nVIDIA GeForce256 and T&L
T&L, part 1: Transformations and GPU
Transform and Lighting (T&L); this is what will do now the flaming graphic card that we have bought to ourselves for an eye of the face and part of the kidney (a lot of money for two words). The idea is simple: to liberate to the mike of the PC on behalf of the work, which falls down in the ultra-specializing chip of the graphic card.
The transformations are necessary because we have a world 3D that he lives on our hard disk in the shape of game, and in the same way that in the planet we have Earth coordinated (with his origin 0,0 in the Gulf of Africa, in half of the Atlantic Ocean), in this world 3D coordinates exist, with 0,0 you are going to know where.
But the problem is that we will play this game, and will handle a personage, and will see this world from a "virtual eye", or camera, where from we will have another coordinates origin to be able to calculate the lighting (the programs 3D illuminate the world 3D throwing beams from our eye towards the rest of the world). To finish of arranging it, we will have to do this world from our screen, which, in addition to having another coordinates system, it turns out that it is in 2 dimensions (at least until they do screens 3D with holograms like those of the movies).
Then: what is a transformation? Simply these changes of coordinates. All the objects move, change size, turn in any direction... to calculate the new position of every point of the object a transformation counterfoil is applied to the initial vector of points, which gives like turned out a final points vector:
So if all these operations (16 multiplications and 12 sums) our processor already does not do them, it will be the whole relief. Well, with the technology T&L now we have a graphic card that does all this, I joust the same that the most recent mikes try to improve, the AMD with his instructions 3DNow! or Pentium III with the SSE.
The fact of being provided now with a chip dedicated specially to these tasks (the GeForce, named sometimes GPU, Graphics Processing Unit) is going to allow to increase the speed so then these calculations are realized, and also it will be able to draw (=calcular) many other polygons. That is translated in a very top image quality, as you can observe in these drawings:
The lower "donut" (a bull, being thinner) is clearly better that the donut of above; if renderizamos and we apply textures, the difference diminishes, but it keeps on being quite visible. The ideal thing will be to be able to draw figures like the lower one, but that is not so easy...
How much difference believes that there is between a donut and other? Do his bets... Well, the answer is that the lower donut has 10 times more triangles that of above, scarcely 2048 triangles, and obviously our computer is going to take 10 times more in drawing it. If this donut is one of 4 tires of our car, and there are several cars on screen, we can already have a rapid mike, a graphic card with T&L... or if not, more it costs to be satisfied with another donut (certainly, right now I am hungry that Homer Simpson).