Hey team. In terms of the actual implementation, how do these two techniques differ? Is the main difference just that VSM renders all shadow maps into a single atlas?
The main difference is the fact that VSM can handle super high poly meshes under 4ms while the same scene with high poly meshes(like nanite level) with regular Shadow maps will cost around 11ms
But here’s something interesting. Lets take City Sample, In editor, say you load up about half a mile by half a mile of the city. VSM’s will cost under 3ms!
BUT, when you play in editor, turn off all AI(get rid of that CPU bottleneck), it cost around or above 4ms? From what I’ve read, it’s supposed to cost more with dynamic objects since it had to recalculate moving shadows at an expensive ms cost.
I’m personally trying to figure out how to get Native 1080p City Sample with no Ai, No AA, Low PP, High Lumen, reflections, and effects, with medium shadows to run VSM under 3ms, I can’t seem to figure out how to scale VSM’s down?
To read more just go to the VSM 5.2 documentation here
Virtual Shadow Maps in Unreal Engine | Unreal Engine 5.2 Documentation.
Highly informative: Virtual Shadow Maps in ‘Fortnite Battle Royale’ Chapter 4 - Unreal Engine