VRay CPU vs. GPU: Bump Map inconsistent

[JonasNöll]

Hi,

having issues with procedural bump maps between V-Ray GPU and CPU where they look very different. Is there a trick to match them so that you can switch back and forth without changing the values? They seem very inconsistent with both scale and intensity.

[vladimir_krastev]

Hi you are correct. On a simple scene the 3ds max's noise renders differently on V-Ray CPU compared to V-Ray GPU. We see differences if we use it both in diffuse and bump maps slots. The VRayNoiseTex renders the same results if we use it in the diffuse slot. However when we apply it to the bump there are differences much like the ones you mention.

V-Ray CPU and V-Ray GPU are different render engines and it is not advisable to switch back and forth between the two in the middle of production. It is best if you could stick to one of the two engines from start to finish of a project.

[Vrayfortech4home]

vladimir_krastev We know that V-Ray CPU and V-Ray GPU are different render engines and we shouldn't switch back and forth between both in the middle of production but there are other scenarios, for example, imagine that we need or want to replicate a shader that someone did in CPU mode (from a tutorial or smomething) but our machine were built to work with GPU (my case), in this case, it is very important that the bump values and noise maps behaves the same. Said that, is there something you are doing to solve this or are you planning to solve it soon?

[^Lele^]

There simply is no guarantee that the two results will look remotely the same, unless you pipe the bump through a -more expensive- vrayColor2bump texture.
See the docs, here.

In V-Ray Next and above, the texture ensures that the procedural bump is fully consistent between V-Ray 5 and V-Ray GPU. Without VRayColor2Bump there might be some difference in the rendered procedural bump amount related to the scene size and scale.

[JonasNöll]

There simply is no guarantee that the two results will look remotely the same, unless you pipe the bump through a -more expensive- vrayColor2bump texture.
See the docs, here.

Hi, actually the result is also very different when going through a VRayColor2Bump.
But alright, I get it that those are different render engines and switching back and forth is not recommended. Though for shader compatibility it would be great if the result between both looks more or less the same

[vladimir_krastev]

Hi I tested what Lele advised and it works just fine. I used VRayNoiseTex combined with VRayColor2Bump and there are no differences between CPU and GPU.



I also created an improvement task for the bump differences. It is logged in our system under VGPU-5715.

[JonasNöll]

Hi, there are differences when using Color2Bump which will you notice with higher bump intensities. I agree they are more similar but not identical, but maybe that's splitting hairs I guess going through Color2Bump could be used as a workaround for now. Is there any information on how much it would slow down the rendering compared to piping in the noise directly as a reference? So let's say there is a more complex scene with for example 100 shaders and I would use Color2Bump in all of those? Would it like slow down the rendering 0,001 / 0,5 / 2 / 5 /10 times? I know it's hard to evaluate but would be interesting to know how much the Color2Bump would affect performance and for me it's not really clear how much of a performance break Color2Bump would be.

[^Lele^]

Try with a non-procedural.
If that isn't identical, then it's a bug.
EDIT: the cost isn't that big. If i didn't say, you likely wouldn't have noticed it.