I’m just wondering if this is a GPU limitation when it comes to Alpha Contribtion. Usually when using the default shadow-catcher workflow, alpha contribution is set to -1 and the alpha is as expected (however it takes rather long until the alpha gets completely noisefree, therefore I’m usually setting the Alpha Contribution to 0 instead when using CPU rendering and I get the identical beauty-render and a nice alpha to knockout all the elements from the shadow-catcher and in post I can set the background to multiply and have the shadow applied to the background then. Alpha Contribution set to 0 on GPU gives a quite different result and I’m wondering if this is a technical limitation of V-Ray RT GPU
We are aware of this limitation, sorry for the inconvenience. The problem with the noisy alpha and matte is resolved in V-Ray Next.