Clear coat behaving strangely with raytraced reflections

It's really cool being able to experiment with DXR features in Unity, but I've been having issues getting metallic materials with clear coat to work and I'm out of ideas on how to solve this.

I'm using Unity 2020.2.2f1, HDRP 10.2.2, although I've also tried Unity 2021.1.0b4 and older Unity versions with the same results.

I created some screenshots to show the issue:

Normal: Without raytracing enabled, if I bake a reflection probe everything looks good and all materials behave as expected.

DXR: With raytraced reflections enabled normal materials [4] and [5] look good and show nice raytraced reflections.

If I enable clear coat then the metallic part of the material seems to stop to ray trace reflections, visible in [1], [2] and [3].

The clear coat receives ray traced reflections, but for some reason the HDRP Lit shader has these blurred (with rough material [3] being more blurred, but smooth material [2] is also blurred, which is plainly wrong.

I tried creating a Shader Graph shader with StackLit master node and enabled clear coat [1] and that works with sharp clear coat reflections on a rough metallic surface, but because the rough metallic surface isn't receiving ray traced reflections it only reflects the skybox.

If I try to fake the metallic reflections by using a reflection probe then the metallic reflections will look good for the HDRP Lit shader, but the clear coat will also receive the reflection probe, which then is sharp (which I want) but has an offset (which I don't want).

The Shader Graph StackLit shader on the other hand will not be influenced by the reflection probe, so that the metallic part will only receive specular reflections from the light source and no blurred reflections at all.

All of this actually looks like there's a system to it, but it's currently not functioning in a way that I'd say I could use clear coated material with raytraced reflections.

 

@auzaiffe have you been able to verify this against any existing stacklit clearcoat materials? I'm getting some strange results after your PR.

Before #3872


After #3872


I've tried various stacklit clearcoat materials and I'm getting the same results so I have reverted the PR locally for now.

 

I guess you need reflection probe now because base layer of metal reflecting sky now

 

Can an option be introduced to use SSR/RTR for both layers (base and clear coat) in StackLit? I think having to setup and bake reflection probes in a realtime ray tracing project is a step backwards (in both quality and workflow) and defeats the purpose.

 

It does, but not for the clearcoat shader if you specifically want realtime raytraced reflections on it (not talking about SSR fallback now which requires the probe fallback regardless).

Only use case where I could see the probe combined approach work nicely with raytracing setup if it's some showcase room / car configurator thing but if you want to render accurate reflections on car paint for vehicle that actually moves in the scene, relying on probes for one layer of the paint is just no go. HDRP doesn't have fast enough realtime reflection probes and manually placing bunch prebaked ones isn't going to work properly either.

 

Hello. Carpaint has clearcoat. Cars move.) This would help to get best in class for all these use cases.
Without i see no possibility to render correct carpaint en move.
Would be nice if you give the possibility that the user can decide to use it or not.

For following use case i could spend lots of performance to have an workable/realistic carpaint en move because raytraced reflections are the most important like you see here for car to car and environment to car reflections.

So there are at least the complete industrial automotive brach and car race gaming.