PDA

View Full Version : ATI cheating with AF


Pages : [1] 2 3 4 5

DSC
05-16-04, 12:11 PM
http://www.computerbase.de/artikel/hardware/grafikkarten/versteckspiele_atis_texturfilter/

Original article in German.

http://www.rage3d.com/board/showthread.php?s=&threadid=33759063

http://www.beyond3d.com/forum/viewtopic.php?t=12486

Discussions on Rage3d and Beyond3d.

vampireuk
05-16-04, 12:29 PM
Well we already know neither company is angelic when it comes to this

DMA
05-16-04, 12:55 PM
Well, i think we all learned something after Driverheavens article about the NV40 cheats in 3dmark-03. Not all is what it seems. :D
Give it a few days before you bring out the tar and feathers.

It's gonna be interesting to see though what sites that'll post this as "news", now that it's ATI. I think it's gonna be pretty quiet this time. ;)
No fun when it's not the big evil NV.. (crazy)

smeagol
05-16-04, 01:07 PM
Ati cheating?

Impossible. They don't understand. Ati never cheats. NV does :rofl

jimmyjames123
05-16-04, 01:08 PM
Very difficult to say exactly what is going on at the moment, but there is no doubt that some of the data has been interesting regarding the effect of AF.

See these UT2004 tests in the Firingsquad review, one with 4xAA and one with 4xAA/8xAF:

http://www.firingsquad.com/hardware/ati_radeon_x800/page21.asp

http://www.firingsquad.com/hardware/ati_radeon_x800/page22.asp

As you can see, the 6800U has the lead with 4xAA, and is 20fps faster than the X800Pro at 1600x1200. However, when you enable 4xAA/8xAF, the X800XT takes the lead, and the X800Pro is only 4fps slower than the 6800U at 1600x1200. Also, notice that the X800Pro and X800XT cards only lose 3-4fps with 4xAA/8xAF enabled vs 4xAA, while the NV card takes a much larger hit!

The FarCry data shows similar results as above, with the X800Pro and X800XT only losing about 3-4fps with 4xAA/8xAF enabled vs 4xAA.

On a side note, some reviews have noted that NV's anisotropic filtering does seem to be slightly clearer/sharper than ATI's, even though both are now angle dependent.

smeagol
05-16-04, 01:10 PM
On a side note, some reviews have noted that NV's anisotropic filtering does seem to be slightly clearer/sharper than ATI's, even though both are now angle dependent.
Got a link for that assumption son ?

vampireuk
05-16-04, 01:20 PM
Back down with the attitude or no ring for you

jimmyjames123
05-16-04, 01:21 PM
Got a link for that assumption son ?

http://www.tomshardware.com/graphic/20040414/geforce_6800-44.html

http://www.firingsquad.com/hardware/ati_radeon_x800/page9.asp

jimmyjames123
05-16-04, 01:23 PM
Back down with the attitude or no ring for you

LOL! That was a great one.

Smeagol says in the distance: My precious! Give it to me!

GamblerFEXonlin
05-16-04, 02:24 PM
http://jooh.no/root/omg_3dcards/3D_quality/9590-9800xt_aniso.gif

rTool is a utility that make driver enable trilinear all the way not just on the first mip-level (nearest texture) but all over.

http://jooh.no/root/omg_3dcards/3D_quality/af-test_xt_bilinear.jpg

I asked ATI about it but they sent me a gaming FAQ that didnt answer anything.


No useless posting please, if there where less garbage here i would maybe known why my 9600 was so blurry to my old GF3 ti. talk about downgrade.

-=DVS=-
05-16-04, 02:37 PM
AF was always Angle debendand on ATI cards you campareing 5950 to R9800 , make comparision to GF6800 becouse it got nearly same AF as Radeon Line right now. :screwy: older GF cards did had better AF but that is over....GF6800 no longer got it , or atleast that what reviews showd us.

Dazz
05-16-04, 02:49 PM
ATi an't cheating persay, they now have 3 modes Trilnear, Bilnear & Adaptive. I believe Adaptive is automaticly used as standard as it provides best visiual qulity and speed blend. As such an't really cheatsing as the USER can pick what to use.

Nutty
05-16-04, 03:06 PM
I thought the whole jist of the thing, was that ATI do things different if they detect those coloured mip layers, to other times.

Thats got nothing to do with what type of AF quality you select. Thats trying to show you do 1 thing, when you dont in benchmarks.

Isn't that what the article was going on about ?

Lfctony
05-16-04, 03:24 PM
Here's how I understood it. Ati is now using Brilinear, and turns it off when it detects a program using colored MIP textures. Apparently, this was done to cover up its usage, IMO. So, by enabling color MIP levels, a 9800 will score the same as without colored MIP levels. But, an X800 based card will score lower - 62 -> 49 / 50 -> 37 in 1280 and 1600 respectively - as its forced to use normal filtering. In the article, we can see it drop respectively in performance, while the 9800 doesn't change at all by the use of colored MIP leves.

The article took it a step further, and clocked the X800 P at 9800XT levels to match fillrate. They had to lower the core more though, due to the extra pipelines. The X800P scores exactly as a 9800XT at the given point.

My opinion? ATI is indeed optimizing their AF. I did wonder why the X800 cards take such a minimum impact when enabling AF, even at such high resolutions. I guess this is why.

-=DVS=-
05-16-04, 03:42 PM
Hmm interesting to bad they didn't make quality comparisions (mag) showing how it should look and how it looks ...

Shuna Sassi
05-16-04, 03:43 PM
ATi an't cheating persay, they now have 3 modes Trilnear, Bilnear & Adaptive. I believe Adaptive is automaticly used as standard as it provides best visiual qulity and speed blend. As such an't really cheatsing as the USER can pick what to use.


