PDA

View Full Version : MIA blur + large Scene Scale = x10 rendertime

Gal
09-18-2009, 01:51 PM
hey, something i've noticed recently. a bug? or just how mr works?
i have 1 object with a few mia shaders on it that include mild blur reflection(0.8, rays 8) and blur refract on 1 object/shader (0.6, rays 30) this renders in normal default grid scale at 3mins a frame. when moving this into a larger scene, which is relatively simple. large ibl and a further back imageplane.. scene is scaled x10 including all nodes(there are no procedural nodes).. rendertime becomes 35mins (x10).
i assume it has to do with the reflection blur/refract blur sampling/tracing which is affected by scale?.
any ideas why this happens?
could it be clipping ranges, must i introduce max distance (the ibl is considered as exit color right? so ibl should still work fine).
ndeboar
09-21-2009, 03:37 AM
from my basic knowledge of ray tracing, the ray does a hit test for every part of the ray, so the longer the ray, the more hit tests = slower render. So, setting a max distance for the rays "should" speed things up.
coccosoids
09-21-2009, 07:24 AM
Hey Gal,

do you think this particular occurance is reproducible? Using simple geometry with similar shader - one setup at scale 1:1, one at 10:1?
Gal
09-21-2009, 07:12 PM
actually it was BSP's fault. since the scene got much larger but the objects stayed the same size... the amount of usage of the scene was very very small comparing to the encompassing environemt(hidden geometry that gets reflected/generates gi) so the bsp depth now didnt have enough subdivisions to work optimized. basically 1 bsp division/square calculated the whole scene instead of many more subdivisions like it did before the scale. raising the depth from 40 to 60 changed it back to 2mins from previous 30mins. using bsp2 also did the same effect and got me 2mins as well.
CGTalk Moderation
09-21-2009, 07:12 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.