PDA

View Full Version : Image quality comparison at Extreme Tech


Pages : [1] 2 3

Skynet
06-07-04, 02:22 AM
Well worth reading. Here (http://www.extremetech.com/article2/0,1558,1607400,00.asp)

weeds
06-07-04, 05:24 AM
There are some problems with their tests.

1. They used colored mipmaps, which we know forces true trilinear on the x800, so a real world comparison it is not.. Its not what is being done in games..

2. They used forceware 61.11 which full trilinear is broken, thats why the full
trilinear on looks the same as it does useing brilinear opt.

The slide show was cool, but otherwise flawed comparisions

mikechai
06-07-04, 05:42 AM
There are some problems with their tests.

1. They used colored mipmaps, which we know forces true trilinear on the x800, so a real world comparison it is not.. Its not what is being done in games..

2. They used forceware 61.11 which full trilinear is broken, thats why the full
trilinear on looks the same as it does useing brilinear opt.

The slide show was cool, but otherwise flawed comparisions

Yep. And they don't even know they should enable AA in farcry game settings instead.
Conclusion: flawed comparisons.

BTW, perhaps reviewer should switch to video capture for image quality comparison ...

jimmyjames123
06-07-04, 08:06 AM
Where did you see that they are using Forceware v61.11?

If they truly did use v61.11 with trilinear optimizations on at all times, then this review actually goes to show that it is arguably fair to benchmark the NV 6800 series cards with trilinear optimizations on vs the ATI X800 cards with their optimized filtering method.

EDIT: it looks like ExtremeTech did use the Forceware v61.11, where trilinear optimizations cannot be disabled. In the review, they mention that these are the drivers that NV supplied to them.

weeds
06-07-04, 08:36 AM
Where did you see that they are using Forceware v61.11?

If they truly did use v61.11 with trilinear optimizations on at all times, then this review actually goes to show that it is arguably fair to benchmark the NV 6800 series cards with trilinear optimizations on vs the ATI X800 cards with their optimized filtering method.

QUOTE:We can see the drastic difference anisotropic filtering makes when we zoom in to 4x magnification on our test scene. Note that the drivers nVidia supplied us for our 6800 cards, ForceWare 61.11, have a bug where anti-aliasing is not applied in Far Cry if you turn it on in the display properties. It must be enabled in the game menu, where the only options are none, low, medium, and high. We chose "medium." This is certainly not a final driver and has not been publicly released, so hopefully this will be addressed before 60-series drivers are officially released.END QUOTE:

http://www.extremetech.com/article2/0,1558,1607382,00.asp

Well you cant compare color mipmaps as ati is useing full trilinear on these and nvidia is not. They should have done it as a optimizations IQ shootout,
and left trilinear out of it.. I could live with that..

Nv40
06-07-04, 08:48 AM
Yep. And they don't even know they should enable AA in farcry game settings instead.
Conclusion: flawed comparisons.

.


yep its unfortunate that Journalists reach conclusions about something they are not fully informed. They even are not aware about the FARcry issue with AA in the control panel. a non-issue if you use the in game control panel. like every reviewer does.

neither they are aware that mipmaps colors aplications shows NOTHING!!!.. since thats the *only place* where FUll Trilinear always work -at all times-,thats the "peculiarity" of the algorithm.

their main message in their article is.. "nothing have happened.. why so much noise about Trilinear optimizations if ATi and NV IQ is the same? "

So they need first to read..

1) the German COmputerbase /3dCenter article.. or understand at least their claims.. the 20%-30% loss ATi suffers only when using mipmaps aplications.

2)the TEchreport article of ATI "saying something ,doing another" which really sumarize very well part of the situation..

3)the second IXbitlabs article about this optimzations ,with real game screenshots wich show abrupt transitions with ATI optimizations and not with Nvidia when doing FUll trilinear in some games.

4)aware that many POpular sites benchmarked NV cards doing what the aplication request -fullTRi *at all times*. but in ATI cards doing what ATi wants. which is Brilinear + texture stage optimizations.

