I've recently discovered that a lot of performance issues are caused by overloading the GPU Cache that Vmix loads up with inputs like HTML sources and video inputs and I'd like to know a little bit more than what is described in this knowledge base post.

Specifically I'd like to know what actually influences the required amount of cache on the different inputs and what amount the different inputs utilize.

I'd also like to know if there are differences between the different graphics cards and if so, what amounts/speed the cards have and if this cache is referring to the GPU L2 Cache (which doesn't seem likely as it's not accessible)

Are there differences between AMD and Nvidia cards?

Are there differences in the accessible amounts of cache between Quadro and Geforce / RTX cards?

Are there differences within the actual GPU generation, eg GTX 1080/1070Ti/1070 are all built using the same GP104 die while the reference GTX 1060 uses the GP-106 - are there cache differences between them and between the cards themselves even of the same GPU die variants?

It's just really unclear what impacts this number and I'd like to know which card offers the largest cache.