ATI caught cheating in Crysis to outperform GTX480

[randomizer]

Now that the wowser thread title has got your attention... let's hope the driver was bugged.

http://www.pcgameshardware.de/aid,743333/Geforce-GTX-470-und-GTX-480-Test-von-Nvidias-GF100-Generation/Grafikkarte/Test/?page=3

 

[shadow187]

Wait...what? ;-;

 

[maximiza]

ATI needs to be spanked... naughty, naughty ATI

 

[AMW1011]

I didn't see anything about that in the link, maybe someone can explain it?

 

[Mousemonkey]

I tried running that page through the google translator and it really messes up the grammar. :lol: Anyone care to summarize? All I could work out was something about dropping frames or some such.

 

[AMW1011]

Yeah same here MM, I tried two different translators.

 

[L1qu1d]

.....k we can all close our eyes for this, cuz Nvidia did it with 3Dmark I believe with the 8800s

 

[Mousemonkey]

Yeah same here MM, I tried two different translators.

Does your head hurt now as well? :lol: I'm going to sink a couple of beers and try again to see if it makes more sense when I'm half cut.

 

[flyinfinni]

Umm... what? lol. I'm totally with you MM! I can't get through that page at all, even with various translators. Its making my eyes hurt lol

 

[Bluescreendeath]

From what I'm getting from google-translate, I think what the article is saying is that ATI is using a jerry-rigged version of AA/MSAA/etc in Crysis, while the GTX480 is using the actual AA that looks much better... ?

 

[roofus]

If there is any truth to the claim, the news will make its way around. Nvidia would make sure it did for their own sake. I have my doubts though.

 

[Mousemonkey]

If there is any truth to the claim, the news will make its way around. Nvidia would make sure it did for their own sake. I have my doubts though.

Or they are just sitting back and giggling their socks off at the concept of the competition having to skew the results when faced with what is effectively the "broken" version of the GF100's. [:mousemonkey]

 

[JackNaylorPE]

Tweaking drivers to fit bench scripts is nothing new....I'd say it's mostly responsible for the truth behind the "driver improvements will increase benchies" mantra.

 

[notty22]

Tweaking drivers to fit bench scripts is nothing new....I'd say it's mostly responsible for the truth behind the "driver improvements will increase benchies" mantra.

Exactly, if you look at Anands launch review of the 5870 and the new #'s generated when compared with the Gtx 480, many are within a frame or 3. And thats 6 revisions ! They are able to say (without lying) a 5 or 7% in Driver notes, increase as long as there is that at some unspecified resolution and quality setting. The next month , they tweek for less AA and a higher resolution and they magically can claim a new 5% increase !. I'm not pointing fingers at either side, but if you added up these %'s in every driver release the number would not make sense.

 

[Mousemonkey]

Tweaking drivers to fit bench scripts is nothing new....I'd say it's mostly responsible for the truth behind the "driver improvements will increase benchies" mantra.

Exactly, if you look at Anands launch review of the 5870 and the new #'s generated when compared with the Gtx 480, many are within a frame or 3. And thats 6 revisions ! They are able to say (without lying) a 5 or 7% in Driver notes, increase as long as there is that at some unspecified resolution and quality setting. The next month , they tweek for less AA and a higher resolution and they magically can claim a new 5% increase !. I'm not pointing fingers at either side, but if you added up these %'s in every driver release the number would not make sense.

:lol: That's why I love taking a poke at that little mantra whenever it gets trotted out, which (in a spooky kind of way) seems to coincide with every release of a new card from ATi.

 

[megamanx00]

Umm... I don't read German, and it doesn't translate well ^_^. Anyway, I guess I'll wait and see if any other hardware sites reach this conclusion, using methods I can read, before drawing a conclusion.

 

[flyinfinni]

+1 to that.

 

[Shadow703793]

Perhaps one of the Mods may want to contact the Toms Hardware Editors and see what they can dig up... they have too much free time imo.

 

[randomizer]

I think the issue is more with the AF than the AA. It seems that there's some sort of bilinear filtering being used from the images nearer the bottom. But yea, translators don't do much for me on this page so I didn't link it to one. There's some discussion on it at XS as well, which is where I got the link from: http://www.xtremesystems.org/forums/showthread.php?t=248755

I am not sure why people say they can't see a difference. It's small but it is there, and it will have some impact on framerate.

 

[shadow187]

Can someone please explain in a summarized form what's happenin'?

 

[builderbobftw]

Even though ATI cards run faster, the image quailty is worse.

But no one noticed this but one german website weeks after the fact.

 

[notty22]

It takes some very 'smart' and in the know people to catch this stuff thats why.
Long story short , it allows the gpu to work less, increasing performance. Best explanation seems like this post
http://www.xtremesystems.org/forums/showpost.php?p=4320571&postcount=42

For all the people saying there is no difference so its not an issue, if nvidia did this everyone would go on and on about cheating...

bilinear filtering does 5 texel lookups per pixel, trilinear does 10 and then does a linear interpolation between 2 colors, anisotropic uses the view angle to vary the sampling range of the texels, each anisotropic level enlarges the sampling range(non-uniformly) and so the texels sampled increase, it then applies trilinear filtering with the new ranges. This allows the textures to look smooth even when moving through a scene and as a result heavily anti-aliases textures and reduces pixel swimming but comes at a heavy cost.

For example in a standard 1680x1050 scene, the texel lookups counts are as follows:

bilinear: ~8.82 million texel lookups
trilinear: ~17.64 million texel lookups
Anisotropic: (depending on AF level) potentially more than 4x the texels of tri-linear depending on the AF level.


In a still image there is no problem, the major filtering artifacts are movement based artifacts such moire patterns and pixel swimming and mip-level pops, static images will never really show the full picture when it comes to filtering. I cant believe ATI would do this!

I'd would love someone used the 10.3s and the previous versions and see if there is a difference, that nice 10% increase across the board the new drivers bring might be

 

[builderbobftw]

Lost me at "trilinear filtering".

 

[EXT64]

Would "cheating" in AF really make a huge difference in performance? I thought the performance hit was pretty small.

 

[randomizer]

Doing trilinear filtering vs AF would make a difference.