5) and finally but not least be aware of the people that already says can notice that their X800 pro shows mipmaps lines abrupt transitions .even some claim noticeable IQ decrease than their Radeon9800 when forcing FULL trilinear in some games or levels.

maxbero
06-07-04, 11:55 AM
Well said, Nv40

particleman
06-07-04, 12:51 PM
There must be something wrong with their painkiller shots, or a bug in nVidia's drivers or something, it looks like the 6800 is stuck in bilinear in their shots.

I would disagree with NV40 with his ATi assessment that there is huge IQ difference with ATi's optimizations. Tom's and most of the sites that have reported on the trilinear opt, have said it is very hard to see in games, but they should have a option for an unoptimized mode to leave it up to the user. So far I have seen only one person complain about the X800 IQ going from a 9800, and that is from a guy called TED on the beyond3D forums, I wouldn't call a guy on a forum a reliable source though. It's just as likely that he messed something up much like I think Extremetech messed something up with their 6800 painkiller shots.

Skynet
06-07-04, 03:42 PM
[H]ardOCP just did a review of the VisionTek Radeon X800Pro, there are lots of screenshot comparisons between the X800 and the 6800. But I am sure many of you here will find something wrong with that review because it paints the X800Pro is a pretty good light. Review is here. (http://www.hardocp.com/article.html?art=NjI4)

Blacklash
06-07-04, 03:54 PM
[H]ardOCP just did a review of the VisionTek Radeon X800Pro, there are lots of screenshot comparisons between the X800 and the 6800. But I am sure many of you here will find something wrong with that review because it paints the X800Pro is a pretty good light. Review is here. (http://www.hardocp.com/article.html?art=NjI4)

Being as I own and have tested one (Hours of game play) in Far Cry, Painkiller, UT2004, Mafia, LineageII, NWN, Halo, SWG, and benched 3dmark03/aquamark many times, I am going to have to go with Skynet on this one. The only b*tch I have about the X800 is flashing lights on some textures in SWG.(driver bug)

I think everyone here knows I have no malice towards Nvidia.

adenosine
06-08-04, 09:06 AM
yea its pretty bad when casual enthusiasts like us (i read up on things maybe 2-3 times a week.. and then its usually only a few threads on various forums) know more about the situation than the damn reviewer!



extremetech has just been removed from my "sites to visit" list

weeds
06-08-04, 09:51 AM
I dont really have anything against Hardocp, but slight manipulation of screenshots is not playing nice, Example, Farcry IQ

1.In the first comparison above we can see that the VisionTek X800Pro and the ATI X800Pro both look identical. The GeForce 6800Ultra’s light reflection quality on the pipe is not as defined as the one on the X800Pro’s.
http://www.hardocp.com/image.html?image=MTA4NjQ1NDUyMmdJZmdKUDFJdlVfM180X 2wuanBn

2.In the second comparison above the VisionTek and ATI X800Pro is exactly same, but the GeForce 6800Ultra does have a problem rendering the floor smoothly.
http://www.hardocp.com/image.html?image=MTA4NjQ1NDUyMmdJZmdKUDFJdlVfM181X 2wuanBn

3.In the third comparison above you can really see the differences in shader quality. The GeForce 6800Ultra has a duller lighting effect on the pipe and what looks like banding on the pipe and floor.
http://www.hardocp.com/image.html?image=MTA4NjQ1NDUyMmdJZmdKUDFJdlVfM182X 2wuanBn

Does anyone else see whats wrong with the shots??

1.Could it be that the light reflection quality being not as defined because
the shot was taken farther away than the x800 shots??

2.Here we see the 1.1 patch in action, dont mention that 1.0 doesnt do this..
but make sure to mention atis shadows problems as a bug, at this point useing farcry for iq comparisions is just stupid considering both companys drivers are haveing problems.
http://www.hardocp.com/image.html?image=MTA4NjQ1NDUyMmdJZmdKUDFJdlVfM18xX 2wuanBn

3.Again this screenshot is at a different location as the ati shots look at the crate to the right, lighting quality isnt as good if your standing farther from the light.

Hardocp has done this stuff before.. Of course this is all in my opinion :)

saturnotaku
06-08-04, 09:58 AM
Do you know exactly where in the game these shots have been taken? I'd like to load up Far Cry and take a look on my 5700 Ultra.

