Is there any use in pooling PBO resources, even though they are orphaned? Or can I as well glGenBuffer, glBufferData, glDeleteBuffer when needed without much performance overhead compared to pooling?
Well, let’s think about this.
The best case scenario means that you will get no better performance than you would with pooled buffers. And in the worst case scenario, you will get a lot worse performance.
I’d suggest sticking with the guaranteed performance that will work fine on any hardware, as opposed to the relies on of how fast frequently allocating and deallocating buffers is, which is platform-and-driver specific and therefore unanswerable.
Sounds reasonable. The only disadvantage is that pooling PBOs might lead to unnecessarily high memory usage.