toms hardware puts up article on ati's filtering game

[reever]

somehow OK when you used to flame nVidia to Hell and back for this sort of thing?

But then how could you flame Ati to hell and back when you were the one defending Nvidia for this sort of thing?

 

[chsh1ca]

Originally posted by: CaiNaM
i did a 15-20 sec movie capture in fraps which shows this. it doesn't stand out quite like it does in the game as the image is sized down, however you can definately see what i'm talking about. unfortunately, the .avi is 75mb (zipped - over 100min uncompressed), and dubbing to divx blends the texture in such a way it's useless to use as an example..

Doh. If I had my FTP server up, I'd set you up with an account, unfortunately it's out with a bad case of the dead hard drives.

edit: i did put up a sshot where you can see the line where the texture detail changes. again, in a still shot, it's not easily noticed, but obvious when you're moving as the line moves directly in front of you (there's actually 3 lines @ 4xaa/16xaf). it's a 2.8mb bmp, but if you have broadband you can view it here.

Yeah, it's pretty obvious there's a line there even without your highlighting it for the viewer's benefit. The issue at the heart of this is showing there was degradation period. I for one would like to see an option to force trilinear just so we can do the quality comparisons and see if that banding in the screenshot you provided is a result of their "trylinear" mode or something else.

Unfortunately, like NVidia, it's likely ATI won't have an option to run full trilinear in the drivers for a year or more, long enough for everyone to have forgotton it. How many people will bother going back and doing FX5800/FX5900 reviews with UT2K3 with those new dets that are purportedly going to offer the ability to enable full trillinear vs brilinear? Probably none. I think you can file this in the same category a year from now.

As for whether what ATI has done is despicable or not, think about it for a second. Both the major players are always telling reviewers to run their hardware in a favourable light, this isn't exactly something new.

Rollo, on your point #4 regarding ATI knowing what is best and etc., it could be that this is how the trilinear filtering algo is implemented on the hardware, in which case it is well more than ATI programming its drivers.

 

[GeneralGrievous]

Since ATI does trilinear on colored mipmaps, the hardware is certainly capable of it.

 

[chsh1ca]

Indeed, however that fits within the algorithm they discussed in the media. The issue is whether it can globally apply trilinear always, rather than defaulting to it when it thinks it needs to.

 

[rbV5]

here some actual avi files you can download, from a guy named tEd. He also knows how to disable try-linear, but wont tell how.

Why won't he tell?

 

[Pete]

Originally posted by: sandorski
It's not an issue of whether he is honest or not. It's all too common for people to attribute one thing as the reason for an issue they may be experiencing. I know I have done it and many threads posted regularly place blame on certain things incorrectly all the time. I'm sorry I am not as accepting of it as others, but as I said ad nauseum already, someone needs to Prove it.

Very obvious mipmap banding has been shown by two separate people at B3D with videos of MP2 and FC. IIRC, they said it wasn't obvious in all games all the time, but obviously there are at least two proven visible defects. Let's see if ATi can correct these "edge cases," or just insert a switch to disable trylinear.

I don't want to provide direct links b/c I think AT would basically rip those personal servers a new one (akin to the /. effect), but you can find them both at the Beyond3D forums. Look for "x800 texture shimmering: FarCry video" in 3D Tech and "we are not done yet ati " in 3D Graphics Boards. I can see clearly see shimmering around the mipmap boundaries in both. Granted, this is just early evidence from less than a handful of sources, but there is visual proof out there.

 

[Shamrock]

Originally posted by: rbV5

here some actual avi files you can download, from a guy named tEd. He also knows how to disable try-linear, but wont tell how.

Why won't he tell?

he posted his reason at NVNews.net, but it's too vague to care about

tEd:
no that's just me being stubborn for now

seriously there are other reasons why i don't like to reveal such reg tweaks which is not in anyway related to ati.

 

[BFG10K]

It does do preset algorithms that are detected, then applied. the drivers are not "smart"

The drivers are the ones analyzing the situation and they make the decision when to activate the adaptive trilinear.

I can see clearly see shimmering around the mipmap boundaries in both.

Are you sure that just isn't a symptom of low quality/resolution videos?

 

[Pete]

No, as the shimmering is not in evidence with a 9700 in the FC video and with AF optimizations off in the MP2 video. Check 'em out yourself, they're pretty obvious.

Of course, both video authors have said these cases are somewhat rare. The fact that they exist, though, is an issue that should prompt either the ability to disable the optimizations or some trylinear improvement (hey, maybe ATi really can take care of most/all corner cases with a refined algorithm).

 

[CaiNaM]

Originally posted by: BFG10K

It does do preset algorithms that are detected, then applied. the drivers are not "smart"

The drivers are the ones analyzing the situation and they make the decision when to activate the adaptive trilinear.

I can see clearly see shimmering around the mipmap boundaries in both.

Are you sure that just isn't a symptom of low quality/resolution videos?

yup, it's clearly visible when moving, and it's clearly visible in the screenshot i linked to earlier. r420 does a very poor job of smoothing the mip transitions compared to nvidia, and even compared to my 9800pro.

it's my hope this situation can be rectified via drivers if ati receives enough "heat" from this issue.

