No Anisotropic filtering with new ATI Catalysts?

[MAsta_KFC]

Hi guys!

After taking an extended break with CMAK, being preoccupied with other games + uni work, I decided to jump back in the other day. However, when I reached the battle screen, I was beset by an inordinate amount of 'sparklies' in the textures, something which I didn't have before. I quit the game, went to the ATI CP and upped the Ani filtering to 16x. To my dismay, when I entered the battle, it was still the same! In fact, i've tried several combinations of ani settings, yet it doesn't seem to make a difference. I remember the old CAT's (like back in early 4's), anisotropic filtering worked just fine.

Has anyone else noticed this? It really does make a big difference to the IQ, IMO.

[MAsta_KFC]

Oh, I should add I am using currently:

Radeon 9700pro

DX 9.0c

Cat 5.1's

[Redwolf]

Anisotropic filtering shouldn't have anything to do with "sparklies".

If we mean the same effect.

Can you post a screenshot?

[MAsta_KFC]

By sparklies, I'm referring to the sparlking effect of textures when they are some distance away. Makes it look like a light show, especially when scrolling around the battlefield. I'll try to post some screenies.

Any more comments, guys?

[Redwolf]

OK, so some things are unclear about your post. What is "IQ"?

Then, in the begging of the post you say you didn't have it before, but later you say that filtering as a fix worked before.

Can you be a little more precise what you had before and how you fixed it?

[Pine59]

I noticed the same thing after the last new drivers I installed for my 9800XT, the 4.12s. I run CMAK at 1600x1200 on a 21" LCD and have always gotten more benefit from AF than AA, plus there is the text problem with AA on ATI cards. I usually run 8x AF and have noticed an increase in sparkling or graininess which in the past AF definately improved.

I'm speaking for someone else and have never seen the abbreviation but I assumed IQ referred to Image Quality.

[Redwolf]

All right, so do I understand that the problem is that AF used to make the sparkling go away in older ATI drivers but doesn't make it go away with the 4.12 drivers?

Screenshot?

Is that sparkling at the edges of textures, especially those that join to form a 3D model with two textures visible?

[Pine59]

Hopefully this will work.

It is from some distance because this is where I notice it. Look at the buildings, the stone wall, and to a lesser extent, the stars on the flags. What I can't show is that the problem really shows up when scrolling around the map between turns. Large surfaces with details "shimmer" all across the surface. It is not the "jaggy edges" I associate with aliasing. Closer views, the phenomenon is much less noticable. But pan out and scroll and it is quite noticable.

Now, turning it off and on doesn't make a difference. It isn't the worst problem Ive had. Mainly, I just wanted to confirm that I think I am experiencing what MAsta_KFC reported. Truth be told, I think AF worked better on my previous nvidia card.

[MAsta_KFC]

Thanks for the screenie, Pine59, that's exactly what I'm referring to. And yes, i meant image quality when I said IQ =)

Turning AF on or off, or whatever combinations of x2, x4, x8 or even x16 has no effect on the game whatsoever, the sparklies still appear. Somewhere along the way, ATI broke the AF for CMAK (and maybe other CM games too, I dont have CMBB installed atm). Can others confirm? Looks like we may have to report this ATI.

[Wilhelm Heidkamp]

Guys, I have an ATI 9600 128Mb with Catalyst 4.12 drivers and I dont feel these kind of problems.

Im thinking the problem isnt the graphic card because this graphical problem isnt universal. I want to say that if this question was dued to the graphic device or latest drivers all users would suffer the same results, but it isnt.

If you have installed the catalyst command center, you can work about all configuration parameters through it. However, you can change configuration too through the advanced properties of control panel in OS windows. Ocasionally, this two ways to set up parameters are contradictories. For example, to play Operation Flash Point I disable (or get down) AF in catalyst command center but when I check the advanced properties of graphic card in control panel, AF is still on. Sometimes, it occurs the same with other parameters like the balance between performance and quality. Check the OS parameters, not only the catalyst command center. I usually, set my parameters in catalyst command center as first step. Secondly, I check my graphic card configuration through the control panel in OS windows XP. I modify it when necessary. No problems here.

