Unity URP vs Standard: Performance?

So this morning I tested a URP setup, dumped in a simple stylized asset kit. I did not test full build, I only eyeballed performance numbers in editor (single window, no editor camera).

The rendering loop performance dropped pretty dramatically compared to standard. 8ms vs 13ms in rendering.

Now, admittedly this was not particularly optimized or anything, I spent a few hours tweeking various settings, poking materials, etc. Nothing made a significant difference.

URPs batch count was much lower and the cpu time looked lower or similar, but the actual render time was like 30% slower. This was running in the last LTS release, I think 2020.3 or something. Both had post running, URP had no AA.

What went wrong here?

- Did I not use some preferred configuration? (I basically just dumped in a toon quality asset store kit)
- Is there a much more significant gap between editor/player performance in URP that I just missed checking?
- Do older versions (like 2020.3) just not have all the performance optimizations?
- Something else?

(Main targets: PC/console, switch)

 

More here etc https://forum.unity.com/threads/will-standard-shader-slow-down-the-game-on-mobiles.1180687/

Seems a common enough thread starter that Unity ought to answer about the common differences. Mine is usually it'll be post effects or shadow settings.

 

StackOverflowException

on post 7? Now, jokes aside. I have found that is the other way around on my work projects. So it's interesting to see people having performance issues with URP

 

From my testing I've avoid Standard Shader like the plague, kills performance on low-end devices and low end mobiles.

 

For sure.

That said, are side-by-side comparisons of the same content even relevant? I'd have thought that to get the strengths of any RP over another you'd need to understand how they work, design content to work with the strengths and avoid the weaknesses, and be working with scenes of at least a certain scale.

So my thoughts are that Unity need to give some clear info about the strengths and weaknesses of each, a set of example use cases with explanations of why each RP is stronger or weaker for each, and at what point that will be noticed.

 

As far as rendering in the editor is concerned, my experience has been the same - new RPs don't match built-in performance.

I'm very interested to see how similar scenes in different render pipelines perform in actual builds rather than in the editor. (Haven't had the time to test that out.)