Eevee displacement behavior changed #101175

New Issue

A pattern of pixels appears on the hair in the "strip" mode in EEVEE. Testing was performed on two computers
1 PC:
system - Windows 10
Video Card - RTX 3060
2 PC:
System - Ubuntu 20.04
Video Card - Geforce MX230

Behavior changed in 80859a6cb2

3.3 & 3.2 result:

3.0 result:

File with hair:
test.blend

A pattern of pixels appears on the hair in the "strip" mode in EEVEE. Testing was performed on two computers **1 PC:** system - Windows 10 Video Card - RTX 3060 **2 PC:** System - Ubuntu 20.04 Video Card - Geforce MX230 Behavior changed in 80859a6cb2 **3.3 & 3.2 result:** ![Screenshot_20220918_230827.png](https://archive.blender.org/developer/F13508178/Screenshot_20220918_230827.png) ![Screenshot_20220918_230830.png](https://archive.blender.org/developer/F13508180/Screenshot_20220918_230830.png) **3.0 result:** ![Screenshot_20220918_230746.png](https://archive.blender.org/developer/F13508184/Screenshot_20220918_230746.png) **File with hair:** [test.blend](https://archive.blender.org/developer/F13508187/test.blend)

Added subscriber: @sanek2005

Added subscriber: @lichtwerk

Something going on with the normal map of the .Human material {which i also on the hair}.
Think this might end up being a duplicate of some other report about changing behavior of normalmaps, will dig further...

Something going on with the normal map of the `.Human` material {which i also on the hair}. Think this might end up being a duplicate of some other report about changing behavior of normalmaps, will dig further...

Added subscriber: @fclem

Behavior changed in 80859a6cb2

And I have to correct myself, it is not the normalmap that is the issue, but the displacement node instead.
I am a bit confused here, because prior of said commit, I dont see the node having an effect at all (even at other values).
After said commit it does have an effect, but it does not seem to match cycles at all (in terms of strength).

Like I said, this could all be working as intended after all, but would like @fclem to check.

Behavior changed in 80859a6cb2 And I have to correct myself, it is not the normalmap that is the issue, but the displacement node instead. I am a bit confused here, because prior of said commit, I dont see the node having an effect at all (even at other values). After said commit it does have an effect, but it does not seem to match cycles at all (in terms of strength). Like I said, this could all be working as intended after all, but would like @fclem to check.

Changed status from 'Needs Triage' to: 'Confirmed'

@fclem: feel free to lower prio again (or close if appropriate).

Added subscriber: @pragma37

The change in behavior seems to be caused because before 80859a6cb2, the Normal Map node used the Strength input to interpolate between the smooth surface normals and the normal map, but now it interpolates between the “displaced” normals and the normal map.

Since the user incorrectly plugged the normal map color into the Strength input, now the displacement is actually visible.

Whether the change in behavior is an actual bug, I guess is up to @fclem to judge.
In Cycles, it seems like the Normal Map is applied ~~on top of the smooth normals~~ the same way as EEVEE does now.
But then the BSDF input Normal is applied on top of the “displaced” normal, while EEVEE replaces it.

The change in behavior seems to be caused because before 80859a6cb2, the Normal Map node used the Strength input to interpolate between the smooth surface normals and the normal map, but now it interpolates between the “displaced” normals and the normal map. Since the user incorrectly plugged the normal map color into the Strength input, now the displacement is actually visible. ![imagen.png](https://archive.blender.org/developer/F13609703/imagen.png) Whether the change in behavior is an actual bug, I guess is up to @fclem to judge. In Cycles, it seems like the Normal Map is applied ~~on top of the smooth normals~~ the same way as EEVEE does now. But then the BSDF input Normal is applied on top of the “displaced” normal, while EEVEE replaces it.

Added subscriber: @Emi_Martinez

Added subscriber: @ThomasDinges

@fclem Hi, can you please check on this report? Would be good to solve or close if the behavior is intended,

After extensive testing, the only thing I see that is different from cycles is the hair support. The rest of the behavior seem to be consistent.

In cycles, bump displacement is just disabled for curves. In EEVEE it is glitchy. I'll fix this discrepancy and close this report.

After extensive testing, the only thing I see that is different from cycles is the hair support. The rest of the behavior seem to be consistent. In cycles, bump displacement is just disabled for curves. In EEVEE it is glitchy. I'll fix this discrepancy and close this report.

This issue was referenced by b2746876f2

This issue was referenced by b2746876f2883798cc3487dcda2ba83e92f0d29b

Changed status from 'Confirmed' to: 'Resolved'

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Eevee displacement behavior changed #101175