Seems to be a lot of odd views about FSAA here.
"Although, pics really don't do FSAA justice. You have to experience it, to be able to judge it."
Several years using dozens of different hardware and software implementations here, yourself?
"And don't even bother trying to base your claims on nVidia's FSAA, because it's not even in the ballpark of the V5."
Yes, they are most definately in the same ballpark, little league. The fact is that both of their solutions are very low end, decidedly weak and not comparable to render engines several years old. I think they look very close to each other, neither of them are very impressive compared to a proper 16X FSAA implementation rendered at 8000x8000 and then resized down to 640x480 for output to video. They both are poor.
"There's more to FSAA than "jaggies". Well, at least on the V5 that's the case. Pixel popping and swimming/shimmering are a couple examples of how it improves overall image quality."
And you introduce haloing(high contrast) and blurring(objects viewed from a distance). The V5 is in reality worse in both cases due to its' useage of RGSS, the same things that give it an advantage to alleviate the flaws that your eyes are trained to pick up on, it increases the flaws that I look for(in terms of FSAA). Which way do you want to go? Neither of them are far off from each other, though I think that the V5 does hold an overall edge in FSAA quality, just a small one.
"I've done many comparisons (FSAA vs NoFSAA/higher res) on many games. And not one, did I find looked better without FSAA. Especially Madden, NHL, MDK2, and in particular, NFS5. That game is simply gorgeous with FSAA enabled."
If that is your taste all the power to you, but on a mathematical basis utilizing the resolution that the pixel data processed would allow is superior. For 1024x768 2X, 1536x1152 would be the comparable resolution. Of course, if your monitor can't handle it then that is a non issue.
"yes, hi-res sure is better, ON A GTS. The GTS's FSAA is weak. The 5500's FSAA kicks ass. That's why hi-res looks so much better on a GTS than FSAA does"
They are both weak, and very weak at that. The V5 is the better implementation if you are looking for the effect it causes, the GF2 is more accurate to the native pixel data. The GF boards also have higher levels of native pixel accuracy then 3dfx boards(required in the pro market), which makes the impact of FSAA less noticeable. Pixel popping is a rarity on my GF, particularly with the latest drivers(this has been an issue that nV has worked on more due to the Quadro market then the games, but the end results benefit all GF based owners).
"agreed, except the FSAA on a 5500 isn't a "mosh of pixels"."
You can certainly argue with a strong point that any FSAA is a mosh of pixel. The effect takes pixel data and blends it together, taking four accurate samples and producing one which more then likely is not true to any of the native pixels. Mainly the problem with current gfx boards FSAA is that they sample at such an incredibly low rate that your averages are weighted improperly, a problem which is compounded by the V5's FSAA.
"in some cases (NFS-PU), FSAA is definitely superior"
I agree with you absolutely in this case. Racing games and flight sims are well known for having serious aliasing artifacts and the accurate represntation of the data looks very poor. When a time comes that they work around these issues, perhaps expanding the level of mips for racing games and increasing the poly count significantly for flight sims, then it should be a case of higher res being superior.