 

[Matthias99]

Originally posted by: CaiNaM

Originally posted by: BFG10K

It does do preset algorithms that are detected, then applied. the drivers are not "smart"

The drivers are the ones analyzing the situation and they make the decision when to activate the adaptive trilinear.

I can see clearly see shimmering around the mipmap boundaries in both.

Are you sure that just isn't a symptom of low quality/resolution videos?

yup, it's clearly visible when moving, and it's clearly visible in the screenshot i linked to earlier. r420 does a very poor job of smoothing the mip transitions compared to nvidia, and even compared to my 9800pro.

it's my hope this situation can be rectified via drivers if ati receives enough "heat" from this issue.

From today's digit-life filtering article:

2.The situation with anisotropic filtering is more difficult. First, ATI plays an unpleasant trick when anisotropy is forced from drivers. Filtering is only based on bilinear samples, so, despite a clearly increased sharpness, borders between MIP levels are unpleasantly obvious. While NVIDIA smoothes these borders with partial trilinear filtering (see pictures with coloured levels), ATI shows clear borders in all modes including Quality. But when filtering degree is set from the application and not forced from the drivers, things get back to normal. So, what's the matter? Such problems have been repeatedly observed in game applications. If an application can't manage anisotropy itself, it can be forced from driver settings, and everything will be OK. If an application can do it, then we disable anisotropy in the drivers and activate it from the application. But if anisotropy is activated in both the drivers and the application, we get sharp borders between MIP levels, annihilating all visual quality of anisotropic filtering.

Is it possible you're getting this effect? I mean, the shots you posted look like it's not doing *any* trilinear. Can you turn on AF from within the game? They also suggested this workaround if you can't:

Alexey Barkovoy's note: It is the control panel that is to blame for this in the case of ATI. If you move the Performance/quality slider from the 3D tab to the right and thus select AF/AA forcing, trilinear anisotropic filtering will always be selected in the first texture, and bilinear anisotropic filtering in others. And after you "play" a bit with the slider (select "use custom settings" again and than AF separately), everything will work all right and trilinear filtering won't disappear at anisotropy forcing.

They should definitely have an option to disable this, though (if only to see how it differs from 'full' filtering in various situations!) -- and they need to fix their AF forcing, too.

 

[CaiNaM]

texture quality is set to full, and af is set to 16x.

 

[Matthias99]

Originally posted by: CaiNaM
texture quality is set to full, and af is set to 16x.

But did you force it in the drivers, thus possibly overriding it with a trilinear first stage and bilinear on every other texture stage?

 

[CaiNaM]

Originally posted by: Matthias99

Originally posted by: CaiNaM
texture quality is set to full, and af is set to 16x.

But did you force it in the drivers, thus possibly overriding it with a trilinear first stage and bilinear on every other texture stage?

it's not "adjustable" from within the game....

 

[Matthias99]

Originally posted by: CaiNaM

Originally posted by: Matthias99

Originally posted by: CaiNaM
texture quality is set to full, and af is set to 16x.

But did you force it in the drivers, thus possibly overriding it with a trilinear first stage and bilinear on every other texture stage?

it's not "adjustable" from within the game....

Hrmmm. Well, there's a registry hack that's supposed to disable the trilinear optimizations with the 4.6 beta drivers... could you give that a shot and see if it helps? It'd be nice to know if this is specifically what's causing that IQ problem, or if it's some other issue.

Also, does this show up everywhere in DAoC, or just in certain areas? The problems reported in other games (such as Max Payne 2 and Far Cry) seem to be isolated to particular areas/textures...

 

[Myrdhinn]

Originally posted by: Shamrock
round 2 for all the naysayers

http://www.ixbt.com/video2/nv40-rx800-5-p1.shtml It has comparison videos, and PLAINLY shows the differences...ATI's are DEFINTELY lowering IQ.

Excuse my poor translation but in the conclusion of that article they state...

"Therefore admirers " purists " (and they, as a rule, admirers either ATI, or NVIDIA) as though were not indignant, but it is necessary to tell, that the lion's share of users in the majority of games (certainly where there are no obvious glitches or artefacts) because of all these optimizations will not see ANY difference."

*note the "ANY"..*

So no matter how you slice and dice both cards are pretty equal and I'd be happy to own either one, all the other serious sites seem to agree on that. The only difference being NV had to totally invent new hardware to slightly beat ATi in some cases so I'd say kudos to ATi on this round.

 

[nitromullet]

CaiNaM: If you're still reading this thread, I have a request...

Will you give the registry tweak to enable full trilinear a try?

http://forums.anandtech.com/messageview.cfm?catid=31&threadid=1327006&enterthread=y

I would be curious to know if the full trilinear registry change fixes the IQ degredation that you have seen in game play. By the way, I have an XT on preorder and I appreciate your honesty in regards to your Pro. You certainly are taking quite a bit of heat for paying $400+ for a video card and being blunt about it's shortcomings.:beer:

Edit: looks like Matthias99 already requested this...

