Is it me or Cycles Bump map does not affect lighting of a material properly?

Found this post here:https://projects.blender.org/blender/blender/issues/86648

These images are excerpts from the post:
Blender Cycles:

Luxcore:

Notice how more realistic the bump map looks in Luxcore (I am not sure if its the same bump map used in both) the scratches seems to affect the lighting on the mesh better and catches the highlights on the surface. Cycles looks very flat.

Try turn down your Displacement strength value to something like 0.1

It’s hard to say that the two are the same. :thinking:
The highlight sizes are not the same at all.

Add

If you look closely, small embossing is better represented on the blender.
However, in Luxcore, only scratches are strongly expressed.

This may be a problem with the settings, but it doesn’t seem to be the rendering quality.

1 Like

Maybe it’ll explain more…

I was using the normal slot of the material, rather than the displacement node.

Yeah, but if the bump map strength was increased in Cycles, like the scratches, shouldn’t both sides of the edges in the scratches be highlighted since the material has a glossy shine to it? I think this is possible in other renderers :slightly_smiling_face:

Thanks, I changed to the Displacement node with material displacement set to bump only. Its definitely looks better but I am noticing white lines when I increase the strength above 0.1 and need a higher strength. :slightly_smiling_face:

Disregard luxcore for now, and temporarily mute the bump node. Instead use displacement node and subdivision with adaptive enabled. In material settings, switch Displacement from Bump Only to Displacement only. Adjust Displacement Scale until you get in the ballpark of where the bump should be. This value is what you should be using for Bump Distance, and leave strength to 1. Mute Displacement, unmute Bump, and remove Subdiv modifier. This works well for me with procedural heights, don’t see why it should with image based. Point is, bump strength does “weird” things, shaves off values or whatnot, whereas adjusting distance gives - for me at least - a more predictable result. For things like scratches, even 0.001 (= 1mm from black to white) may be too much, maybe use a math/divide (n/1000) node to finetune it better.

3 Likes

I think luxcore has a strong metallic feel.

The image below is metallicity enhanced in cycles.

Each rendering software may have its own characteristics, but I don’t understand the completely different results.

1 Like

@CarlG Thanks. That was very helpful. Noticed a lot of improvements in the bump appearance. :+1:

@oo_1942 You might be right. It definitely does. But I do feel even matte surfaces with bump map in renderers, stroke edges are suppose to catch the light kinda like beveled edges of the scratches but that’s probably me paying too much attention to details.

Thanks for the vid, that’s very interesting. Definitely demonstrates how different the renderers are.

1 Like

you should make your own demonstration using your own bump map so you can be sure

1 Like

Other people swear by the exact opposite and either can’t tell me why or I’m too dumb to understand their explanation. These look the same to me:
image

Perhaps related to this unsolved old reported issue:

Good idea. I tried it and think the bump map looks better and more realistic in Luxcore even for matte materials. Notice how it affects the lighting details of the models surface better. In Cycles, it looks sort of flat. I used the Disney material for Luxcore and principled shader for Cycles. I didn’t touch any other shader parameter except for the bump settings which I tried to get similar (not a pro in doing renderer comparisons btw)

Cycles:

Luxcore:

@YAFU
I think you might be on to something here. Why is this being ignored by Blender devs though? Its been 3 years the issue was posted. At least a response from one of the Blender devs would have been nice or they aren’t aware of this issue?

I am aware the principled shader has been modified in 4.0 so maybe it is better now with bump maps?

Would appreciate more comparisions between Luxcore and Cycles for bump maps. I am sure someone can make a better comparison than the one I did.

would be better if the cameras matched


Then there is also the issue of light leaks caused by either bump or displacement as bump. At one point I think it happened with both but now it only happens with one of them.

That’s true. Done.


Voronoi color outputs a sharp transient which is a horrible way to do a comparison. At least try something will well defined gradients. Here is yet another one:


I’m pretty confident saying adjusting strength instead of distance doesn’t come close to ground truth displacement, whereas setting bump distance to the same as displacement scale gets pretty close. Other than self reflection and shadowing for obvious reasons, but not much of an issue in this example.
Blend file attached:
BumpVsDisplacement.blend (156.4 KB)

Who swears the opposite? All I’ve seen are most tutorials getting this wrong, mostly by ignorance, and I believe it’s because the strength slider is more convenient to adjust.

3 Likes

@CarlG Would you by any chance know a way to convert a bump map to a normal map in the material editor and also be able to convert the resulting normal to direct X without using Blender’s baking feature?

I am using the Bake-A-node addon here:

And I find it faster and more convenient to use than baking from Blender.

edit:

Would this be correct? Would also using an invert node for the Green channel be accurate?

Sorry, don’t know, but I’m pretty sure I’ve seen it done. Give it a search. Since I’m usually dealing with “random style details” and don’t need the additional benefits of normal maps, I prefer bump maps as they offer me better flexibility; easy to mix, compatible with triplanar mapping, debug’able with microdisplacement and so on. But that’s for what I do.

1 Like

@melvi You might want to take a look at this thread:

https://blenderartists.org/t/how-do-you-convert-the-world-normal-to-tangential-normal/1242210

1 Like

@CarlG Thanks for the info.
@Zebrahead Thanks.

1 Like