texture shimmering on 6800 GT even w/ optimizations disabled

Does this file require a special codec? WMP10 doesnt seem to want to play it.

 

xvid

 

realized after that I didnt have the latest Divx codecs and grabbed them

Thanks

 

I have downloaded the UT2004 demo just to look for that texture shimmering. And I safely can say, that it does not occur on my computer! And that even though I'm verry sensitive to graphical glitches!
I use a GeForce 6800GT (from Galaxy) with 4xAA and 8xAF (all "optimizations" in the drivers disabled) with the resulution of 1280x960.

Of course there is still a certain amount of flickering on polygon edges especially on stairs e.g. in doom 3 (as I said, I'm really sensitive to image quality ), but that is normal and heavyly reduced by AA... (I heared that in movies they use 64xAA and higher to reduce those effects! ) That of course is -I presume- something completely different from your texture shimmering problem, isn't it!

 

The 9700Pro videos (Painkiller and UT2003) you can see here:

http://www.oc-zone.com/modules.php?name=Reviews&rop=conteudo&id=87&page=9

Note that the NV40 videos are the same that I posted earlier in this thread...

 

Ok I might be wrong due to unsufficient testing, yet I'm really running out of time. I need some of you to give it a shot and for the time being it seems to be OGL only:

Fetch Rivatuner if you shouldn't have it and under the OpenGL tab--> Intellisample and under "Override Intellisample texture compression control" set it to "determined by application". I have my doubts because I quickly tested it only in the Painkiller demo and was constantly alt+tabbing in and out to see what the heck would change anything with LOD to "0".


***edit: forget the whole thing *sigh* that's what you get when you try to test after an exhausting day.....Painkiller isn't even OGL arrgghhh.

 

just for shits and giggles I will try it in CoH tonight as people have complained about shimmering in that.

 

I have both cards, and the shimmering is clearly visible on the 6800U, but only on certain ranges from the "camera" viewpoint.

 

Does nothing for City of Heroes shimmering. Only 4xS or 8x + LOD +1 reduces it enough (doesn't eliminate it).

 

any updates on this ?

is it only an issue SOME people experience or everyone ?????

Edit: My first thought was..."Nvidias version of Antialiasing optimization".....and *IF* it is some "optimization" - well there should be a way to turn it off ???

It looks definetly like parts of the image dont get AntiAliasing.

Edit: Is this problem driver dependant ??????

 

Particleman what wattage power supply do you have?

 

The image will be further improved if you set:

UseCompressedLightmaps=False

 

I have played around with Unreal Tournament 2004, where the levels Asbestos and Grassy Knoll provide excellent examples of the shimmering effect, and been able to rectify the shimmering problem using methods stated earlier in this thread. By setting the DefaultTexMipBias to 0.2 (in UT2004.ini)or using Rivatuner to increase the LOD, I can remove the shimmering completely.

Has anyone studied this solution to see if the resulting image is the intended and correct one? It would be interesting to see how the final image compares to that of ATI' s R300 and R420 hardware and earlier Nvidia GPUs.

It seems that NV40 is using a negative LOD bias (sharp) by default and this, coupled with some games, causes sharpness that is beyond the ability of the hardware to smooth out. In testing I tried forcing the extremes (-3/+3) using Rivatuner in both OpenGL and Direct3D. Some games become so pointillated at -3 that everything begins to sparkle. This may be expected, but I don't recall the effect being this strong on other hardware.

As a further note, I was tempted to start a new thread for analysis of this problem and I am glad to see that this thread has been revived. However, it would be more useful if we could come to some sort of plan of attack on this issue, to explore it and understand it. There are many threads like this on various fora, but they seem to be mainly about complaining about the issue or claiming it doesn't happen at all and some dwindle to become an Nvidia v ATI slugfest with the issue as ammo for the flamewar.

I am really puzzled by the statements from some people that they don't see this effect at all. Could this be the case or are they simply not paying attention to it? Using high levels of anisotropic filtering and super-sampled anti-aliasing may reduce the shimmering effect, but this is not a proper solution as my UT 2004 experimentation has shown me. Once the LOD is 'correctly' set the shimmering vanishes and it looks good. The question here is does it look right or are we not compensating, but lowering the LOD to where UT 2004 is a low LOD blurry soup compared to what it should be?

Here I wish to make a special mention that when first playing UT 2004 on the 6800, I was taken aback by how detailed and crips everything looked. Maybe it was the FSAA and AF, but I will try to go back and check how it plays by default on a GeForce 4600 to make sure.

It seems to me that the issue is one where the LODs used are too high by default and that manually forcing a higher bias cures it. Another example is Painkiller* where sharp textures are too sharp and the moire/pixel crawling/shimmering can easily be seen. By using Rivatuner to set the bias at 1.0 these problems vanish, but you get a softer overall look and, again, I am not sure if this is acceptable.

So, I propose that instead of just reporting the problem, we try to analyze it to find out what is going on. There are two points I would like to see addressed:

1. By using an increased LOD bias ( > 0) in game configurations or by forcing it through Rivatuner, are we compensating for a LOD error or are we simply hiding it by decreasing LOD?

2. Because some people are reporting that they never have the shimmering effect, is it possible that a certain driver installation sequence exposes the problem? I have tried a few drivers, but they all have this effect as far as I can recall. Could it be that those who do not have the problem are using another hardware revision? (shudder) Of course, this could all be due to some people simply not seeing the shimmering for whatver reasons.