 

[Robor]

I agree ATI should've announced this feature so it wouldn't have looked like they were trying to hide something but IMO if it improves speed without degrading IQ then it's a feature, not a cheat. And as far as benchmarks go if Nvidia can't produce equal or better IQ without trilinear enabled then they need to have it enabled when they benchmark.

 

[nitromullet]

Originally posted by: Robor
I agree ATI should've announced this feature so it wouldn't have looked like they were trying to hide something but IMO if it improves speed without degrading IQ then it's a feature, not a cheat. And as far as benchmarks go if Nvidia can't produce equal or better IQ without trilinear enabled then they need to have it enabled when they benchmark.

The fact of the matter is that they did try to hide it - it doesn't just look that way. Hopefully, everyone can agree on this by now. In addition, an owner of the card has stated that they do see IQ degradation during gameplay. Whether or not this should have an effect on your decision to go w/ ATi or nVidia remains a different story and will probabaly include other factors, such as price/availability.

 

[BFG10K]

Check 'em out yourself, they're pretty obvious.

I don't want to suck them down on a low-bandwidth pipe.

The fact that they exist, though, is an issue that should prompt either the ability to disable the optimizations.

I'm almost certain the next drivers will include a setting to disable it as there are reports of registry keys that can control it in the current beta drives.

yup, it's clearly visible when moving, and it's clearly visible in the screenshot i linked to earlier.

Okay.

 

[CaiNaM]

Originally posted by: Matthias99

Originally posted by: CaiNaM

Originally posted by: Matthias99

Originally posted by: CaiNaM
texture quality is set to full, and af is set to 16x.

But did you force it in the drivers, thus possibly overriding it with a trilinear first stage and bilinear on every other texture stage?

it's not "adjustable" from within the game....

Hrmmm. Well, there's a registry hack that's supposed to disable the trilinear optimizations with the 4.6 beta drivers... could you give that a shot and see if it helps? It'd be nice to know if this is specifically what's causing that IQ problem, or if it's some other issue.

Also, does this show up everywhere in DAoC, or just in certain areas? The problems reported in other games (such as Max Payne 2 and Far Cry) seem to be isolated to particular areas/textures...

i wouldn't say particular areas as much as i would say particular textures... there's certain textures for the ground that don't have alot of detail, and it does not stand out in those areas - only in areas that have a lot of texture details, like in areas where the ground is made up of pebbles/rocks or something similar.

as for the beta drivers, i'll wait for the 4.6 release and try it; getting tired of swapping drivers heh. they are supposed to be out in short time.

 

[CaiNaM]

Originally posted by: nitromullet
CaiNaM: If you're still reading this thread, I have a request...

Will you give the registry tweak to enable full trilinear a try?

http://forums.anandtech.com/messageview.cfm?catid=31&threadid=1327006&enterthread=y

I would be curious to know if the full trilinear registry change fixes the IQ degredation that you have seen in game play. By the way, I have an XT on preorder and I appreciate your honesty in regards to your Pro. You certainly are taking quite a bit of heat for paying $400+ for a video card and being blunt about it's shortcomings.:beer:

Edit: looks like Matthias99 already requested this...

lol.. a second request.. perhaps i'll try again tomorrow. little grumpy as i put a new heatsink/fan in my wife's rig, booted windows, and it bluescreened right after the login prompt. now it just hangs at "verifying dmi data" and won't boot. grrrr.....

 

[CaiNaM]

Originally posted by: BFG10K

Check 'em out yourself, they're pretty obvious.

I don't want to suck them down on a low-bandwidth pipe.

The fact that they exist, though, is an issue that should prompt either the ability to disable the optimizations.

I'm almost certain the next drivers will include a setting to disable it as there are reports of registry keys that can control it in the current beta drives.

yup, it's clearly visible when moving, and it's clearly visible in the screenshot i linked to earlier.

Okay.

one thing i'd like to make clear is that while it's a problem, i for one am being quite picky. it's not like it ruins gameplay. however, given the current trend (and the fact these are $400 cards i suppose), i only feel fair we give this as much attention as the ati camp did to nvidia, depsite the fact i felt it was being a bit blown out of proportion - brilinear really didn't bother me that much, although the mipmap transisition didn't seem as obvious as here, tho perhaps it's just my memory fading a bit - my nv35 has been gone almost a month now

 

[Matthias99]

Originally posted by: CaiNaM

Originally posted by: Matthias99
Also, does this show up everywhere in DAoC, or just in certain areas? The problems reported in other games (such as Max Payne 2 and Far Cry) seem to be isolated to particular areas/textures...

i wouldn't say particular areas as much as i would say particular textures... there's certain textures for the ground that don't have alot of detail, and it does not stand out in those areas - only in areas that have a lot of texture details, like in areas where the ground is made up of pebbles/rocks or something similar.

Thank you for clarifying. As I don't have an X800 and I don't play DAoC (though I'm rapidly getting hooked on City of Heroes...), I'm trying to get as much info as I can about the behavior of these new cards. It sounds then like this *is* probably related to their filtering optimizations -- there seems to be something about certain textures that causes it to not work properly (or, rather, to apply a too-low level of filtering to them). Did you file a bug report with ATI?