Nvidia GF4 Aniso and FSAA

[KimB]

Because the 8500 doesnt make performance compromises with LOD Bias like the NVIDIA cards do. We already had this discussion earlier. I took shots at default LOD Bias for both cards, but this can, of course, be reduced down.

What? Texture aliasing is a rendering artifact that should be avoided. The Radeon 8500 (like the Radeon before it...) quite obviously uses too aggressive MIP LOD settings. This will usually result in superior screenshots, but there will also be situations in-game that will produce horrid texture aliasing.

The shots your provided show the EXACT same thing- little to no change by doubling the anisotropic filtering samples.

There is a change. This proves beyond a shadow of a doubt that your statement that there is no change between 4x aniso and 8x aniso was false. That you continue to argue the point baffles me.

Especially given how the GF3 in early drivers, the 8500 and SGI/USparc graphic workstations do not exhibit this "GF4-only" behavior. Just the GF4.

Try showing this, not just stating it.

You really need to try an 8500. From your commentary it's plainly obvious you have *never* used, seen or operated an 8500. This much is clear. And from that basis, I discount your opinions as such from someone that has zero experience, and this rings in more and more true with every post you place.

No, I have not used a Radeon 8500, and I don't intend to. Unless, perhaps, I happen to be near somebody who owns one (which hasn't happened yet...). This is, in fact, why I am here (Well, half the reason...I also like to argue ). I don't want to have to try out every piece of hardware in existence to get an idea of which hardware I would like best. For example, if I wanted increased texture detail at the expense of texture aliasing, I would adjust the texture LOD on my GeForce4.

 

[John Reynolds]

LLB-
John-

and I'm finding that I really like 4xS and 2x aniso (LOD at -05 to -1, depending on the game) in Morrowind, Sacrifice, Freedom Force, Independence War 2, and Aliens vs. Predator 2. And I often use those two settings in conjunction with high resolutions, such as either 12x9 or 12x10, while still getting very playable frame rates.

I love the quality and use the same settings, but just hate the performance. I'm really waiting for the R300/NV30 in hopes to see about 15-20% improvement with similar visuals. That would most definately be a "sweet spot" if we could have this level of visuals and 40+ fps to boot.

Whoa, can't believe I inlcuded Morrowind in that list <g>. While the other games I mentioned run just fine with 4xS and 2x aniso (I'm breezing through AvP2, for example, so frame rate definitely isn't hindering the gameplay in this shooter), Morrowind chokes with 4xS enabled.

 

[Dave Baumann]

This is, in fact, why I am here (Well, half the reason...I also like to argue

Well that can go elsewhere - this is a discussion forum.

 

[Sharkfood]

No, I have not used a Radeon 8500, and I don't intend to.

But instead you would rather make direct statements of idiocy concerning the card, it's behavior, it's method of performing things, and it's LOD bias defaults and aliasing produced therein. sigh.

Thanks, appreciate the conflicting information as it paints what the true case with your Radeon "commentary" truly is derived from- zero experience and based on an agenda from a preference to a competitor.

In the future, you really *might* want to try using both products before defining yourself as an expert in the process and making bold, wrong, baseless assumptions. It might lend one to just a hair of credibility that way.

Cheers,
-Shark

 

[SirPauly]

Too bad it still doesn't completely clear 'em up, and ATI's FSAA isn't very useful because of its performance.

As you mentioned, you did pick the worse-case example and created an entire map of this texture; what did you expect?

To say ATI's FSAA isn't very useful lacks logic because you're blanketing this and I question your objectivity. The FSAA may not be the fastest solution but that doesn't mean it is not an enjoyable and viable solution for countless titles that exist today, imho. I use this feature on every single gaming title I have with high levels of anisotrophy. I think both are very useful together.

ATI *really* should have done a much better job on the MIP LOD selection. Of course, nVidia's isn't perfect, either, as you can see from my screenshots.

I can't speak what nVidia or ATI really should do..... I can only offer my subjective opinion what is on my screen and share my view. I am not bright enough to think I know more than 3d engineers from any company considering it isn't my field.

 

[Bambers]

As 4xs 2xRGMS applied to a screen with doubled vertial res does mean there will only be a gain in anisotropy (of 2x) only on floors and ceilings and only horizontal edges look better than normal 4xAA.

