I stumbled upon the latest blog post on rendering, and then on the HDRP Forward vs Deferred documentation page, and was quite surprised that Forward now seems to be the better looking shader mode: “If performance is not so important for your Project, use Forward rendering mode for better quality rendering.”
In built-in, deferred offered the best shadow fidelity and had much less limitations than forward. So my first question is, is forward the recommended mode now in HDRP when the goal is to have the best visuals?
My second question is, I tried to toggle on the MSAA at 2, 4 and 8X. That does not seem to reduce any aliasing in my game view. Is there anything to be known about setting up MSAA except toggling it on, plus setting the mode to Forward?
I’d suggest improving the tooltip for the shader mode. If forward does indeed look better, it should be told as many experienced Unity developers would intuitively think the opposite.
I dunno about that, going for deferred everything in this industry has made many things worse visually and people have needed less ideal workaround to deal with some very basic things like translucent materials. There’s a reason why many still do forward passes on deferred rendering.
One of the main benefits from deferred used to be being able to handle more light sources but that’s gotten better with more modern forward renderers.
That being said, there are practical limitation now with DXR coming into play and Unity only supports DXR with deferred path in HDRP, making this even more mixed bag when you try to achieve the best visuals from it.
Oh, that is good to know, but there is nothing preventing us from building our game with support for both and using Deferred when enabling DXR.
Our current scene has 50 lights (only one directional) and moving to forward doesn’t seem to affect performance that much. But Unity claims it opens up MSAA support and better shadowing, which we can’t see in our context for some reasons.
One thing to note about current HDRP and MSAA is that there are currently bunch of things that don’t currently work when you use it, SSR coming into my mind immediately.
The return of forward rendering doesn’t quite mean the return of MSAA. A lot of effects need access to data like depth and normals. Producing and consuming those buffers with MSAA is very expensive. Bandwidth-wise, it’s the same as using a buffer with 2X/4X the amount of pixels.