The tests were done with the only WHQL benchmarks at the time, not 'OLD' drivers, and those 'new' drivers you and nV wanted them to use were full of cheats, and even nV said they weren't good, and wait for the 53.03 for the first TRUE forceware drivers. Even the 52.16 had bugs, and they wanted them to use the 51.75!
The fact that you don't ackownledge the DX9 hole in the FX hardware shows your nV bias. The run-time compiler is impressive in how it can make up for alot of the shortcomings of the hardware but it needs to be programed to handle those shader perfromance issues. Look at the Tomb Raider and Max Payne2 benchmarks of late and see that there is still that same hole. And they will have to 'fix' the run-time compiler for each new game (it doesn't work evenly across the board).
I think nV's done a good job with the run-time compiler, but don't ignore that that problem exists, despite your conspiracy theory BS. The fact that even with the 53.03 forceware drivers the HL2 leak still plays much slower on nV hardware shows that it's not just the drivers. Also, remember that Carmack backed up those results by valve, saying that he's seen similar performance results with D]|[ which has the ATI leading the FX under equal paths. Only with the FX-centic path does the FX outperform, and even then it's just barely better at lower precision.
The conspiracy theory stuff is getting pretty old by now.
I'll give credit to nV's software guys for doing all they can to make up for the hardware shortfalls, but I'm not about to turn a blind eye to the whoel issue.
- You need a licence to buy a gun, but they'll sell anyone a stamp <i>(or internet account)</i> ! - <font color=green>RED </font color=green> <font color=red> GREEN</font color=red> GA to SK
