View Single Post
Old 11-22-04, 04:25 PM   #44
Registered User
Join Date: Jul 2004
Posts: 575
Default Re: Official HardOCP HL2 benchmarks

Originally Posted by jbirney
I am talking about 2003 (about may) when a few sites got a chance to benchmark Doom3 earliy.

Why are you talking about a benchmark of a demo? ATi's are not of a demo we are talking about the full final benchmark releases

Originally Posted by jbirney
Thus if it was ok for NV to have all the time in the world to work on Doom3 and not ATI last year, turn about is fair play no? .
Actually ATi blew its chance when they leaked alpha, if I was making software and a company I was showing it to leaked let be accidental or not I wouldn't trust that company anymore. And as you recall ATi now how has the game they aren't fixing their drivers to deliever better performance but doing shader replacement instead of fixing whats wrong. But from Carmacks view I'm sure Doom 3 was made with alot of communication with ATI, If you remember the 9800 and 5900fx perform very close in all situations on that game. As for all other Ogl programs too. Its not just Carmacks engines don't be blind and make a statement that the ID's engines run better because nvidia has optimized for them this is not true to any degree, you know that.

Originally Posted by jbirney

We have differnt timedemos recorded by a few different sites and they all seem to back it up ATIs demo and you still think the are biased?
Originally Posted by jbirney

Dave over at Beyond3d has a possible reasoning:

Which seems pausable.

Not saying they are biased I'm saying something else is going on. As I explaind reflect world and reflect all work similiarly. Unless reflect world is just a premade cubemap. Which you seem to have ignored. Its the process of how real time reflections are done that you and dave seem to have over looked.

See its one shader thats occuring in either of the two types of reflections but how are they occuring. pre made CM's is the same cube map at every single frame reguardless of camera location. on the other had real time refections are snap shots of the world at the every frame (a render to texture at every single frame) both use fresnel's equation to do refraction so thats not going to change on what texture you use. The texture look doesn't change because the refraction is the same with either type of reflection.

The only difference is that render to texture at every single frame.... between those two shaders.

Since thats the only difference, what operations are occuring, a fresel refraction, a render to texture,

draw water plane

then a render to texture is aplied in second pass, along with fresnel refraction.

this second pass is the problem area for the geforce so lets take that apart

Dave said is not presumably possible, render to texture would be taken care of by the primary shader unit MUL or SFU or a texture operation, and the fresnel operation is taken care of by second operation since its a MAD and Dot operation.

Also you would see the same huge speed drop in other games that use the same types of shaders and thats not happening in Far Cry is it?

Last edited by Razor1; 11-22-04 at 04:46 PM.
Razor1 is offline   Reply With Quote