Page MenuHome

Performance regressions
Open, Confirmed, HighPublic

Tokens
"Love" token, awarded by dgsantana."Love" token, awarded by xdanic."Dat Boi" token, awarded by hitrpr."Love" token, awarded by razgriz286."Orange Medal" token, awarded by Zino."Love" token, awarded by hadrien."Love" token, awarded by Hickz."Dat Boi" token, awarded by shader."Love" token, awarded by zhruith."Love" token, awarded by avpiedra."Love" token, awarded by lcs_cavalheiro."Love" token, awarded by staughost."Love" token, awarded by Krayzmond."Love" token, awarded by 0o00o0oo."Love" token, awarded by aliasguru."Love" token, awarded by ixd."Love" token, awarded by ace_dragon."Love" token, awarded by Astiero."Love" token, awarded by TheRedWaxPolice."Love" token, awarded by dpdp."Like" token, awarded by elbox01."Love" token, awarded by zappedfish."Love" token, awarded by Alrob."Love" token, awarded by amonpaike.
Authored By

Description

List of performance issues that we would like to fix or investigate for 2.80:

  • Transforming vertices in edit mode on high poly meshes is slow, due to overhead of depsgraph, mesh updates and and batch creation. Verify if there is anything unexpected happening here.
  • BMesh to mesh conversion was happening all the time at some point, is this still the case? See check_rendered_viewport_visible, it now always happens when a 3D viewport is visible, and could be quite problematic for edit mode performance.
  • Verify if custom data is only being preserved through modifier stack when needed (vertex colors, vertex groups, UVs)
  • Manipulate modal callback performance may be problematic: T55458

Details

Type
To Do

Event Timeline

There are a very large number of changes, so older changes are hidden. Show Older Changes
rndmnm added a subscriber: rndmnm.Nov 28 2018, 12:51 PM
This comment was removed by rndmnm.
Brecht Van Lommel (brecht) raised the priority of this task from Normal to Confirmed, High.

Funny, turning off overlays increases performance in edit mode a little bit, why is that?

Because there's fewer things to draw then, that's expected.

Looked into Custom Data layers.

Layers which are dependent on am modifier stack (like deform groups) are not preserved if they are not needed. So this part is good.

However, the UVs and vertex colors are always preserved. This is actually the same in Blender 2.7, so i see potential being more memory efficient than 2.7.

Hi.

Is there any reason for the decision to not fix these regressions for the 2.8 final release? Unless the solutions are more complex than we users think they are, it seems like an unforced error that really dampens what would otherwise be a truly major milestone.

Further performance improvements are more complex than you think, that's exactly the reason they will be for later releases.

I´ve tested the different render possibilities in blender and compared blender 2.79b with blender 2.80 5f140e61c28c.
It´s the same scene in blender 2.79b and blender 2.80.

Blender 2.79b

CPU (AMD Threadripper 1950X 16Cores)
bmw27_cpu.blend (tiles 32x32) 01:20.53
bmw27_cpu_denoise.blend (tiles 32x32) 01:24.32
2xGPU (Geforce Gtx 1080 Ti 11GB / Geforce Titan X 12GB)
bmw27_gpu.blend (tiles 240x180) 01:09.15
bmw27_gpu_denoise.blend (tiles 240x180) with denoise 01:12.15
GPU (Geforce Gtx 1080 Ti 11GB)
bmw27_gpu.blend (tiles 240x180) 02:01.35
bmw27_gpu_denoise.blend (tiles 240x180) with denoise 02:05.07

Blender 2.80 5f140e61c28c

CPU (AMD Threadripper 1950X 16Cores)
bmw27_cpu.blend (tiles 32x32) 01:46.65
bmw27_cpu_denoise.blend (tiles 32x32) 01:51.20
2xGPU (Geforce Gtx 1080 Ti 11GB / Geforce Titan X 12GB)
bmw28_gpu.blend (tiles 32x32) 00:34.24
bmw28_gpu_denoise.blend (tiles 32x32) with denoise 00:40.69
GPU (Geforce Gtx 1080 Ti 11GB)
bmw27_gpu.blend (tiles 32x32) 00:49.69
bmw27_gpu.blend (tiles 32x32) with denoise 00:55.06
CPU+GPU (AMD Threadripper 1950X 16Cores + Geforce Gtx 1080 Ti 11GB / Geforce Titan X 12GB)
bmw28_gpu_cpu.blend (tiles 16x16) 00:30.40
bmw28_gpu_cpu_denoise.blend (tiles 16x16) with denoise 00:39.05

Summing up blender 2.79b renders the scene on cpu ca. 25% faster. Why?
Compared one GPU to two GPUs in blender 2.79b there is a 75% boost with two GPUs.
In blender 2.80 the boost is low, only 44%. Why, what is slowing down the second GPU? Blender 2.8 is also using a newer CUDA version.
GPU rendering in blender 2.80 is more than twice as fast as in blender 2.79b!
I couldn´t find any significant slowdown with denoise.

here is another bizarre and interesting phenomenon …
the first ship is composed of many objects but all edited simultaneously … the second ship is a single object …
notice the differences …

here is another bizarre and interesting phenomenon …
the first ship is composed of many objects but all edited simultaneously … the second ship is a single object …
notice the differences …

Looks like you have proportional editing turned on, which doesn't apply between objects even if they are edited simultaneously. The additional calculation of proportional editing for the entire ship vs. one part could be slowing it down.

Looks like you have proportional editing turned on, which doesn't apply between objects even if they are edited simultaneously. The additional calculation of proportional editing for the entire ship vs. one part could be slowing it down.

It was a test to make bottlenecks more evident on how a single object with a high number of polygons is currently structured.
Proportional editing turned on has nothing to do with it, it is enabled to make this slow down clear compared to when multiple objects are being edited simultaneously.

it is a test done on purpose to perhaps inspire some enlightening mind that can find a solution.