I have this problem with lightcache. If I set the render to view and have it calculate the light cache for the entire image, the quality, especially around these curtains is fairly good. Pardon the crop, can’t show entire image.
If I change nothing other then set the render from view to region and set region box just around the curtains and render, it looks really bad.
I thought the Lightcache was resolution independant and this sort of behaviour shouldn’t be happening. Any ideas?
When rendering a region, V-Ray will reduce the number of light cache samples proportionally to the area of the region. If you use relatively low number of light cache samples, this may be not enough.
I see. That would explain why raising the subdivisions from 1000 for the region to like 10000 did in fact help. Has LC always been this way? ALso, this makes it tough to do region renders when I need to. Sure raising subdivisions will help, but the regions also look darker then the LC calculated for the entire image. This seems counter-intuitive.
Oddly enough, I just started noticing this yesterday too, when I had to render some small regions to patch over a large base render. The lighting was so completely different, much more than I remember ever happening before. I used to not have a problem when doing small region renders.
This apparently is new in between 1.5 final and Service Pack 1, which we installed this week and is why both Percy and myself noticed the problem.
Here is a comparison. These are both small regions of a high res scene. In 1.5 the region renders pretty much the way you’d expect it to, and would match the full view render. But with SP1, because the light cache samples are proportionally reduced, the quality is not high enough and you get really wacky results. This is not good, because we do region renders all the time and I would rather not have to guess at how to adjust the subdivisions each time. Can we get it changed back to the way it was?
It is intended to work as it does in the service pack; I’ll make some tests to make sure that things are not off somewhere, but I’m pretty sure that it is correct.
Hmm…is there a reason you decided to change the way this works? From our perspective it is not an improvement, since this way requires you to guess how many subdivisions you’ll need based on the region size you are rendering. In previous Vray versions, small regions usually matched the full renders pretty closely, and now I’m worried that they won’t match as well, even if we turn up the subdivisions on the regions.
Well, the samples were reduced in the same way in the 1.4x versions, and while I myself have no preference on this, I was asked to return the behaviour of the 1.4x versions; if there is use for both methods, I can add an option for this.
Oh, I had never noticed that in 1.47 before. Strange…well, I hate to make more work for you, but I guess it would be nice to be able to use both methods if some prefer it the other way.
Vlado, I think this is related to the DMC bug, which you fixed.
DMC is selected as default path sampler, hence it behaves like it does int he ppt, stopping earlier.
What we see is just undersampling that is inherent to this (released)version fo the DMC algorithm.
If you people try and switch the sampler to legacy, you’ll notice the differences are minimal, if at all there.
-wild guess, forgive the language, it’s late nite XD -
I think there’s definitely a need for the old method at least. I frequently render regions to “fix” small areas of a rendering that have been changed rather than completely rerendering the entire scene. With the new way these regions would be unusable.
so is this bug fixed now in the SP? or will this come in another SP?
I have to patch large renders with regions and up till now I have not used light cache for this exact reason, the patch will always look different with different samples.