Catalyst 9.1 news on the spot...

And you are still forced to live without maximum image quality, because ATI ist refusing to allow the user to control texture filtering quality independent from A.I. which is a must for Crossfire. And even if A.I. is disabled the texture filter is still not on par with NVs HQ.

I simply can't understand why this is issue is debated so little. We now have over a TFLOP of computing power and nobody cares about quality.

 

Some games were pretty "meh" in the improvement scale, but HOLY HELL on Crysis...

 

Well didn't "thingy" (ATi guy that I cant remember name of now) say that they

"aren't where we (ATI) want to be in Crysis at present"

It seems now they are!

 

your saying ati are in crisis

 

I'm not sure...this is the 8.12 Hotfix driver I'm using (Vista x64, 4850 CrossFireX):

8.561.8-081216a-073652E-ATI

This was the second 8.12 HF driver I installed, and it works just fine. The first and third 8.12 x64 Hotfix drivers caused me to bsod at boot with the ATIMKDIAG.SYS cpu stop error. I manually got the other two drivers to function by booting into safe mode after the bsod, installing the drivers manually through the device manager, and then setting the ATi External Event service to "disabled," and then rebooting (steps must be followed in that precise order.) At that point I had to reboot a second time so that both 4850's were initialized and would function under CFX. The second 8.12 hotfix is the only one of the three that installs and boots properly with the ATi EE service set to automatic. No other troubles with the driver to report.

The third hotfix driver was supposed to be 081217a, but I never verified this before reinstalling the second hotfix.

So I count three separate hotfix drivers for 8.12 for Vista x64. Anyone encounter any more of them?...

Update:

'kay, for some reason I cannot fathom I have successfully installed the 3rd and final x64 8.12 hotfix driver:

8.561.3-081217a-073402E-ATI

All I did differently is that from the normal desktop I disabled off the ATI EE service, and then installed these drivers normally right over the former driver version as listed above, and rebooted to no bsod's whatever, rebooted a second time successfully with also no bsod's, and checked the CCC Crossfire setting and noted that it was indeed checked, indicating Crossfire is working fine. I then went back into services and set the ATi EE service to automatic, and started it, and everything is now running like it should, AND with the 3rd set of 8.12 hotfix Crossfire drivers running. Finally...

 

It only matters if you can actually tell the difference without using a special benchmarking tool.

Unlike some of Nvidia's and ATI's past angle dependent or brilinear attempts at more efficient AF methods, the one AMD currently uses is pretty much undetectable by the naked eye for almost all people.

That's the main reason why you don't hear anything about it. Because virtually noone can actually see a difference without special tools.

Now whether or not there should be an option to just turn it off is debatable. Nvidia went with the power user/enthusiast method of "more options = better" and put it in. AMD/ATI went with the consumer mainstream thinking of "more options = more confusing" and didn't put it in.

/shrug. If there's enough demand/outcry I'm sure ATI would put it into the drivers. As it is, aren't there 3rd party tools available that allow tinkering with this? Which most power users/enthusiats are probably already using? BTW - I don't actually know if any can since I stopped caring and stopped using 3rd party tweakers about 3 or 4 years ago.

Regards,
SB

 

Why AI sucks:
There is a nice Parallax Mapping Mod for Doom 3 and Quake 4. But if I want to use this mod, I have to turn AI off. And now imagine, you are a CF or X2 owner: What happens, when you deactivate AI?
Next example: You want to play HL2 with AI off, because you heard, you get then a better filtering quality. But wait, you can't: With AI off you disable the Source Engine's Fog bugfix for HD Series.

To test some CUDA stuff I replaced my HD3850 by my 8600GT. And I played HL2 a little bit (HQ of course). Ok, there was a little shimmering, but all textures looked much better than with HD3850. And this is the point: The 8600GT can deliever much better texture quality than HD4870X2.

More options = more confusing? Is it a joke? All, what I want, is an option for "Evil Filtering Optimization" ON or OFF. Not more. I mean: There is an HQ switch for X1000 series in CCC and no one got confused.

 

So true, of more settings confuse you, put down the controlpanel and step away from it!

 

If anyones interested :

Doom3
http://www.xtreme-lab.net/gtx/gtxd3.htm
ps: dont patch doom3 past 1.3

Quake4
http://www.xtreme-lab.net/gtx/gtxq4.htm

 

Not only that but even Nvidia's list of options aren't that vast or confusing at all. "Texturing Quality" isn't really some ambiguous concept. Actually the all or nothing Cat AI approach is much more confusing because you have no idea what's being turned on or off when you flip the switch.

 

I'm glad your eyes are much better than the rest of us.

Even Scott over at Tech Report could find no real world (read game) situation where ATI's texture filtering was worse than Nvidia's (2900 XT vs. 8800 GTX). Even though the standard mip map viewing tools showed that 2900 XT was theoretically worse.

http://www.techreport.com/articles.x/12458/5

But he could still find no real world cases where you can actually tell in real games. I've found the same myself. With Cat AI on or off. Granted it's been a year or so since I've cared enough to investigate it myself so things may have changed. But I really haven't noticed it getting worse so meh... [edit - Also notice that the default quality now is EXACTLY the same as HQ setting in the drivers when you could change it].