Nv40
06-08-04, 10:17 AM
[H]ardOCP just did a review of the VisionTek Radeon X800Pro, there are lots of screenshot comparisons between the X800 and the 6800. But I am sure many of you here will find something wrong with that review because it paints the X800Pro is a pretty good light. Review is here. (http://www.hardocp.com/article.html?art=NjI4)


[H]ocp numbers are just an example how you can confuse people about products real potential. remember the many people that believed the Geforce6800ultra was "slighty" faster than the radeon9800Xt at launch.. when they saw [H] site.. the mistake in the review is not their methology.. they want to show gamers experience..the issue is that they use other IHV cards too. with diferent settings.. they should use better Beyond3d style and only compare hardware againts products of the same company..

ATi "optimizations" fits very well with [H] "ways" of benchmarking cards since they dont care about "apples vs apples" comparisons.
they use the sicentific method of if it "looks "similar in a *still screenshot * to their eyes" then its ok.the problem with that analogy is that what looks similar to them ,not necessarily looks similar to others..so can make their reviews extremely subjective. and Brilinear or even bilinear in NV cards will look similar to ATI best IQ ,since once you use AF is more dificult to notice diferences in many games in still screenshots .

here is why their results doesnt show real performance...

1)they doesnt use Rtool for ATI cards,neither edit registry settings or games provide settings for Trilinear and AF.. just CP HQ settings.. so you get there. "Texture stage optimizations".. ("trylinear" in the first texture#0 and bilinear in textures 1-7 the rest")

2)they benchmark NVidia cards with 60.72 and force full trilinear there ,but with ATI cards they allow them to use al their optimizations.. even thought they *know* about the ATI tricks..they simply dont care .. for fair comparisons.they only care about the benchmarks ..

3)finally they dont test highest settings + resolutions.. if people remember the RAdeon9700pro release.. it was not until HIgest settings was used that it was a lot more faster than the geforce4.
and medium settings at times the geforce4 neck to neck.

so that review show How fast the X800pro at medium settings resolutions with "optimizations" againts highest HQ AF settings in NVcards.

Sazar
06-08-04, 10:23 AM
[H] makes it very clear they are looking @ IN GAME image quality actually nv40...

they say that IN GAME the iq does not look different and when you see the still caps presented its the same thing...

they always seek to give you testing with highest res/LOD while remaining fully playable... I know it takes a lot of getting used to but as i get more used to it.. it starts to make more sense...

doesn't mean i like it but i can read it now without going "d'oh"

piranha
06-08-04, 10:36 AM
Where in the review do it say it force trilinear on nVidia? And since they investigate image quality that is really a non issue. I am interested in real game performance and image quality not theoretic performance and image quality. So this kind of reviews tells much more than those 3DMARK and other scores we have learned not to thrust for a second.

And the settings are sure well above medium ;) What is the point testing resolutions the games aren´t playable at???

The 9800 256 versus the 128 at 1600x1200x32 with 4x aa and 8x aniso is 30 % faster than the 9800 128 in Far Cry. Who cares? Can you play the game at that res? You get frames into the single digit at times!

That is exactly what Hardocp try to do. Evaluate real world performance except for theoretic performance. And those "medium" settings are surely enough for showing a difference in performance between the video cards mentioned in current state with current drivers.

Hardocp show real world performance something other review sites fails to do. Sure 3DMARK people may not get much out of Hardocp reviews but gamers sure do. And if we want apple to apple they aren´t exactly hard to find elsewhere are they?

Sazar
06-08-04, 10:45 AM
wot?

you dont want to see benches showing a 400% improvement albeit 4fps to 1fps :eek:

how can that be :)

dan2097
06-08-04, 10:47 AM
Yeah where on hardocp does it actually say they are using full trilinear?

3)finally they dont test highest settings + resolutions.. if people remember the RAdeon9700pro release.. it was not until HIgest settings was used that it was a lot more faster than the geforce4.

As said above what the point of testing with settings which are unplayable, with the 9700 pro the 9700 pro was playable with added af/fsaa. in this case what are you actually suggesting that they start doing 1600x1200 16xaf 6xAA benchmarks with 15-30fps results? I know you dont like those types of benchmarks as Hardocp would logically compare against Nvidias 8xAA as its the nearest AA mode to ATIs 6xAA.

