In case you were wondering, like I was, the precise performance tradeoffs for various AA settings on a GFTi4200 (64MB in my case), I've posted my UT2K3 Demo benchmark results. The fancy graphs show my FPS in flyby and botmatch at 1024x768, 1280x960, and 1600x1200 with all of the Ti4200's AA options (None, 2x, Quincunx, 4x, and 4xS). I ran all the resolutions, even though 1600x1200 was unnecessary for me -- my 8 year old monitor shut off during those tests because it couldn't handle the massive res. Anywho, check them out, it took me a while to do. And could anybody tell me exactly what 4xS is?
http://members.cox.net/thatbox/UT2K3Benchmark.htm
[FLYBY]: I found it interesting to note that the FPS was actually higher with 4xS on the jump from 1280x960 to 1600x1200, that 4xS had a better framerate than 4x at 1600x1200, and that the performance hit from 2x and Quincunx (I semi-kinda know what this one is, but feel free to explain it to me too!) is virtually the same, although I did not check image quality. I've heard the first two explained away by saying that at 1600x1200 some tiny details (and therefore jaggies, too) become unnoticeable, so the GPU doesn't have to worry about them. Is that true?
[BOTMATCH]: I was surprised to discover also that the only relative difference with the botmatch results was at 1024x768: I expected 4xAA to be closer in performance to 4xS, at least judging from the flyby results, but its performance is much closer to that of the lower AA settings than in the flyby. I guess I'll have to assume that my graphics card is my bottleneck here, since the flyby and botmatch (which requires heavier CPU involvment) graphs are so similar (I believe that if the GPU weren't the bottleneck, the botmatch graph's bars would all be closer to being even, ie, the frames per second for every res and AA setting would be closer, right?).
Oh, and sorry that the exact FPS aren't on the bars; I couldn't legibly include them. Enjoy!
Edit: Omitted a fairly important word
http://members.cox.net/thatbox/UT2K3Benchmark.htm
[FLYBY]: I found it interesting to note that the FPS was actually higher with 4xS on the jump from 1280x960 to 1600x1200, that 4xS had a better framerate than 4x at 1600x1200, and that the performance hit from 2x and Quincunx (I semi-kinda know what this one is, but feel free to explain it to me too!) is virtually the same, although I did not check image quality. I've heard the first two explained away by saying that at 1600x1200 some tiny details (and therefore jaggies, too) become unnoticeable, so the GPU doesn't have to worry about them. Is that true?
[BOTMATCH]: I was surprised to discover also that the only relative difference with the botmatch results was at 1024x768: I expected 4xAA to be closer in performance to 4xS, at least judging from the flyby results, but its performance is much closer to that of the lower AA settings than in the flyby. I guess I'll have to assume that my graphics card is my bottleneck here, since the flyby and botmatch (which requires heavier CPU involvment) graphs are so similar (I believe that if the GPU weren't the bottleneck, the botmatch graph's bars would all be closer to being even, ie, the frames per second for every res and AA setting would be closer, right?).
Oh, and sorry that the exact FPS aren't on the bars; I couldn't legibly include them. Enjoy!
Edit: Omitted a fairly important word
Facebook
Twitter