num passes in lightcache

[flipside]

What is this for? Is it for speeding up LC on multi processor machines? If yes, I think you loose all the gained time during the 'merging LC passes' phase. This takes a long time already for low subdiv LC solutions, must take vray long on high quality LC's...

[Micha_cg]

num passes in lightcache

... an other disadvantage of more than 1 pass is, that I can not estimate the quality of the LC, because every new pass start at a low quality level.

[flipside]

num passes in lightcache

I don't notice that it starts over again?

How many passes do you set it to? I think you shouldn't go higher than the number of processors you have. A HT enabled proc means 2, a dual machine with HT = 4.

[Micha_cg]

num passes in lightcache

Thanks Wouter, I know.

[flipside]

num passes in lightcache

Ok, but why then does your LC calculation start over?

[Micha_cg]

num passes in lightcache

I use a dual core machine with 2 passes and after the half LC pass, the noise level jump and clear up slow again. It's not a big problem, but with a single pass I can see the full quality of LC at the end of LC pass.

I think, if the LC pass is to noisy, the primary engine need more time to clean up the rendering or? What do you think?

[flipside]

num passes in lightcache

The primary engine will try to use the LC info, and if it is too noisy and incorrect, this will get carried on in the primary GI calculations, I think.

Any idea what the adaptive option in LC parameters is?

[flipside]

num passes in lightcache

Mmm that merging is getting on my nerves here It takes about the same time as the actual calculation time!

[vlado]

num passes in lightcache

Mmm that merging is getting on my nerves here It takes about the same time as the actual calculation time!

V-Ray has always been doing the merging and it's typically very fast, although this was not indicated in the progress bar. I suppose that if there are many samples, then constantly updating the progress may slow things down (there was a similar issue with "building static accelerator" progress some time ago). Hopefully this can be corrected.

Best regards,
Vlado

[flipside]

num passes in lightcache

Ah yes, I see. The progress also starts very fast, but it slows more and more towards the end.

[joe_b]

num passes in lightcache

Mmm that merging is getting on my nerves here It takes about the same time as the actual calculation time!

V-Ray has always been doing the merging and it's typically very fast, although this was not indicated in the progress bar. I suppose that if there are many samples, then constantly updating the progress may slow things down (there was a similar issue with "building static accelerator" progress some time ago). Hopefully this can be corrected.

Best regards,
Vlado

I'll look into it. I thought I only incremented after a certain percentage had been completed - but I could be wrong.

[garrett_jsaa]

num passes in lightcache

I am using Micha's awesome tutorial for rendering exteriors. I initially set the number of phases to 1, but the Render Progress box says:

"Performance loss: number of rendering threads (2) is less than number of light cache passes (1); some threads will be idle."

Does this mean anything for the quality in the end?

[Micha_cg]

num passes in lightcache

Best you set the phases at the number of CPUs/Threads you can use. For a dual core is it 2 here.