1)they doesnt use Rtool for ATI cards,neither edit registry settings or games provide settings for Trilinear and AF.. just CP HQ settings.. so you get there. "Texture stage optimizations".. ("trylinear" in the first texture#0 and bilinear in textures 1-7 the rest")

I dont think hardocp actually benchs with trilinear although feel free to prove me wrong, and the reason being most users wont enable trilinear, the same with this, most users wont download rtool/radlinker. If hardocp actually benched a game where ATIs texture stage optimizations made a noticeable difference with the settings they were using they would comment on it. In many games trilinear on the first stage is sufficient.

Yes the review is subjective but from other reviews you get no idea what settings offer playable performance especially as min fps is usually not displayed. I mean the reason why I think it was Bf1942 vietnam is benched on lower settings than the X800 pro is that a certain part cuts the 6800s down to like 15fps, that sort of experience can only be gained by actually playing games on the cards which from many reviews you have no idea whether they have done.

Nv40
06-08-04, 10:54 AM
In FarCry..We did encounter that one shadow bug with shadows on “Very High”. We found out though that you can fix it by setting shadows to “Medium” and at the same time you get a large performance boost in FarCry on the X800Pro. Other than that the VisionTek X800Pro provided the best image quality in the game.


With Battlefield Vietnam ...we found 1600x1200 with 4XAA and 8XAF to be the best quality settings on the VisionTek X800Pro. We still experienced the texture filtering bug when Anisotropic filtering was enabled. You can fix this by either setting the quality settings to “Low” or not enabling Anisotropic filtering, but both options severely reduce image quality.

In Painkiller ....we found 1280x1024 with 4XAA and 8XAF to be playable through every level in this game. There was a noticeable difference in Anisotropic filtering however. The GeForce 6800Ultra returned a slightly sharper texture upon close examination.

Splinter Cell ...is a very demanding game and unfortunately does not support Anti-Aliasing. We found 1280x1024 with 8XAF to be the highest playable setting on the VisionTek X800Pro. As you can see from the graph performance did start out low, but increased along the way. Shadows do work correctly on the X800Pro video cards, however there is a sky glow that is not working so again due to "bugs," the GeForce 6800Ultra pulls out on top.



ALbeit [H] doesnt mind to use lower IQ settings or optimizations in ATI hardware ,they force [non optimizations] and higher quality settings in NV cards. They clearly notice Lower IQ in X800 hardware in most of the games,but still believe the X800pro hardware is the faster like in their previous reviews.. any relation of IQ with the performance is merely a coincidence :rolleyes:

if they *really* want to show "gamer experience" but using diferent settings in diferent hardware HIgher here ,lower ,"optimizations"here and not there ,there they should simple drop comparisons between diferent IHV. and focuse only with products of the same company (unless they are not really interested to show gamer experience like they say but the benchmarks) just like other sites .otherwise they are misleading the average gamers (the majority )that doesnt visit forums or doesnt understand too much about video cards.

Sazar
06-08-04, 11:08 AM
They clearly notice Lower IQ in X800 hardware in most of the games,but still believe the hardware is the faster .. any relation of IQ with the performance is merely a coincidence :rolleyes:

if they want to show "gamer experience" but using diferent settings in diferent hardware HIgher here ,lower there they should simple drop comparisons between diferent IHV. and focuse only with same company products..just like other sites.otherwise they are misleading the average gamers (the majority )that doesnt visit forums or doesnt understand too much about video cards.

open your eyes and read again nv... you're going back to your hyperventilating troll days now...

[h] clearly benched twice... there are 2 sets of results wrt the shadows ... with it on high and with it on medium.. had you actually bothered to spend half as much time reading through instead of constantly jumping to conclusions you may well be taken more seriously...

http://hardocp.com/images/articles/1086454522gIfgJP1IvU_2_3_l.gif

http://hardocp.com/images/articles/1086454522gIfgJP1IvU_3_3_l.gif

You can clearly see huge gains in performance with Shadows on Medium. In the second half of the graph starting at 81 seconds and on is when we are closest to and around trees in the forest in Fort. As you can see this is where we gained the most performance.

