Thanks for the great work of the developers. Here are some tests and comparisons of the principled v2 branch. Soap bubble scene. Real-world soap bubbles rarely seem to be thicker than 1000nm.
Cycle: filter glossy: off, noise threshold: off, clamping: off, 400 samples
GJettG shader 500nm 33s:
Where did you download it? I missed @ git.Blender.
I’ve been following the thread and it’s super-exciting work. Nice to see some tests pulled out here so other people can see how amazing the new developments are. 
Hi! The developer has now completed the build, you can go to blender’s branch category to download.
The picture shows the effect of different options of Principled V2:
Here is my test. I used AGX color transform and a color balance node just to increase contrast. Thin film Thickness is 1.17 micrometers and noise textures driving the thin film IOR.
And now with lower film thickness 569nm. Sunset HDRI, DOF, filmgrain etc.
Just did a comparison between an iridescence shader I created a few years back in the thread linked below - that uses overlapping offset RGB sine waves - and compared it to the principled V2 shader.
I wasn’t too far off the mark (my shader is on the left).
The colour bands are broadly correct and in the right order.
Here is a comparison of the Principled V2 branch with other metal shaders from different versions of blender. Luxcore and Cycles have very different lighting and shader parameters, so I just tried my best to make their renders look as close to the same. For Cycles and Luxcore, diamonds have the same parameters, both with an IOR of 2.3818 and a dispersion of 0.01. There seems to be a problem with the refraction shader of the blender in the spectral branch, its red dispersion is too vivid. Having said that, how is the spectral branch doing now？ It seems like it was a long time ago when it was last updated.
None of the following images have noise reduction or toning.
Since the difference in the effect of the metal in the picture is relatively small, if you want to see the difference carefully, you’d better download them, put them in a software such as PS, and quickly switch back and forth between layers to better perceive the difference . If you’re too lazy to do that, the metallic color of the spectral branches is cleaner than the non-spectral ones. The metal edge of the Principled V2 branch is similar to that of Luxcore, which is different from the old principled BSDF. The rest of the differences between Luxcore and Cycles are likely due to lighting and ground material issues, so I won’t comment too much.
Spectral branch, with spectral rendering turned on, 8000 samples 5min19s:
So not just film interference, we are having even dispersion in principled v2?
The dispersion in Cycles’ rendering is implemented with a shader I made myself, and the dispersion in Luxcore is implemented by Luxcore’s default glass shader. The developers of Principled V2 say a dispersion shader might be added, but not sure yet. The dispersive shader seems to have a lower priority in the developer’s mind, but the dispersive shader is a relatively simple shader, at least simpler than the functions developers are currently doing. So I think in the later stage of the development of the Principled V2 branch, Cycles will most likely have its own dispersion shader, and the shaders made by developers are often several times faster than the shaders written by ourselves.
So is the spectral cycles dispersion shader is the same user made one?
Yes, all Cycles-rendered images use the same shader, and the refraction shader of the spectral branch has a problem, resulting in too brightly dispersed red light.
