Re-assigning, as @Richard Antalik (ISS) is working on this.
Tue, Jan 15
You can make the object appear diffuse in the reflections if you want, by using the light path node and a mix shader.
I think this is a known limitation of baked probes, see https://docs.blender.org/manual/en/dev/render/eevee/limitations.html#indirect-lighting
Mon, Jan 14
Sun, Jan 13
Thx for getting back, looks like this can be closed then...
Alright after 8 hours of rendering, I think the problem has fixed itself.. Must've been the updates from AMD drivers.. I rendered as images for 8 hours and nothing happened until I decided to use a little bit too much cpu when I was on the pc and it crashed. I'm sure that rendering as a video would most likely yield the same result. I'm certain my problem has been fixed by something new in blender builds or AMD drivers over the months. Thanks for your help!
I have noticed, that when you create new file from video editing template, render result is shown in new window, sequencer drawing code is not executed.
Proxies for scene strips should be replaced by cache dumping system, so I would also close this.
This is because cache system will manage validity of a frame, so we can do minimal necessary rebuild instead of manual re-render.
Sat, Jan 12
Fri, Jan 11
Thu, Jan 10
Speak of the devil... as I made that reply it just crashed. I didn't get a chance to read the command log, I assume it's the same thing but is there a way to read the command log again? It seems to have gone for a lot longer this time, it reached frame 600 until it crashed, so something went right..
I apologize for not being active in this thread, I'm making this reply for now before someone starts figuring out what's wrong. I've downloaded the latest build and updated my drivers, I retried rendering the test scene again and it seems to work. I'm rendering a scene I made a few months ago that I didn't get to render, and seeing if it'll crash after a night of rendering.
Wed, Jan 9
GPUTetxure: texture alloc failed. Not enough Video memory
I don't see a bug here:
- The shortcut shows as Shift+S and that one works, it's not Ctrl+S.
- The default is now to open a new window, you can change this in Render menu > Display Mode.
Tue, Jan 8
@Lukas Stockner (lukasstockner97) I tried the factory startup with no result.
I'm sorry if i failed to see the output, but when i start Blender with the given arguments and start the rendering Blender crashes and doesn't show anyting in the command line (no output).
Mon, Jan 7
Ok, so it is render. In fact, can reproduce it now on my computer. Looking into it.
@Milan Jaros (jar091), i've worked around the crash by tweaking an actual object which was causing OpenSubdiv to crash. We are looking into a more robust solution, but that'd take a bit of time.
Sun, Jan 6
You can check the results on your server. I used BVH2 with primitive=LINE_SEGMENTS, but it happens without those parameters too.
It happens on render. I tried it on commit from Jan 4.
Does this happen on render? Animation playback?
What is the SVN revision of the production repo used?
Sorry for noise.
Oh got it now.
I made a new file from scratch in -4b7596ec914e- and it's not just the text strip that caused the crash but a camera keyframe bind:
Can not reproduce.
Fri, Jan 4
why not control it with the options in the performance section of blender? set it to use a single thread and off you go, leaving it on auto (which in your case is probably 12) and limiting affinity manually to a single core will lead to excessive context switches ruining performance for the whole system.
Ability to manually control affinity is crucial if one needs to use the PC for other tasks while also waiting for some long render to finish. In some cases, setting CPU priority to low is not enough to free up enough resources, fast enough so that other software does not lag.
@Brecht Van Lommel (brecht) did you try with an asan build? It seems that by simply turning on ASAN I'm getting the issues mentioned above. It can be an ASAN false positive too, of course.
@Brecht Van Lommel (brecht) For the records, I still run into a crash with this file (and any other eevee file actually).
Our current goal is to give best performance for artist on a machine with lots of cores (> 64) and on CPUs with non-trivial topology (i.e. Threadripper2 and certain configurations of EPYC computers). Currently this implies following aspects:
Oh. I forgot about that. I usually set them both to the same value so this won't happen. I guess I missed this one. Sorry for the fake bug.
This is not an issue. You have different subdivision counts on view port and render in you first subd modifier. So the result is different in viewport and the actual render.