[MAsta_KFC]

Interesting, Wilhelm.

I dont use CCC actually, I just use the normal CP (ie normal advanced properties tab) version as I didn't like the way the previous CCC was set up, although I've heard it's gone better. What I'll do is download the CCC version tonight and see if that works. If not....

Anyone else using Cat 4.12 or 5.1 and not getting AF on CMAK?

[Redwolf]

If you get the sparkles (it's still not clear to me from the screenshot, do you mean the while pixels on the flag's mast?), that doesn't mean AF is on or off.

The purpose of anisotropic filtering is not to correct graphics errors at texture edges.

It may "accidentally" do something about graphics glitches because it re-paints textures, but that be purely coincidental, in fact it would actually be a bug. AF on or off should not do anything about glitches.

Maybe I still misunderstand what you mean.

[Pine59]

No, its not on the edges but in the textures themselves. It is really hard to appreciate in a still shot. I know what you mean about the flagpole in my pic but that isn't the effect I've tried to describe.

I discovered a long time ago while playing CMBB that turning on AF seemed to help but not eliminate the effect. That was on an Nvidia card and every time I updated the drivers AF would be re-set to off. So, I would fire up CMBB, say "yuck" and realize I had to go in and turn AF on again (4x back in those days).

Updating ATI drivers doesn't turn off existing settings so I was surprised when I fired up CMAK and. saw shimmery, crawly, whatever you want to call it building and wall textures. I assumed something must have turned off AF but when I went in it was still turned on at 8x. I don't use CCC because they removed the option to use lower refresh rates on DVI monitors. Not being able to select this option causes my screen to go black when Windows begins loading (known issue with DVI on this monitor).

So anyway, I'm not motivated enough by this problem to roll back my drivers but whether it was an accidental improvement or not it does appear that my in game screens have resumed a "no AF" look. Haven't noticed a problem on any other 3D games but I don;t play many other 3D games either.

[Redwolf]

It's very odd that you had that with NVidia cards, too. I certainly don't and I don't remember it from my ATI days either.

Do you maybe had/have some program installed that would mess with the Direct3D libraries themself?

Are you able to see whether the improved texture quality (other than the sparkling) of AF is there? I've never been able to see it with CM but I'm not very sensitive to image quality.

[MAsta_KFC]

Okay, after installing and testing with ATI Catalyst 5.1 with CCC, AF is still not working on CMAK. Very dissapointed atm, I really hate seeing the sparklies while playing, especially when scrolling around the battlefield. Looks like I'll have to report it to ATI. I'll post a link up here too for anyone who wants to chip in. Doubt they'll do much about it, but we can still try.

[Redwolf]

Again, anisotropic filtering doesn't have anything to do with sparklies.

The sparklies must be an independent driver bug which in past releases has been accidentally covered up by AF.

You need to verify whether you AF is on and off by other means first.

[Schrullenhaft]

If this is what I think it is, then it may be an issue with mip-mapping. I'm not really sure if AF will affect mip-mapping and 3D operations on mip-mapped textures.

I believe the "sparklies" issue is when highly-structured patterns start "swimming" when the camera position is moved/adjusted. The "swimming" effect is when the patterns start 'crunching' together and lose their original pattern (noticeably). The somewhat random patterning in the stone wall probably makes this effect more distracting and/or noticeable. Where I see this often is in the tile roofs of CM buildings and it becomes more prominent the further the camera is away from these patterned textures.

If I'm not talking about the correct thing here, correct me.

It's possible that the AF in the past has smoothed out the mip-mapped textures in the past to the point that the "sparkly/swimming" effect is quite reduced (the textures don't seem to change as drastically when the camera is moved). But in the ever-important race for benchmark supremacy, the AF functions in the Catalyst driver may have been cutback for mipmapped textures at particular distances; distances that other programs rarely have.

