It is also unfortunate that some of you siimply refuse to acknowledge that there is a difference between an "Honest" Mixed-Mode for Dx9 used in a benchmark and a "Dishonest" mixed mode.
Throw out this notion of an "Apples to apples test" Its nothing but a misleading bloatware statement.
Running mixed-mode FP16/fp32 aginst standard DX9 path is ok as long as the Iq results are pointed out. Its still all very legal and pretty much everyone by now understands to some degree why its being done. Great....
That is clearly not what is happening here anymore. Nvidia is not using a proper "Mixed mode". But replacing shaders ad hock. Which also afects the output. Replacing as many PS2.0 shaders with PS1.4. I can also virtually Gurantee you that there is not even one instruction in all of Nvidias Replacement shaders that are ever running FP32. Its all FP16 and LOWER.
This is the problem. There is already a DX8 path in the game that uses PS1.4 Thus to take the High Quality DX9 path and Subvert it through drivers to 90% DX8 while Still reporting results as "Mixed-Mode" DX9 is outright FRAUD.
Why is it Fraud? Because this is a "BENCHMARK" Not a normal Game situation. Its the kind of thing that can easily only be pulled off in a limited Benchmark situation with a Controled set of variables. Not something that can be easily reproduced in a full game.
Overam Mirage 4700
3.2ghz P4 HT
SIS 748FX Chipset 800mhz FSB
60Gig 7200RPM HD
Radeon 9600M Turbo 128 (400/250)
Latest good read. [url]http://www.hardocp.com/article.html?art=NTc4LDE=http://www.hardocp.com/article.html?art=NTc4LDE=[/url]