Pipet / Texture Split Performance


found a strange performance behavior using the pipet / texture split with different texture sources.
could anyone be so kind and explain me the differences here?

as soon as i disable the dynamic texture the performance increase a lot but for the renderer it does´nt make any difference if its enabled or disabled here…



PipetTest.v4p (9.2 kB)


as mentioned in other threads before some nodes performance index cannot be trusted. among them are info and pipet. the high number you see in the pipet that is connected to the dynamic texture actually belongs to the dynamic texture.

so in your first scenario most performance is lost while filling the texture. in the second scenario (with the renderer) most performance is lost by the pipet itself. this can be explained as pipeting on pool-default texture (as dx9texture is) is way slower than pipeting on pool-managed textures (as dynamic-texture is).

in both cases you should be able to disable the pipet to save performance. but here i see there is a bug in betas <=28.1. that won’t completely disable the pipet. fixed for betas>28.1.