Yeah, the GPU is just a bunch of cheap, identical cores that all run the same shader code in parallel. Every frame gets split into thousands of tiny tasks that sit in a pipeline—fetch, execute, write. The trick is that each core is cheap enough that you can put dozens or hundreds of them on a chip, so even a complex shader gets enough workers to finish in time. So it’s not a mystical thing; it’s sheer parallelism and clever scheduling.

Sure, because nothing beats another layer of abstraction and a manual driver tweak for every little slowdown.

Sure, debugging at two a.m. on a dying GPU is my idea of a good time.