Leed el texto de abajo sacado de http://www.hardocp.com/article.html?art=NDI4LDM=
Analizemos 2 frases importantes del asunto:
1-. Here is the kicker, if using Pixel Shader 1.4 it will render approximately 150,000 polygons, but if it has to utilize Pixel Shader 1.1 then it will render approximately 250,000 polygons.
Significa que si la targeta tiene la capacidad de usar Pixel Shader 1.4 (actualmente ATI 9500 o superior) el Benmarch (básicamente pq le sale de los cojones) pone aproximadamente en juego 150.000 poligonos, mientras que si usa Pixel Shader 1.1 (básicamente TODAS LAS targetas excepto esas dos), pone (tb pq le sale de los cojones) un total de 250.000 poligonos... Recordad cuantos más poligonos menos rendimiento... hasta hay llegamos todos... y la diferencia de poligonos es ENORME con lo que es normal esta JODIDA diferencia de puntuaciones... Y referente a la calidad de imagen, no es apreciable con un número TAN ALTO de poligonos, la diferencia entre 150.000 y 250.000 con lo que si ambos metodos usaran la MISMA cantidad de poligonos las puntuaciones por mucho DirectX 9 o 8.1 (recordemos es el que se usa en los test 2 y 3) no serian las que salen... Simplemente aprovechan esa capacidad para detectar si poner o no más o menos poligonos... pero recordad, los juegos... USEN 1.4 o 1.1 de Pixel Shader, pondrán la MISMA cantidad de poligonos... no os ENGAÑEIS...
2-. Respecto al otro os lo dejo q traduzcais vosotros es lo mismo pero más descarado aun: If Pixel Shader 1.4 is being used then approximately 280,000 polygons are being rendered, but if it falls back to Pixel Shader 1.1 then it jumps up to 560,000 polygons.
Vamos que hace el test haciendo que las 9500 o superior muevan 280.000 poligonos mientras las demas targetas deven mover 560.000 poligonos... Igual que que en el test 2 a santo de nada... ya que los juegos(motores muchos de ellos todavia en desarrollo) no cambian la cantidad de poligonos utilicen pixel shader 1.1 o 1.4.
Total: No se si con buena o mala intención o con dienro o no de por medio SIMPLEMENTE en el Test ATI 9500+ o superior mueve un 45% menos de los poligonos que deve mover las DEMAS targetas...
Són matematicas... simples las que te hacen ver que si las demas targetas tuvieran los mismos poligonos a mover que estas con el PS1.4 la diferencia de puntuaciones seria similar a la mostrada en el 3DMark 01
Test Número 2:
Game 2 - Battle of Proxycon
This game test is geared for DirectX 8. As soon as you see it run you will get the feeling of the Doom]|[-esque FPS style of game. This game test completely uses shaders and a DirectX 8 card is required to run it. Vertex Shaders 1.1 handle all of the vertex processing. This game test can use Pixel Shader 1.1 or 1.4 where supported in hardware. If a card capable of 1.4 pixel shader, it only needs to do one pass for depth buffer setup and a single pass per light source. But if it does not detect a PS 1.4 card then it uses PS 1.1 which uses one pass for the depth buffer but three passes per light source. This test does use normal mapping to make it look better despite having a low polygon count. 128MB of video ram is needed to run this test without texture uploads during test execution; 80MB for textures, 6MB for vertex buffers and 1MB for index buffers.
Here is the kicker, if using Pixel Shader 1.4 it will render approximately 150,000 polygons, but if it has to utilize Pixel Shader 1.1 then it will render approximately 250,000 polygons.
This test does use the stencil buffer to perform shadow volumes "kind of like" Doom]|[. However, its method is not exactly the best way to do things. Their algorithm is very redundant, there is a portion called “Skin Object in Vertex Shader” which basically does the exact same skinning calculations over and over for each object. It is an inefficiency that would not be translated to a real world game such as Doom]|[. Because of the way their algorithm works the vertex pipeline becomes a huge bottleneck.
One other issue about this game is the fact it uses Pixel Shader 1.4. I can’t think of one game that is out there that uses Pixel Shader 1.4. The fallback is to Pixel Shader 1.1, why it wouldn’t fall back to Pixel Shader 1.3 instead is beyond me.
Test Número 3:
Game 3 - Trolls' Lair
This game test is based on DirectX 8 as well; in fact it uses the same range of vertex and pixel shading as Game 2 test. One of the most impressive things in this benchmark is the women’s hair which each strand has force calculations applied to every vertex so that it waves and moves in real time depending on her movement. The rest of this benchmark is based on the same Pixel Shader 1.4 used, if not then fallback to Pixel Shader 1.1 again. 128MB of video memory is needed to run this smooth. If Pixel Shader 1.4 is being used then approximately 280,000 polygons are being rendered, but if it falls back to Pixel Shader 1.1 then it jumps up to 560,000 polygons.