Page MenuHome

3D view rendered mode differs from F12 render
Closed, InvalidPublicTO DO

Description

Using 64bit Ubuntu 12.04 -r46479

If you examine the attached blend and image files, you can see a difference between the F12 render, and the 3d editor "rendered" mode view of the same cube.

I tried 2.63 and 2.62 blender.org releases, along with 2.63 and 2.63a svn releases that I compiled, and all have the same results in that they don't clip the F12 rendered image. The r46479 version in trunk however, has the clipping issue.

To save you some guess work in the attached files, r46479_rendered.png shows the problem. But in general (and in all versions tested), there seems to be a difference between all maximized views and the final render in the way the black environment reflections on the surface are done. My expectation would be for them to at least have the same artifacts in preview and final render, despite the fact that the artifacts are caused by mesh collisions.

Anyways, I wasn't entirely sure if this new behavior was expected or not, nor if you intended for there to be differences in the 3d editor view and the rendered output, and thought that I would error on the side of caution and let you know, just in case it is a bug. cya o/

Event Timeline

Okay, I was able to bisect the commits and narrow it down to the cycles branch merge just after 2.63 release:
-----
r45996 - works (2.63 release)
r46022 - works
r46023 - image broken (cycles branch merge)
r46025 - image broken
r46050 - still affected
r46100 - still affected
r46200 - still affected
r46433 - still affected
-----

I noticed mention of filter glossy changes, so I tried change it to 1.0 and 0.0 without change to the problem. Also, the 3d editors cycles render still was different from the rendered output anyways.

I hope this helps narrow down the problem enough.

Something that might help you as well, I just noticed that when I change from dynamic BVH to static, the 3d editor cycles rendered preview gets the same artifact that the rendered image got. I assume that dynamic/static only applies to the viewport and not F12 renders?

Brecht, can you look into this please?

Dan: Right, in F12 renders only Static gets used.

I can confirm the different renders as well, it's related to the BVH.

Moving this to the todo list of raytracing precision issues. Neither the viewport or final render works well, generally it's advised not to have faces touching like this. Either put the cube slightly above or below the plane, and you don't get self intersection issue.

Brecht Van Lommel (brecht) changed the task status from Unknown Status to Unknown Status.Oct 24 2012, 10:08 PM