So it seems that Shadows greatly effect the performance of the X800Pro, and if you need some more performance and want to sacrifice Shadow quality a bit, which by our evaluation wasn’t much difference at all, then set Shadows to “Medium” and enjoy the added performance.

um oh gee-whiz... did someone jump to conclusions too quickly?

instead of purposely posting FUD and wasting people's time m8 consider reading through the very things you want to point out to people please...

and let me include a :rolleyes: to make you feel better too... thank you and good day...

-edit-

picking and choosing selective portions of the reviews comments also does not make a case better though undoubtedly it is easier to manipulate someone to see one's POV if they can't be bothered to read for themselves..

evilchris
06-08-04, 11:16 AM
ALbeit [H] doesnt mind to use lower IQ settings or optimizations in ATI hardware ,they force [non optimizations] and higher quality settings in NV cards. They clearly notice Lower IQ in X800 hardware in most of the games,but still believe the X800pro hardware is the faster like in their previous reviews.. any relation of IQ with the performance is merely a coincidence :rolleyes:

if they *really* want to show "gamer experience" but using diferent settings in diferent hardware HIgher here ,lower ,"optimizations"here and not there ,there they should simple drop comparisons between diferent IHV. and focuse only with products of the same company (unless they are not really interested to show gamer experience like they say but the benchmarks) just like other sites .otherwise they are misleading the average gamers (the majority )that doesnt visit forums or doesnt understand too much about video cards.

[H] is a joke. Even if they were on NVIDIA's tip I'd still think their review style is BS. They are trying to be "cool" by being "different". Look at the BF:V apples to apples scores. When the 6800 dipped into the teens on zooming ( probably a driver bug ) but SMOKED the X800 in 16x12 everywhere else, they dropped its "playability" down to 1280. Yet ATI's card has rendering errors and all sorts of other crap but they don't drop its "playability" setting for that. Tools.

Nv40
06-08-04, 11:18 AM
picking and choosing selective portions of the reviews comments also does not make a case better though undoubtedly it is easier to manipulate someone to see one's POV if they can't be bothered to read for themselves..



picking selective settings in diferent Hardware Higher here or lower ,optimizations here and not there is not the best way to show real performance of video cards.. Doing this is manipulation of benchmarks.. is that simple. there is no diference between [H] style than (DH.) "style"of 8xS AA in NV cards vs 6xAA in ATI and still giving the crown in performance to ATI hardware.. :screwy:

piranha
06-08-04, 11:20 AM
Sometimes you wonder if NV40 is part of the nVidia FUD or market PR team? You really sound exactly like those sales guy who are trained to death delivering the right messages to sale your product.

Oh and according to Brent they used brilinear for the 6800 ULTRA...

Sazar
06-08-04, 11:20 AM
picking selective settings in diferent Hardware Higher here or lower ,optimizations here and not there is not the best way to show real performance of video cards.. Doing this is manipulation of benchmarks.. is that simple

they are trying to achieve a setting they believe is a playable level for gaming.. and yes wrt what they are doing it is manipulation but consider what you are posting as well.. where is your response to the far cry results?

Nv40
06-08-04, 11:25 AM
they are trying to achieve a setting they believe is a playable level for gaming.. and yes wrt what they are doing it is manipulation but consider what you are posting as well.. where is your response to the far cry results?

i dont trust in "believing" as an accurate way of benchmarking..
pay close attention at what [H] "believes"..

http://www.hardocp.com/image.html?image=MTA4MzU2NDE4OTg4OEFkazcwdGVfOF8yX 2wuZ2lm

a radeon9800Xt faster than a geforce6800ultra? mmm..
and x800pro near Twice the frame rates at highest settings?
and thats without a doubt what they believe is the "Gamers experience".. :rolleyes: and FArcry is were more people complain with shimmering of textures and lower IQ with ATI Trilinear "optimizations" ,there are videos ..but also the bugs in shadows .. albeit NV cards also have issues with precision there using the NV3x path with IQ.. so until NVdrivers fix those .. and ATI fix them nothing is conclusive in that game. ;)

BTw.. this thread was about Extremetech findings ..so sorry about the off topic discussion.. :)