As a bit of information, it's also possible that different Radeons may show different behavior with the same driver (although that is pretty rare). Because of the different structures and pipelines on the different Radeon cores it's possible for the same Catalyst driver to behave differently on different cores. The most common reason for this is the tweaks that the programmers put in themselves to support different functions and/or optimizations in each core (typically the latest ones). So a problem may show up in a newer core that doesn't show up in an older one. The "What's Fixed" Readme's seem to point to a lot of times when one particular chipset is having a problem with a program (though this may sometimes be due to specific reports rather than unique hardware differences).

When it comes to comparing CM to other games, it is quite different from a majority of games on the market. The horizon length/depth, the number and type of mip-maps and the color depth of CM (16-bit vs. 32+ bits) probably combine for a lot of potential problems that aren't seen in other programs. CM requires more fidelity from the drivers because of its characteristics and limitations compared to other games. This makes it quite susceptible to all of the programming shortcuts that are often used by driver engineers to optimize the driver performance.

[ January 31, 2005, 11:06 AM: Message edited by: Schrullenhaft ]

[junk2drive]

When I try to make a building mod with corrugated metal or bamboo I get the "swimming"

I think he is talking about the twinkling christmas tree lights in the distance effect.

Juju's website suggest adjusting AF as a fix for that.

[Redwolf]

Is that "swimming" only when you move the camera around or does it appear if you just sit there?

[junk2drive]

For me, only on the move, at an angle.

When not moving, things with close lines apear distorted and wavey, but stationary, again from an angle.

i.e. roof tiles

[Redwolf]

Allright, that would be anisotropic filtering then, or lack thereof.

What we need now is some test textures that allow people to observe in the CM engine whether AF is actually on at all.

What would be needed, would e.g. be a zerba stripe along a road (not crossing it) and then look along a very long road.

Anisotropic filtering should be clearly visible there by not "washing out" the textures when different levels of mipmapping meet each other.

After you verified whether AF is on at all or not on you would know whether this is a matter of broken AF filtering or of not turning AF on for CM.

Or just go and buy a NVidia card...

[ January 31, 2005, 12:55 PM: Message edited by: Redwolf ]

[MAsta_KFC]

Schrullenhaft, you are correct, it is indeed a mip-mapping issue here. As mentioned by junk2drive, the sparkly/swimming effect was negated by forcing AF on 4x or 8x, maybe even 16x if your vid card could handle it. And that was what I was doing 'in the old days' before taking a break. Like I said, by forcing AF in the new Catalysts, it now does absolutely nothing to negate the swimming/sparklies.

As to buying an Nvidia card, I still prefer ATI. Besides, I'm broke. =(

[Redwolf]

So did you verify that AF is on at all?

There are two explainations here:

- AF is on but doesn't fix it anymore

- AF for some reason doesn't turn on for CM

[MAsta_KFC]

After altering the settings, I have confirmed: AF DOES work for CM series (tried both CMAK and CMBB) but the distance it is in effect to reduce the sparklies has been toned down GREATLY.

EG. In the previous days, 16x AF would cover the entire map, eliminating all sparklies. However, now that I enable 16x AF, in terms of game distance, it only reaches 200m or so from the viewing point, anything beyond that is a sea or stars. 8x seems to work at about 100m-150m and 4x is dismal. This also means that view distances zoomed out are also affected, most noticeable from the top down view.

I'm very disappointed and I doubt that ATI will fix this if the reasoning behind scaling back 16x AF is to imrpove benchmarks. I'll wait for the Omega Catalysts to come out and see if I can play with the settings, but I'm not keeping my hopes up

[Redwolf]

Ah, thanks for looking into that.

Clearly, this is ATI playing with their drivers to increase performance. By not even trying to use the filtering on objects that are "obviously" too far away they get big performance gains. Except in this game they are not too far away and they trample all over the mipmap textures. Kudos to Schrullenhaft for expecting this.

Not the first time, here's another "optimization" they did and then lied about it until hard proven guilty:

http://graphics.tomshardware.com/graphic/20040603/index.html

Don't read it if you "like" ATI, you won't anymore.

And of course the sparklies shouldn't have been there in the first place, even with no anisotropic filtering. They sure are not on my NVidia cards.