@OmegaSupreme (View Post)
Yeah... That wouldn't help at all :)
First-person shooters (FPS) have the advantage of having lots and lots of surfaces (floors, walls, hills, cars, ...), for most of which the process of rendering can be silently pushed to the graphics card.
(So the programmer says "There's a wall right here and there's a dimple in the wall here and there's another wall here. This is what the walls have to be plastered with and this is what the dimple has to be plastered with. Now go, do your work.")
The graphics card happily takes all those instructions and does them silently somewhere out of sight because those can be done independently from what the CPU thinks. (Stuff like turning is just told to the graphics card with glTranslate() and the graphics card figures out the rest)

Here's where the fundamental flaw lies in the very system of TPT. If it had anything to off-load to the graphics card, it would've already been done, honest.
It's just...
there's nothing to be off-loaded.
At the very least, things like rendering the thumbnail images or the look of the legs of a stickman can be. Those will certainly benefit from being done separately and let's say 15 at once.
But particles...
They all interact with each other, they need to constantly know about the entire playing field and thus can't be sent to the background of an entirely different chip's mini-processor.

You see, graphics cards in practice are not much more than bundles of thousands, even millions low-speed processors (which are really good at doing lots of separated stuff at once, like factory workers, each with their own product to finish, so they happily do it but at a slower rate than a conveyor could)
What TPT needs is a super-fast conveyor and factory workers with the efficiency of a thousand [insert battery company] bunnies.

_{correct me if i'm wrong which I probably am!}