Well, the article basically examines the pros and cons of a programmable GPU or a fixed hardwired chip. a GPU will have a ability to excute different code/procedures, just like your CPU can do many different tasks, this gives you the benefit of taking advantage of specific situation and written codes that's fully optimized for it. Generatlly it's more versatile. It's disadvantage is it might be slower then a specific hardcoded(inflexible) units, since the hardwired unit can only performance the tasks it's designed to do, it can perform faster(but in special cases that's not true either).
then the article talked about explicit and procedural rendering.
let's say you want to render a sphere, in explicity mode you'll
need to define every dot on the surface (it's coordinates), depends
on the resolution, it could be quite a lot of data. in procedural
mode, you could just define where's the center of the ball, and it's
radius, and it'll be enough for GPU to generate the ball for you.
the good thing about prodedural rendering is that it reduces the
data size and bandwidth significantly. but the down side is that
not every situation is describable.. where explicit mode is more
suitable for the job.
couple examples for this, Corel Draw (Vector/procedural) vs. Painter (bitmap/explicit) and N64(Vector 3D mostly/prodecural) vs PSX (Bitmap/explicit, most of the time anyway)
I hope I didn't confuse you more... 😛
