Firstly, I would like to state that it is not my intention to fuel any burning fires. I simply think these are things that should be brought to people's attention. Secondly, I would like to state that it is obvious that none of these reviews that I will be pointing to are perfect. I don't claim them to be, I just think they inadvertently raise some interesting points.
As most of you probably know, FiringSquad released a review of the 5900 (linkage: http://firingsquad.gamers.com/hardware/msi_geforce_fx5900-td128_review/default.asp )recently using custom Timedemos made for Q3A and Serious Sam 2. While these benches are traditionally dominated by the NVidia line of cards, the roles reversed when the custom timedemos were used. I found this somewhat interesting. Furthermore, after the review was released, readers requested some of the IL-2 benches be looked at a little more closely. While the radeon cards dominated the benchmark without any AA, the NVidia cards blew them out when AA was used. So, the reviewer went back, looked a little more closely, and took some screenshots (which are included now at the bottom of the appropriate page of the review). It would appear that either NVidias 4xAA is only, at best, 2xAA, at least when compared to the radeon line of cards. Personally, I would have liked a direct comparison between the 5900 using 2xAA and 4xAA, just to see if the 4xAA is in fact any different. This, to me, can explain the relatively small performance hit that the FX series of cards seems to take when doing FSAA.
Also, GamePC (a site I hadn't heard of before, but read the review anyways) posted a review of the 5900 Ultra. (linkage: http://www.gamepc.com/labs/view_content.asp?id=fx5900u&page=1 ) They used several games that are not generally standards in the realm of reviews (warcraft III, GTA:VC). Again, in the less-than-standard benchmarked games, the ATI cards were simply faster than the 5900 variety. The only times that the 5900 took the performance lead was when it was taking extremly small performance hits from the FSAA, which, to me, now seems a little suspect. The NVidia cards continue to perform better in UT2k3 which is using a standard timedemo.
So, I'm just bringint this up for discussion. By all means, discuss. And, as an addendum, digit-life just did a very in-depth look at the entire 3dmark fiasco, comparing the various game test performance and image quality differences between the r3x0 cards and the NV3x cards. I'd say it's very well done and a good read, and in some ways, pays some compliments to the NVidia driver team. It shows that, in a couple tests, the NVidia drivers were increasing performance without IQ loss. (linkage: http://www.digit-life.com/articles2/radeon/herc-r9800-r7500.html )
edit: forgot the digit-life link
As most of you probably know, FiringSquad released a review of the 5900 (linkage: http://firingsquad.gamers.com/hardware/msi_geforce_fx5900-td128_review/default.asp )recently using custom Timedemos made for Q3A and Serious Sam 2. While these benches are traditionally dominated by the NVidia line of cards, the roles reversed when the custom timedemos were used. I found this somewhat interesting. Furthermore, after the review was released, readers requested some of the IL-2 benches be looked at a little more closely. While the radeon cards dominated the benchmark without any AA, the NVidia cards blew them out when AA was used. So, the reviewer went back, looked a little more closely, and took some screenshots (which are included now at the bottom of the appropriate page of the review). It would appear that either NVidias 4xAA is only, at best, 2xAA, at least when compared to the radeon line of cards. Personally, I would have liked a direct comparison between the 5900 using 2xAA and 4xAA, just to see if the 4xAA is in fact any different. This, to me, can explain the relatively small performance hit that the FX series of cards seems to take when doing FSAA.
Also, GamePC (a site I hadn't heard of before, but read the review anyways) posted a review of the 5900 Ultra. (linkage: http://www.gamepc.com/labs/view_content.asp?id=fx5900u&page=1 ) They used several games that are not generally standards in the realm of reviews (warcraft III, GTA:VC). Again, in the less-than-standard benchmarked games, the ATI cards were simply faster than the 5900 variety. The only times that the 5900 took the performance lead was when it was taking extremly small performance hits from the FSAA, which, to me, now seems a little suspect. The NVidia cards continue to perform better in UT2k3 which is using a standard timedemo.
So, I'm just bringint this up for discussion. By all means, discuss. And, as an addendum, digit-life just did a very in-depth look at the entire 3dmark fiasco, comparing the various game test performance and image quality differences between the r3x0 cards and the NV3x cards. I'd say it's very well done and a good read, and in some ways, pays some compliments to the NVidia driver team. It shows that, in a couple tests, the NVidia drivers were increasing performance without IQ loss. (linkage: http://www.digit-life.com/articles2/radeon/herc-r9800-r7500.html )
edit: forgot the digit-life link