Moving OT here, interestingly ATIs 4xquality smoothvision under d3d-nofog in 90xx drivers and d3d and opengl in 7206 drivers seems to use a similar technique only its a 2x diagonal pattern AA (used in 2xquality in the same situations) on a screen rendered twice as wide.

It also seems to be able to 'flip' the 2x diagonal pattern round 90 degrees but how often it does this depends on the drivers and api. In 7206 d3d it can change every pixel giving 'dirty' looking edges at low resolutions although the effect is quite good at higher res. In 7206 opengl and 90xx d3d-nofog it changes less regularly, about 5 times for the whole screen.

I think it would probably be too hopeful to expect ATi to say what is going on with SV at the moment and if its ever going to be like the whitepaper specifies.

Still, 2xq does a perfectly good job of clearing texture alaising and its still perfectly playable in most games. I run morrowind with 2xAA and 16xaniso.

 

[Randell]

I run all my games in 2xquality, 16x aniso, 1024x768 and they are all very playable. Some games look better with 3 x performance, but I cant be bothered to enter the display app to flip settins all the time.
I cant run any higher rez because its an LCD monitor.

On older racing games (eg Midtown Madness 2) 4x is also eminently playable.

I cant say I've ever noticed the moire in other game other than CS and the yellow stripe textures on floors when they are at a distance.

 

[Ichneumon]

I run all my games in 2xquality, 16x aniso, 1024x768 and they are all very playable. Some games look better with 3 x performance, but I cant be bothered to enter the display app to flip settins all the time.
I cant run any higher rez because its an LCD monitor.

On older racing games (eg Midtown Madness 2) 4x is also eminently playable.

I cant say I've ever noticed the moire in other game other than CS and the yellow stripe textures on floors when they are at a distance.

Laf.. you need Rage3D Tweak so you can set profiles for your games that automatically apply the right Smoothvision and AF settings each time you launch one of them.

 

[KimB]

To say ATI's FSAA isn't very useful lacks logic because you're blanketing this and I question your objectivity. The FSAA may not be the fastest solution but that doesn't mean it is not an enjoyable and viable solution for countless titles that exist today, imho. I use this feature on every single gaming title I have with high levels of anisotrophy. I think both are very useful together.

The problem here is that the GF4 is faster in these scenarios.

I can't speak what nVidia or ATI really should do.....

It's called ethics. It's about not making your card look slightly superior in certain situations (most screenshots that are shown in reviews of the video card), and vastly inferior in other, albeit rare, situations. To me, this is the exact same as the Quake/Quack issue that went on a while ago...

It has nothing to do with knowing about how to engineer a graphics card.

 

[Randell]

[quote="IchneumonLaf.. you need Rage3D Tweak so you can set profiles for your games that automatically apply the right Smoothvision and AF settings each time you launch one of them. [/quote]

I know I know I have it, I just found it hard to work out and gave up after I buggered the default settings and uninstalled it I'll go back to it, but it could be easier to use

 

[Ichneumon]

I know I know I have it, I just found it hard to work out and gave up after I buggered the default settings and uninstalled it I'll go back to it, but it could be easier to use

I'll stop the OT in the thread now... but yea you are right that it isn't exactly an entirely intuitive program.. but then again it does have fairly good built-in help files and once you understand the basic operation of it, everything is simply an extension of that. laf, it does a lot of stuff, and the more stuff you make something do, the less easy it tends to be to make it entirely intuitive. Feel free to drop me an email or post to the rage3d tweaker forum if you have specific questions.

 

[SirPauly]

It is impossible to discuss rationally with another when they don't own the product so this discussion is over with you. You don't want to discuss; you want to rant how unethical ATI is and is your right and so be it. I just wanted to share that ATI's AA does a nice job on texture aliasing artefacts, nothing more.

I can't discuss things rationally when a person says this to me:

It has nothing to do with knowing about how to engineer a graphics card

But yet offers this earlier:

ATI *really* should have done a much better job on the MIP LOD selection.

 

[KimB]

It should be obvious. I have two separate arguments here.

1. No matter the MIP LOD selection algorithm, there is always the possibility of producing an image with little to no aliasing.

