Problem with denoising on multiple devices is that there are visual edges between the parts.
So my proposal is, that we can select one device, that should do the denoising after all tiles are rendered.
Maybe it would be a nice addition to add a Composition node for crowdrender.
This would allow us to automatically define that denoising should be done afterwards.
Hi @BiggA94 the denoising node in the compositor will work as you correctly deduced :). Once Crowdrender has rendered tiles on each node, they are merged back into a single image which then passes to the compositing stage where denoising can be applied using the intel denoising node I believe.
We're actually going to look at a different render mode for Blender's other denoising algorithm to stop artefacts on edges as you discovered. Put simple, nodes currently use tile splitting, but we're looking into having four splitting modes in the rebuild of our addon. Two of those will be compatible with the cycles denoising algorithm.
Just experimented a bit. Correct me if I'm wrong, but if i start a render with crowdrender, only the "Render Layers" node of the composition is executed remotely?
So it's working, that I can do the denoising with the rendered image after that :)
But I don't know, what should be the best in case of animations.
Here, a composition node could still help (one CrowdRender-Render node and one CrowdRender-Denoising node).
(Just thinking loud :D)