In some scenes, you can either have the scene render really slow with the cache off or have the cache on and get artifacts.
Artifacts easily appear once you start rendering real complex scenes that are almost completely lit with indirect lighting, also for some reason the size and number of faces on coplanar surfaces affect the cache results when it seems like they shouldn’t. We need either an improved cache with more options or something like bilateral blur done during the rendering instead of using the compositor so it can be used with reflecting and refracting objects.
Ceiling light is completely wrong as expected from a non GI renderer. I definitely need another renderer. This renderer can ruin any nice sculpture in a few moments. Pity…
Your render was named Testscene_64smp_cache_1h17min_2.jpg and it was at 1024x576 pixels.
So I opened the blend file and changed the render resolution (1024x576 at 100% instead 640x480 at 37%). I changed also in Gather, samples from 8 to 64 as your render name suggests. I did no more changes that these two.
Render time you asked: in Phenom 2 X4 965 3400Mhz was of 9min15sec
So I suppose there is some more change you did or you have a really slow computer. Perhaps turn off cache too?
The one from The last rookie looked fairly good when it came to ceiling light. I think it can do simple scenes with lots of diffuse surfaces well and at least a small amount of visible light sources, it’s mainly highly complex scenes with a ton of indirect lighting, reflections, SSS ect… that the GI solution has trouble with.
@ace dragon
Just place a model with some displacement map - heavy subsurface and all this process will come to an endless calculation, sorry… but is what I’m facing now. OK the truth is that I’m waiting.
I assume that what they do will be mainly for animation purpose.
If you need product / object stills where you will see artifacts
because it is a still and not animation I would go with a different
system.
But I am pretty sure that what we have right now is just at an initial
stage and not finale version of GI in Blender.
There’s still work being done in the render25 branch, so there’s still time to polish the GI system.
However, if this is close to being final, then it would be a PR disaster for the Blender Institute being that the GI only works in specific cases and would give the impression that Brecht doesn’t know how to properly implement rendering features.
There’s the discussion in the 2.5 regarding whether or not the devs. feel they should just put in hacky code for an open project rather than implement it properly, I hope such a case won’t be true for the GI, that Farsthary won’t have to come in and fix Brecht’s GI coding.
Also, the full shading method creates more problems than it solves making it not worth the extra realism in bumpmaps.
They are close to finish it? thats news to me. I was under the impression they just start working on it.
The material system rework is also left to be evaluated.
I never said that, it’s very likely they aren’t close to finishing the GI.
The scenes like the ones of the city in the Durian short and Sintel’s cabin will likely have tons of polygons and introduce a lot of cases where the GI needs to be improved, so Brecht will need to go back to the GI code in those cases.
Tea Monster: There is the fact about more than half of the commits in the render25 branch is not related to the Internal Renderer itself, probably because the Durian team as a whole is using it now, check the commit logs for yourself if you don’t believe me, not all commits are related to the rendering, lighting, and shading systems.
I don’t get those kind of artifacts with the cache on, it’s more like small blotches of incorrect lighting when I try to use the cache, more like what a lot of people can get with regular photon-mapping.
Or Thea rendering.
