Just in case anyone stumbles across this thread --
I have had success at eliminating the problem by clamping my normal maps between .001 and .998 in their respective materials. I can confirm the problem only ever seems to happen in DX11 mode like mAIkAv!An said.
Technically you should only clamp the blue channel, red and green need the full [-1,1] range.
However this will only remove NaN/white dots but artifacts (=black spots) remain. That's why I prefer to mask the red+green channel and simply append a scalar of 1 afterwards, at least with only little color information in the blue channel. This is also cheaper on instructions.
For those of you who have this issue are you making changes to the DiffusePower by any chance ?
If I take another output from my diffuse texture sample and connect it to DiffusePower it recreates the problem a little and if I take the B channel output from the diffuse texture sample it's covered in white spots.
I am not convinced that it's a normal map problem, calls for more tests because it's only happening on some textures at my end. It could be a levels problem in the diffuse map aggravated by the normals.
, calls for more tests because it's only happening on some textures at my end.
Yes depends on the texture, I guess it's less of a problem for flat normal maps. You can however use sharpened MipMaps (Texture Properties -> Mip Gen Settings -> Sharpen10) to exaggerate artifacts for testing purposes.
Replies
yes I have been looking it up and found a workaround:)
cheers
I have had success at eliminating the problem by clamping my normal maps between .001 and .998 in their respective materials. I can confirm the problem only ever seems to happen in DX11 mode like mAIkAv!An said.
However this will only remove NaN/white dots but artifacts (=black spots) remain. That's why I prefer to mask the red+green channel and simply append a scalar of 1 afterwards, at least with only little color information in the blue channel. This is also cheaper on instructions.
If I take another output from my diffuse texture sample and connect it to DiffusePower it recreates the problem a little and if I take the B channel output from the diffuse texture sample it's covered in white spots.
I am not convinced that it's a normal map problem, calls for more tests because it's only happening on some textures at my end. It could be a levels problem in the diffuse map aggravated by the normals.
below is the blue channel from the texture in the first 2 pics only here it's not tiled
Yes depends on the texture, I guess it's less of a problem for flat normal maps. You can however use sharpened MipMaps (Texture Properties -> Mip Gen Settings -> Sharpen10) to exaggerate artifacts for testing purposes.