Page MenuHome

Cycles Rendering Tiles Indicator Limited to 16
Closed, ResolvedPublic


System Information
openSuse 13.1; nvidia GTX 750

Blender Version
Broken: 2.76b
Worked: (optional)

Short description of error
When tiled rendering is used with cycles (blender render is fine) the orange indicators of the individual tiles are limited to 16.

Exact steps for others to reproduce the error
The default blend file works, only difference is setting up samples to 20k or something so it does not immediately finish rendering in Cycles. All other settings are default.
And you need probably more than 16 cores/threads,

Event Timeline

Pavol Klacansky (pavolzetor) created this task.
Pavol Klacansky (pavolzetor) raised the priority of this task from to Needs Triage by Developer.

Since I do not know how to attach an image within xmonad (drag & drop), here is a link to screenshot. Also, blender crashes a lot on that many threads.

Thomas Dinges (dingto) triaged this task as Needs Information from User priority.

I cannot confirm this. Setting threads manually to e.g. 64, results in 64 orange tiles.

@Thomas Dinges (dingto), not being able to reproduce the issue does not mean report is incomplete.

@Pavol Klacansky (pavolzetor), some questions:

  • Does it happen when you run blender --factory-startup ?
  • Does it happen with Blender Internal rendering?
  • Does using blender-softwaregl make any different to the behavior?
  • no (although I can't set more than 64 threads, but it still start rendering 144 tiles - same us number of available threads)
  • no
  • no

This is the blend file I am using.

So what’s the status here? If issue does not happen with --factory-settings then it’s just an addon doing stupid things probably…

Bastien Montagne (mont29) raised the priority of this task from Needs Information from User to Normal.

I cannot imagine that there is a bug in Cycles itself here...

if there are more threads active then there are cores then code execution might not be ideal, perhaps tuning on parallelism ea the tasker might solve it. (i've seen this in other parralel code that wasnt optimal, but i'm into the dark of how blender deals with it)

  • merely an advice for parallel coding. *