2. The Radeon 8500's MIP LOD selection algorithm is inferior to the one seen on all GeForce cards. You can see this by simply looking at some r_color_mip_levels shots in Quake3. All GeForce cards use a per-pixel range-based MIP selection algorithm. The 8500's looks much like the range-based per-fragment algorithm seen on the TNT/TNT2 cards.

 

[Sharkfood]

The problem here is that the GF4 is faster in these scenarios.

More ignorance spread from an NVIDIA fan, which is getting more and more obvious with each and every post.

So let me correct you-
"From my ZERO experience ever using, seeing, experience or benchmarking an 8500 and comparing with equal modes on a GF4 Ti4600, I conclude with such zippo evidence that the GF4 is faster in these scenarios... just because."

The Radeon 8500's MIP LOD selection algorithm is inferior to the one seen on all GeForce cards. You can see this by simply looking at some r_color_mip_levels shots in Quake3.

Additional ignorance. Discussion closed. When you are ready to take off the NVIDIA-blinders, please feel free to rejoin the discussion. Until then, you can consider yourself ignored by anyone that has actually used these products.
8500 Quake3 color_miplevels "1"

Moreover, the "3 tier" mipmap boundries of drivers past were replaced with the dithered approximations in current drivers not to alleviate a "cheat", but to simply deal with all the clueless ninnies that took colormips and didn't understand the many different methods of texture filtering that can be used to accomplish the same thing, but defy colorized mipmap levels (which doesn't occur in 3D graphics naturally).

If mipmaps are simply different resolution/detail levels of the same texture, many faster algorithms simple 'mipmap' a mipmap, by approximating a 4x4 square at one mipmap level to be a 2x2 square in the next. So therefore, traveling outside the current texture reads isn't required (and is expensive in segmented memory models) so the next mipmap level is based on the current for the samples. This yields no color blending at mipmap borders since the samples for filtering are from within a single mipmap, extrapolated by using knowledge of what mipmaps are.

Unfortunately, the band of the NVIDIA collective ingoramuses decided this somehow meant texture filtering wasnt "correct" and therefore dubbed it improper. Of course, having broken anisotropic filtering, "as expensive" 4xS mode to yield 2 samples (yet an equally expensive 2-sample AA is somehow "too slow" if it's on a card that doesn't have a little "N" on the chip) and other driver issues are all okay though. lol.

Well, it's not even worth discussing anymore. If some people wish to go through life making blatantly absurd statements with absolutely zero credentials or reason behind them, and settle for partial delivered features for their hardware expenditures, more power to 'em. I just happen to insist my hardware works and works well.

Cheers,
-Shark

 

[KimB]

So let me correct you-
"From my ZERO experience ever using, seeing, experience or benchmarking an 8500 and comparing with equal modes on a GF4 Ti4600, I conclude with such zippo evidence that the GF4 is faster in these scenarios... just because."

No, I was actually quoting from some performance tests I did with another member on this board, where my Ti 4200 outperformed the Radeon 8500 with 2x FSAA and full anisotropic. I'm sure those tests aren't far, if you care to look.

Additional ignorance. Discussion closed. When you are ready to take off the NVIDIA-blinders, please feel free to rejoin the discussion. Until then, you can consider yourself ignored by anyone that has actually used these products.
8500 Quake3 color_miplevels "1"

That screenshot shows exactly what I was talking about. Unfortunately, my TNT isn't available to me right now, but it should be obvious to you that the MIP map boundaries consist of two straight lines intersecting at the center (Which is what you see on a TNT). If you do this on any GeForce card, you will get a smooth semicircle, which is visual superior (albeit very slightly).

Regardless, just from those screenshots, it should be obvious that when the two cards are using the same filtering, the GeForce card will product a better aliasing to texture clarity ratio (although, again, very slightly).

You were, apparently, talking about a totally different idea than I was.

 

[OpenGL guy]

That screenshot shows exactly what I was talking about. Unfortunately, my TNT isn't available to me right now, but it should be obvious to you that the MIP map boundaries consist of two straight lines intersecting at the center (Which is what you see on a TNT). If you do this on any GeForce card, you will get a smooth semicircle, which is visual superior (albeit very slightly).