Then why is this mode being disabled when colored mipmaps are detected?

What I find humorous about the whole situation, true or not, is how DaveB's scampering around quickly trying to disprove this theory...<Damage Control>
There's plenty of negative rumors \ speculation flying around @ Beyond3D about all things NVidia and yet he, the great impartial man, never tries to dispel any it. (infact sometimes he'll interject with a cryptic remark that does nothing but fan the flames of idiocy).
Though, if anything remotely damning about ATi is said, he'll be all over it like a fly on manure....

Oh and stay tuned for the “ATi’s doing it, so now it’s ok” fanb0y remarks from those previously vilifying NV’s actions.

Apologies for the slight OTC.

Have a nice day

smeagol
05-16-04, 03:47 PM
Back down with the attitude or no ring for you
Just asking for a link. Everyone states opinions now a days...
Linkies... precious...

reever2
05-16-04, 03:59 PM
I'm afraid people need screenshots of some sort of quality loss to actually call it a cheat

MUYA
05-16-04, 04:13 PM
http://translate.google.com/translate?u=http%3A%2F%2Fwww.computerbase.de%2Fart ikel%2Fhardware%2Fgrafikkarten%2Fversteckspiele_at is_texturfilter%2F1%2F%23einleitung&langpair=de%7Cen&hl=en&ie=UTF-8&prev=%2Flanguage_tools
Computer base's article.

Nv40
05-16-04, 04:39 PM
i dont need to read any review.. to know some special "optimization" is going on..

http://www.nvnews.net/vbulletin/attachment.php?attachmentid=5890&stc=1


The X800 XT and X800 Pro offer unparalleled performance with anisotropic filtering enabled.
http://www.tomshardware.com/graphic/20040504/ati-x800-33.html



As we will see repeatedly throughout our benchmark section, anisotropic filtering is the greatest strength of the new Radeon X800 cards. Even the "small" X800 Pro beats the GeForce 6800 Ultra hands down in 1600x1200.
http://www.tomshardware.com/graphic/20040504/ati-x800-14.html


There is no surprise for me in all this stellar performance with [high quality] AF ,for me was always dificult to believe that ATI R420 can do free Trilinear AnisoFIltering all The way to 16x. While Nvidia lose some performance -as expected- . the overuse of
BI Linear filtering decrease the IQ in motion ,can't be seen in in game still screenshot and is very cheap in performance . so asking for comparison in still shots is useless.The real issue is that almost all reviewers like Toms/HOCP/ANAnds and many others of the Bigger sites ,(where most people visit) have used for years the control panel [HIGH] quality settings ,because they expect to get exactly what they selected.and that even if REviewers knew how to FOrce the [HIghquality] editing in game registry settings, very few games offers the option to force TRilinear throught APlication preference. So its a near impossible task in the majority of games. And the benchmarking "HIGH quality" AF with ATI cards againts others IHV is fruitless.

Evildeus
05-16-04, 04:45 PM
so ati is cheatin, that's not possible, Ati is not the evil :rolleyes:
Interesting to see how it's done. Good pick BTW Nv40.

SmuvMoney
05-16-04, 04:53 PM
Then why is this mode being disabled when colored mipmaps are detected?

What I find humorous about the whole situation, true or not, is how DaveB's scampering around quickly trying to disprove this theory...<Damage Control>
There's plenty of negative rumors \ speculation flying around @ Beyond3D about all things NVidia and yet he, the great impartial man, never tries to dispel any it. (infact sometimes he'll interject with a cryptic remark that does nothing but fan the flames of idiocy).
Though, if anything remotely damning about ATi is said, he'll be all over it like a fly on manure....

Oh and stay tuned for the “ATi’s doing it, so now it’s ok” fanb0y remarks from those previously vilifying NV’s actions.

Apologies for the slight OTC.

Have a nice day

If this ends up that ATi is using brilinear AF where true trilinear AF is being requested in the x800 series, I'd cancel my X800XT CompUSA preorder in a heartbeat for the same reason I wouldn't touch the nV FX series. I don't like drivers f'ing with app settings no matter who does it - even though others may not feel that way. Hopefully, people can wait for some commentary from ATi - Catalyst Maker, OpenGL Guy, andypski, Dio, et al. on this on why this might be occurring.

OWA
05-16-04, 05:43 PM
It's gonna be interesting to see though what sites that'll post this as "news", now that it's ATI. I think it's gonna be pretty quiet this time.
No kidding. I'm still waiting to see something on a few of the news type sites that post even the ridiculous rumors about nvidia but still haven't seen any mention of this so far.

gordon151
05-16-04, 05:55 PM
Then why is this mode being disabled when colored mipmaps are detected?

What I find humorous about the whole situation, true or not, is how DaveB's scampering around quickly trying to disprove this theory...<Damage Control>
There's plenty of negative rumors \ speculation flying around @ Beyond3D about all things NVidia and yet he, the great impartial man, never tries to dispel any it. (infact sometimes he'll interject with a cryptic remark that does nothing but fan the flames of idiocy).
Though, if anything remotely damning about ATi is said, he'll be all over it like a fly on manure....

Oh and stay tuned for the “ATi’s doing it, so now it’s ok” fanb0y remarks from those previously vilifying NV’s actions.

Apologies for the slight OTC.

Have a nice day

Yeah that ATI fanboy actually helped Quasars claim by doing his own tests and actually finding that ATI is using brillinear optimizations similar to nVidia on the x800.

zer0xp
05-16-04, 06:29 PM
Am i the only one who cannot see any of beckman's posts?