Oldfart
BF1942 is unplayable with 16X AF as is MOHAA. In fact, BF1942 stunk with any amount of AF with the Ti4600.
Which driver revision had the option of 16x AF? I don't have BF1942(online based game, still stuck on dial up
) or MOHAA(played it, couldn't stand what I did play). I have tested fourty or fifty games so far, with the exception of the Unreal engined games(
all of them for some odd reason) I'm not seeing a huge performance hit(all the games remain quite playable using 8x Quality AF at high res).
RTCW also took a massive hit. I didn't expect a GF4 to perform as good as a 9700P. However, I wanted a card that gave me the ability to max out all detail, and use AA/AF filtering. because of the noise/heat issue with the 5800, that left two choices. 9700 and 9800.
I don't have any problem with people liking what the R300 does better, it just does it wrong(as does the NV3X boards). Given what you were looking for obviously you got a viable part. Saw mention of you offering Rollo a trade for your Ti4600 +$100 for a R9700Pro for your kids computer, I would have traded you a R9500Pro even up
BFG-
Except the indicator doesn't look anything like that in a real game.
It is displaying what the anisotropic filtering implementations are doing. The excessive bowing and improper mip transitions are simply proof of what is evident in game.
That thing can't even show depth correctly for heaven's sake and even if you use the checkerboard view you still have untextured walls and an untextured roof so again it really shows nothing.
Do you understand what the test is doing? It is displaying how the AF implementation is working, and it does an extremely good job at it.
If you want to buy cards based on theoretical colour wheels then go right ahead. I prefer to test real games and determine how things work.
Me too. I spent my $200 on a R300 powered board and was shocked at how poor the texture filtering was. Listening to people all over they raved about how great it was. Had a lengthy discussion about this over @B3D prior to seeing the color wheel tests and I thought it was extremely obvious what they were doing wrong, the tunnel tests and color wheel simply backed me up.
I'm all for theoretical discussion but at some point you have to get back into the real world and simply look at what's in front of you.
I probably could have tollerated the level of IQ the R300 board produced(actually, it's 2D was quite poor also, was a luck 'rolling lines' winner and had to use the DVI output, the cheap BFG Ti4200 I'm on now has better 2D then the DVI>VGA out on the ATi board, the Gainward GF2 that was in before was so much better that my wife was griping about how bad the 2D was), but combined with all the games that don't work properly it was too much. When Sacrifice had massive image corruption and had been in that condition for
six months it was pretty clear that they had very different priorities from what I was looking for.
That's interesting considering even 2x anisotropic was unplayable in most games on a Ti4600, and 8x was absolutely out of the question.
Was there a configuration issue, or a driver problem previously with the Ti boards? Running Q3 16x12x32x8 without AA I hit 100FPS, far removed from what I would call a slideshow. For RTCW, running a Ti4200 @stock speeds(mine is @290/560) 1280x1024 8xAF averages over 60FPS according to Digit-Life (
Link).
Please give me an example of a mainstream game (ie one that I'm likely to have) including your tested area and I'll be glad to take a look.
With noticeable mip banding? UnrealII, SS, SS:SE, Mafia, NOLF, NOLF2, Half-Life, CounterStrike, Blue-Shift, UT, Unreal, JKII and Sacrifice for mainstream games off the top of my head. All of them exhibited easily noticeable mip banding, the grates in the HL powered games were horribly bad. Pretty much the only circumstance I didn't see noticeable mip banding is if I was playing games in which the AF was agressive enough to push the mip transitions back beyond visible range(Quake3 for example).
High resolutions. Also if you like it's a simple matter to move the LOD from -0.5 to 0.0 and you'll get the level you get on nVidia cards but you'll still keep your lightning fast rendering when using AF. Personally I like the LOD how it is because I run very high resolutions and the image quality is simply pristine.
You can say the higher resolution is a way to solve anything, although it doesn't eliminate texture aliasing for me. Adjusting the LOD bias is a hack, you can do it on any board and you introduce rather severe texture aliasing adjusting it. If you don't mind that, then this discussion really won't go anywhere.
I'd also like to add that texture shimmering/sparkling will always exist regardless of your anistorpic filtering methods since it's one of the associated artifacts of converting from world co-ordinates to screen co-ordinates. And again, high resolution will reduce it the higher you go.
No, transformation is not the main reason for texture aliasing nor is it directly associated with it(unless you are doing something horribly wrong). It is based on the sampling rate(in relation to physical space and in physical space terms exclusive), contrast and movement. Aliasing can be eliminated in terms or perceptible to a typical person running a resolution of ~8,000x8,000(upper limits of human perception for a typical viewing area) or a sampling that is comparable to such resolution. Running 1600x1200 you could pretty much eliminate all noticeable aliasing running ~35x SS FSAA. Obviously this is not viable for quite some time to come, not to mention the board vendors are moving away from SS AA anyway. Using AF you have to have sampling that is capable of dealing with the level of contrast in the textures and the amount of movement. The problem with the NV3X and R3X0 implementations is that they are adaptive and cut too many corners. On ATi's part it is worse as the LOD is adjusted for optimal angles, which introduces higher levels of contrast, while the amount of samples taken is noticeably off on worse case angles. nV's approach uses a fairly flat implementation(with their NV3X parts), you don't run in to issues with optimal and worse case angles, although it still falls short of the NV2X's proper AF. Running ~64x anisotropic filtering on the NV2X style boards you should be able to pretty much eliminate mip mapping altogether, although obviously the performance hit would be staggering. With the adaptive implementation we will likely need to see at least twice that level, and even then you could run in to odd cases where it wouldn't work properly.
You are also talking to someone who has exhaustively tested the Ti500's/Ti4600's 8x setting. I've seen what you call the best and I don't believe it's the best at all.
I've done the same with the R300. In terms of anisotropic filtering the NV2X is hands down the best. It's performance hit may not be worth it to the majority of people, but the implementation is definitively superior.
That's hopeless as you'll easily dip below 20 FPS on the heavier maps. Even 2x was unusable on my Ti4600 in Quake III because it really hammered the framerate on the tougher maps. And to add insult to injury, 2x is largely a useless setting anyway because it doesn't really improve things much at all.
What settings were you running? For me, AF comes first
always, then AA if it is doable. Either there were some serious issues with earlier driver builds in terms of performance, or something was out of whack on your rig. Check the Digit-Life article I posted a link to above, the Ti4600 with max AF is faster then the 9500Pro with max AF running Quake3 1600x1200. I know the 9700Pro is quite a bit faster, but it's not twice as fast as the R9500Pro.
Facebook
Twitter