 

I decided to install UT 2004 on another system with a GF 4600 to see how it compares. Being lazy, I only installed the CD distribution and did not apply any updates. It turns out that UT 2004 Asbestos, using the default settings (DefaultTexMipBias=-0.5) plays without shimmering on the GF 4 using ForceWare 56.72. Overall I would say that the default GF 4 look is the same as GF 6800 with the LOD compensation, but it is difficult to be precise without spending considerable time taking and analyzing screen shots.

So, my overall impression after this quick look-see/test is that by altering the LOD you get the correct image on the 6800 (as compared to GF4. Radeon 9xxx/Xxxx comparisons would be nice, but I only have a 9200 handy) and that you do not suffer from a degraded visual experience by making this alteration.

 

I seem to have forgotten about my asterix next to Painkiller so I will eloborate in this new post.

Painkiller is one of those games suffering from the shimmering/moire effect very badly. It uses quite sharp/detailed textures to begin with and this may have something to do with why it can become unbearbaly sparkly when playing. I recommend downloading the latest single player demo for this title here.

This demo map (Military Base) is a good example because it has large expanses of tarmac and corrugated iron that helps bring out the moire effect. It also features SM 3.0 functionality with HDR so it may be interesting that way also, not to mention that this game is exceptionally fun to play.

What makes this interesting is that I can see no way to manipulate the LOD bias in the configuration file or by using console commands. However, the effect is made worse by using anisotropic filtering, whereas bilinear and trilinear do not show the same level of shimmering or even the same type of shimmering. Therefore, you must use something like Rivatuner to force the LOD bias (to +1.0). Doing so causes the majority of the shimmering to go away. Note that upping the LOD bias by 0.5 reduces ground shimmering, but leaves it on the vertical (best seen by looking at walls of corrugated sheet metal). As stated, setting it to +1.0 should cure almost all of the shimmering.

Once again, are we seeing what we should be seeing when we make this compensation? What can we learn from the fact that anisotropic filtering brings out the problem? Try using Rivatuner to force a LOD bias of -3.0 and prepare to see something that resembles software rendering, where the shimmering is so bad it makes the game look entirely pixelated.

Note that using trilinear or bilinear does not fix the texture aliasing on the vertical plane as walls still show terrible crawling when moving towards/away from them or panning across them. Also note that bilinear seems to work better against the shimmering on the ground as trilinear still leaves some of it there. Naturally, the look is much blurrier, but I will reserve my judgment as anisotropic filtering has spoiled me and if the 6800 has a sharpening effect by default I may have skewed expectations of what the final result should look like.

PS. A really good area to see the difference in ground and wall shimmering is the room with the radar. Clear it of baddies and hang out for a while. Take some pictures

 

I post these screen shots as illustrations of the effect in Painkiller. They were taken from the demo mentioned in my previous post (Military Base level) using trilinear texture filtering. It is therefore more or less devoid of shimmering on the ground, especially in these shots, but exhibits the moire effect on the walls and general blurriness of the floor (which is probably to be expected when using trilinear filtering without any sharpening ehancement). As these images are compressed JPEG they should not be used for definitive image quality assessment. The moire effect is preserved.

Picture 1: Reference picture of the locale with radar previously mentioned.



Direct link to image

Picture 2: Shows the moirse effect on the corrugated walls. Pay close attention to the area left of the crosshairs.



Direct link to image

Picture 3: More moire (say ten times quickly) shown on the corrugated walls. This time it is evident on both sides of the crosshairs and definite ringing can be seen.



Direct link to image

 

Some Nv40 owners don't see it. On the forum of Nvnews i read that a guy stated that this "crap" is spread by some Ati owners, being jaleous not having Sli etc.

 

Wireframe, excellent work. Do you have Bf:1942 or COH? My experience in those games is that using AF makes the sparkle/shimmer worse.

Only increasing LOD and sticking to tri with opts off make it bearable to my eyes.

 

The thing is it is mainly former ATI owners complaining, like myself. Bearing in mind how 'correct Gf4/NV3x filtering' was rammed down ATI owners throats by the fanboys it's laughable that they don't 'see' it.

I'm almost tempted to fo a format and reinstall to see if some 'ATI ghostly residue' is to blame.

 

I don't have either of those titles, but I do have the demos for Battlefield 1942 and one of its expansions. I might give that a whirl later, but I'd be more interested in doing so if I knew the variables oiverning texture filtering can be controlled from within the game or by editing a config file. Otherwise I am not sure I need to see more of what I am already seeing in UT 2003/2004 and Painkiller.

Yes, AF makes it worse. This is extremely evident in UT 2003 where, if I set the DefaultTexMipBias to 0, from -0.5, I have no shimmering but it becomes very blurry and out of focus in places. Applying anisotropic filtering by forcing it in the control panel sharpens the textures to what I would call acceptable levels, but you get shimmering. Again, you need to increase the LOD to at least 0.2 to have it clear up, but at this setting you have to use anisotropic filtering or you get very flat, blurry and boring looking textures on some floors.

So, I am in agreement with you that AF triggers more shimmering, but not using it when compensating through LOD manipulation leaves you with poorly defined textures in some areas.

I am curious why UT 2003/2004 would default to a TexMipBias other than 0. Being TWIMTBP titles, I am wondering if this has something to do with making the game look correct for older Nvidia hardware, but perhaps this is not needed or desired on the newer NV4x series. Ahh, but I must resist speculating until I have at least a better grasp of what is going on.

The fact that some 6800 series users are reporting no shimmering has me very puzzled. I suppose they could be playing titles that are not susceptible or less so. They could also not be seeing it or not thining it a problem when they do. They may have lucked out in driver installation history, but I doubt this one. I also doubt this next one, but it would be horrible to find out they are all using some other, newer revision of the hardware itself.

I am using a 6800 Ultra (NV40 revision A1 according to Rivatuner) with ForceWare 61.77.