Maybe my eyes just aren't that good at small details. Maybe Scott's eyes at Tech Report aren't that good at small details. I believe even Hard|OCP (granted they don't have much of a rep here but they do tend to go into detail over IQ differences) couldn't find real world examples where it was worse.

So I have no doubt that there are probably people with really sensitive eyes that can tell the difference in a game. But apparently most reviewers and most people in general just can't.

As to the sharpness of textures and shimmering or no, I believe someone postulated that it was mostly due to LOD clamping and ATI and Nvidia using different values for this. And that the texture filtering quality had virtually nothing to do with it (with all recent hardware and drivers).

Regards,
SB

 

So, you think, I'm lying?
Ok, this is G80@HQ:

And this is RV770@AI off:


You see the difference? No? Then I help you:

Hm, if there are no differences between G80 and RV770@AI off, how could The Compressonator find some big differences (dark/black = no differences, brighter parts = differences).

And here RV770 AI on vs. AI off:


And here RV770 AI on vs. RV570 AI on:


There are huge differences... Don't trust your eyes. And don't trust screenshots. These Compressonator's pictures aren't perfect, too. What really matters: How can this affect your games? Because you don't stand still, but you move: You see this Evil Texture Noise, if you move.

Remember this?
http://forum.beyond3d.com/showthread.php?p=1187613#post1187613
What happened? Nothing.



I've bought my X1950GT in March 2007 and replaced it by HD3850 in August 2008. In these 17 months I played some games, but the most time HL2 and the Episodes. So when I switch to HD3850, I was surprised and disappointed about what I had to see on my screen: There is a game, which was released in the year 2004, and it looks worser on the HD3850 with its Unified Shader Stuff and 16 "fat" TMUs than on the X1950GT.

 

While I can't say I disagree with the filtering issue (HQ should be an option), those screenshots are an extremely bad example of that if at all.

For instance, a great deal of the differences isn't even in the clarity of the textures but rather how light they are. Second of all those screenshots are worse than bogus since you can easily tell that the way they did the filtering on the alpha textures is also different.

The fact that you need to show a diff comparison just highlights how ridiculous this particular example is.

 

Arnold, there are bound to be some differences in the rendered image regardless of the filtering, i think they even have slightly different gamma settings on nv vs ati, has been as long as i can remember (can't remember which one rendered the darker image, though)

 

Thank, Arnold, for demonstrating how ridiculously picky one has to be to see the difference between G80 HQ and RV770 AI on. If you know what to look for then shimmering while moving can be roughly judged from static screenshots, and it's pretty clear to me that RV770's shimmering is very minor compared to past offenders like G7x.

It is odd, though, that AI affects the fog blending of distant objects. I can understand differences in ATI/NVidia, but not AI on/off because there's no performance gain to be had.

 

Ive never really felt a need to play with Cat' ctrl panel options. Most are left at default and the games I play look awesome and play well.

I can get most of my desired options inside the game's video options window in order to get the experience Im looking for in the game. Ive played with cat it in the past and have either seen little differences or the changes made the gameplay performance worse.

So what I do when I get a new set of drivers is put some options on quality and the rest are set to let the game decide.

 

Agreed.

 

Catalyst AI contains some fixes for legacy apps via shader substitution with incorrect depth fog in Source engine on DX10 hardware being one of them. So if you want to disable texturing optimizations you'll end up with broken depth fog in HL2 and with some other problems in other apps. Even moreso depth fog was broken in HL2Ep1 with or without AI I don't remember for how long (with AI on it's simply isn't there, with AI off it's everywhere).
While one can say that it's a minor problem, I've been really sick of this approach from ATI/AMD since X2 cards appeared, where with AI off you get no Crossfire and half of performance.
They really should make several options in CCC instead of one super-uber-option that controls everything.

 

Arnold Beckenbauer, what are you trying to prove here? Yes, there is a difference in how they actually filter and how the fog is handled. But which one is 'correct'?

 

The performance gains from the news miracle driver set is at least in a small part due to the driver not accepting forced control panel AF in some D3D10-games. instead, it uses the applications default setting. AMD has been able to reproduce this bug and is looking into a possible fix.

http://www.pcgameshardware.com/aid,673266/Practice/AMD_confirms_Crysis_AF_bug_in_Catalyst_Hotfix/

Also, Nvidia seems to think, that soft shadows are too soft sometimes (or a driver bug has slipped by unnoticed) and that Ambient Occlusion should be emulated by a driver instead of using pre-calculated textures.

(unfortunately not yet in english, but you can judge from the mouseovers for yourself).
http://www.pcgameshardware.de/aid,6...n_keine_optimalen_Schatten_in_Half-Life_2_dt/


Welcome to the world of the abolition of confusing switches for the end user & believing in whatever a given company tells you is "superior image quality compared to everything else out there including GOD"