see with corona.

But of course it is a personal though.