There's more to texture quality than the D-curve a particular card uses...

Regardless, just from those screenshots, it should be obvious that when the two cards are using the same filtering, the GeForce card will product a better aliasing to texture clarity ratio (although, again, very slightly).

Not necessarily. There are other factors such as LOD bias to take into account.

 

[KimB]

Yes, if the LOD bias selection algorithm is incorrect (MIP map lines aren't properly-spaced), there will be additional problems. But, I don't really think that's an issue with any of the cards I'm talking about.

But, given the same exact texture filtering algorithm and LOD bias, nVidia's per-pixel MIP map level selection will produce superior results (But, again, very slightly so...).

Of course, we do know that ATI's MIP LOD bias is significantly more aggressive than nVidia's, and it also is beginning to look like there is additional aliasing when textures' u/v axes are not parallel to the screen axes. Why doesn't ATI lower MIP LOD in these situations? Is the hardware incapable of it?

One question, can you adjust MIP LOD on a Radeon 8500? If not in the driver, then in software with Vogel's OpenGL in UT?

 

[KimB]

Just something that might be interesting.

I decided I wanted to try and find out what MIP LOD I could select on my GeForce4 that would produce absolutely no texture aliasing (On the level I posted before). It turned out that I had to select a LOD around +2 for me to be utterly unable to see any texture aliasing, and this is apparently independent of the use of anisotropic filtering. Unfortunately, this made the display into a smeared blur. Obviously nVidia does have a little ways to go for optimal texture filtering. Fortunately, the level I posted is will, by far, show more texture aliasing than pretty much any real gaming situation.

 

[Althornin]

and it also is beginning to look like there is additional aliasing when textures' u/v axes are not parallel to the screen axes.

Where are you getting this?
I dont think its beginning to look like that at all. Please show me some of the evidence you've collected. Oh wait, you dont have any.

Make two maps in UT, very much like the one you just made. However, have the texture be one that looks the same from a 45% angle - IE, say, an octagon pattern. This way, it will look the same whether viewed from 45% to U/V "axes" and parrallel to U/v "Axes". This could put your claim to rest right now - or verify it. But there is NO evidence of this currently, in fact, there is plenty of evidence to the contrary, that Bambers came up with, that he showed to you earlier (remember....? its NOT rip-mapping...)...

 

[KimB]

Where are you getting this?
I dont think its beginning to look like that at all. Please show me some of the evidence you've collected. Oh wait, you dont have any.

It's called screenshots, though I will admit it is not certain that the issue is based on the texture axes not being parallel to the screen axes.

Anyway, look at the screens that Sharkfood posted on page two of this thread. Most of the texture aliasing occurs at the center of the image, and tapers off towards the sides. It turns out that for this texture, the texture axes are rotated 45 degrees from the screen axes in the center, and are parallel at the sides of the shot.

Additional evidence can be found in some Serious Sam screens that were floating around, that show an opposite effect. The texture was aligned parallel with the screen axes in the center of the shot, and skewed at around 45 degrees towards the sides. Guess what? The texture aliasing was only really visible at the sides of the shot.

For these textures, then, it should become obvious that the texture aliasing is only present when the texture axes are rotated with respect to the screen axes. This would seem to be indicative of the anisotropic filtering only filtering along one texture axis.

I went ahead and adjusted the LOD bias on my GeForce4, and it tended to show aliasing much more evenly across the directions (although there was much less to the right of the image...more to the left, in the level I designed to test aniso/aliasing). I believe this is a result of the texture chosen. Here's a shot:

http://169.237.254.233/aliasing.jpg

Update:
The above shot was taken with no anisotropic filtering. Since we are comparing anisotropic, I decided to take another shot with 8-degree aniso. To exame worst-case texture aliasing here, I further moved the LOD bias to -16.

http://169.237.254.233/anisoaliasing.jpg

Make two maps in UT, very much like the one you just made.

Not really necessary. Just take my map, open it up in tournament mode, then move so that you're facing one of the axes of the texture (preferably in the direction that showed the most aliasing in my screenshot above). There should be enough difference here that the shots don't need to be in the exact same position. If there's not, I'll edit the map (Or you could, if you really want to...all you have to do is move the playerstart actor).