Performance question regarding sampling the depth texture vs tex2d

Hi guys,

If a scene depth texture is already being generated, is sampling that depth texture in a shader any more expensive relative to sampling a regular texture 2d?

If so, why and how significant is the difference?

Platform - mid range iOS/Android
URP

Thanks in advance!

 

It depend on the hardware, and the pipeline implementation, so there is no straight answer. Some tile based GPU on mobile generally don't like sampling the depth, some like the new mali GPU have pixel storage cache: https://community.arm.com/developer...og/posts/pixel-local-storage-on-arm-mali-gpus But the pipelines implementation have to take advantage of that by not flushing the memory between pass. SO it's not a straightforward question.