Page MenuHome

Blender 2.79 vs 2.8 - Viewport orbiting speed is heavily impacted
Open, Confirmed, MediumPublic


System Information
Operating system: Ubuntu 18.04
Graphics card: GTX 1080

Blender Version
Broken: 2.8
Worked: 2.79

Short description of error
Viewport orbiting speed is heavily impacted

Exact steps for others to reproduce the error

  1. Open attached blend file in blender 2.79
  2. middle mouse drag to orbit around, notice how it is slightly laggy, but not too bad to work with
  3. Open attached blend file in blender 2.8
  4. middle mouse drag to orbit around, notice how it is significantly laggy, and unworkable for a production scene.

This is the same production scene i have uploaded for other bug reports. I am trying to make blender 2.8 just as good as 2.79 for archviz style production scenes.



Event Timeline

Just for reference:

There is nothing in particular that seems completely wrong to me here.

YAFU (YAFU) added a subscriber: YAFU (YAFU).EditedApr 29 2019, 2:00 PM


I can reproduce the problem here on Kubuntu 18.04, It is laggy and choppy in 2.8.
i7-3770. GTX 960

My CPU graph:

GPU usage While Orbit:
2.79 - peaks of 40%
2.80 - 100%

Sebastian Parborg (zeddb) triaged this task as Confirmed, Medium priority.

The issue is the way we render the particle instances. We are iterating over all the particles and the depsgraph is taking 1/3 of the drawing time to iterate over them.

Draw Cache & Render time could also be reduce if treating particle instances as a special case but that's more complicated because a lot of the complexity is abstracted by the depsgraph currently.

Tagging @Brecht Van Lommel (brecht) & @Sergey Sharybin (sergey) maybe you have ideas on how to fix this.

In 2.79 the particle dupli stuff takes even more of the time.

So I'm not sure if fixing that performance is needed to get things back to 2.79 level, I think there is something else going on here as well.

To improve the (particle) dupli performance we should ideally cache the duplis on the object that generates them. There may also be some improvements possible in the particle code to generate them. I'm not sure we can do this for 2.80 though.

I am not 100% sure how the code works behind the scenes, but are all of the bugs I have reported in regards to performance all linked to this same code slowdown? (T62511 T62511 T58188 and this one T63925 )

Would be great to get this in 2.80, but totally understand that it may require a bit more of a rework.

The selection issue is a different bug.

This report may be the same as T58188, something in the wire overlay drawing code that is not optimal. I hope we can fix that for 2.80.

Is this still targeted for blender 2.8?

Not really, I wanted to but refactoring was too deep and would have made too many changes too close to release. It will be in 2.81. See progress made by D4997 (not even up to date).

Is it still a target for 2.81?