Originally Posted by grunion
I think not, I use x16AF in every game.
I think what he was referring to was the fact that it is commonly said that the HD3k series has poor AA performance when this isn't entirely true. It is mistakenly interpreted that way because review sites tend to turn up AA and AF at the same time, and conclude that the performance drop is due to AA being applied. In my opinion this is because they find out that AA is done on shaders instead of ROPs and without doing any additional investigation assume that this new method is the cause. Really, the problem lies with the relative lack of texturing capability on the part of RV670 and so it is the AF that drops the performance
That being said, true or not, the benchmark graphs do show that only 1xAF is being used. Not applying a lot of AF while using AA should indeed skew the results in RV670's favor and make it look like these cards are less of a generational leap. However, I'm undecided as to whether or not I think this is "fair". If these results are to be used as a comparison against HD4870, which should be a bit faster than the tested HD3870X2, then I think it is appropriate. My reason for this is becauase RV670 really did choke with AF, and I think doubling the amount of TMUs available should alleviate the performance issues with AF enabled. Add in the fact that Nvidia cards typically take little to no performance hit with AF enabled, and this becomes sort of a "simulated HD4870" comparison.
Note: I'm making no claims as to the validity of this bencmark, because noone really has much of an idea as to how either the GTX280 or HD4870 will perform at this point. I'm just expressing my thoughts about the given data, true or false.Edited by darkcloud89 - 6/15/08 at 9:55pm