I’ve read a lot of information about these techniques and nobody seems to be able to reach an outright factual answer to whether we should use them or not.
For example: Should I just make every texture I have that is 2048 res and higher into a Virtual Texture?
And should I turn my landscape material into a Runtime Virtual Texture material, and turn every assets material that needs to blend into it also a Runtime Virtual Texture? Like rocks and trees and things…
Is the overall performance better with them, or worse off?
Someone said that the extra texture sample per texture when using Virtual Texture Streaming method has negligible performance hit on the CPU but decreases the VRAM usage significantly.
Yet I can’t see any evidence of this and my small environments don’t hit hard enough for me to notice any performance differences when testing.
Does anyone know for absolute sure when to use these and why please?