8800GT falls apart at 8XAA

[Distant]

Sadly whether intentional or not, whether Nvidia bought all these wins or they just got them out of the pure ignorance of hardware reviewers it is happening.

What am I talking about? Surely you saw the latest anandtech 8800GT article and surely if you are like most gamers and have a 19/20 inch monitor looking at the numbers these cards put out you'd probably imagine at a resolution of 1600 by 1200 or lower you seem to have more then enough head room to go up to the 8x AA level right?

So why all this testing on monitors for 24/30 inch monitors? Certainly it's nice that anandtech tests to those levels, but I think it's more then obvious that 8xAA tests at resolutions of 16x12 and under are of much more use to PC gamers then what a these cards push out on 30 inch monitors, monitors.


So what is this 8xAA testing I speak of?


http://www.computerbase.de/art..._radeon_hd_3850_rv670/

You can clearly see in the 8 tests that they did do 8xAA on

Anno1701
Clive Barkers Jericho
FEAR
Oblivion
Prey
Company of heroes DX9 and 10
Lost Planet

In every single one of those with the exception of FEAR and Company of heroes in DX9 mode the 8800GT's framerates literally drop like a freaking rock, in some cases it's performance getting cut in more then half while the 3870's really takes much of hits at all and because of this the 3870 overtakes the 8800GT at this level of AA.

Now call me crazy, but I think most of us don't have a 24+ inch monitor and if you do you really need two cards anyway, my point is looking at the frame rates in most games your going to be wanting to game at 8xAA and if you just read this article you wouldn't have known that the 8800GT appears to be garbage at high levels of AA

 

[imported_Scoop]

If you're gaming with 1600x1200 or up there's hardly a need for any level of AA.

EDIT: WAIT! was this thread a joke?

 

[AzN]

Maybe if you had a CRT but LCD is more noticeable if you don't add some kind of AA even high resolutions.

 

[Keysplayr]

I see where your coming from, but at such high levels of AA, and smaller memory bandwidth, I think it's expected that framerates will tank. We aren't even sure which mode of AA was used. Unless I missed that detail. And I noticed that Transparency AA was Aus/On for Nvidia cards and Adaptive AA was Off for the ATI models. Does this make a difference? Unsure. Anyone know?

Ahh, QxAA. But still, what about transparency/Adaptive?

 

[error8]

Originally posted by: Distant
Sadly whether intentional or not, whether Nvidia bought all these wins or they just got them out of the pure ignorance of hardware reviewers it is happening.

What am I talking about? Surely you saw the latest anandtech 8800GT article and surely if you are like most gamers and have a 19/20 inch monitor looking at the numbers these cards put out you'd probably imagine at a resolution of 1600 by 1200 or lower you seem to have more then enough head room to go up to the 8x AA level right?

So why all this testing on monitors for 24/30 inch monitors? Certainly it's nice that anandtech tests to those levels, but I think it's more then obvious that 8xAA tests at resolutions of 16x12 and under are of much more use to PC gamers then what a these cards push out on 30 inch monitors, monitors.


So what is this 8xAA testing I speak of?


http://www.computerbase.de/art..._radeon_hd_3850_rv670/

You can clearly see in the 8 tests that they did do 8xAA on

Anno1701
Clive Barkers Jericho
FEAR
Oblivion
Prey
Company of heroes DX9 and 10
Lost Planet

In every single one of those with the exception of FEAR and Company of heroes in DX9 mode the 8800GT's framerates literally drop like a freaking rock, in some cases it's performance getting cut in more then half while the 3870's really takes much of hits at all and because of this the 3870 overtakes the 8800GT at this level of AA.

Now call me crazy, but I think most of us don't have a 24+ inch monitor and if you do you really need two cards anyway, my point is looking at the frame rates in most games your going to be wanting to game at 8xAA and if you just read this article you wouldn't have known that the 8800GT appears to be garbage at high levels of AA

So you signed up on anandtech forums just to show us how bad 8800 gt is ? Well first of all, nobody really uses such a high level of AA, because you won't see any difference above 4X anyway, no matter what resolution is being used. The second thing is that at 8X AA and above there is some blur added that lowers image quality a bit.
I know you love ATI , but just try to understand that the 8800 GT just scales better than the HD3870 in 99% of the games out there, with all the eye candy turned on, except for 8X AA which doesn't really matter, because both card produce low fps at this setting.

 

[Capt Caveman]

I play with BF2 at 1600x1200 with 8xAA and high settings.

 

[Ackmed]

From another thread, nicely typed up, so Ill just paste it.

Originally posted by: tuteja1986
DX 9 benchmark

Anno 1701 1600x1200 4xAA/16xAF
HD3870 : 35,7
8800GT : 31,3
HD3850 : 30,6

Clive Barker's Jericho 1600x1200 8xAA/16xAF:
HD3870 : 15,6
8800GT : 5,6
HD3850 : 1,3

Colin McRae Dirt 1600x1200 4xAA/16xAF
HD3870 : 15,8
8800GT : 4,5
HD3850 : 4,3

Company of heroes 1600x1200 8xAA/16xAF
8800GT : 64,2
HD3870 : 58,7
HD3850 : 18,6

F.E.A.R 1600x1200 8xAA/16xAF:
HD3870 : 40
8800GT : 40
HD3850 : 33


Gothic 3 1600x1200 16xAF:
8800GT : 37,6
HD3870 : 20,8
HD3850 : 18,0

Oblivion 1600x1200 8xAA/16xAF:
HD3870 : 39,7
8800GT : 35,7
HD3850 : 34,3

Prey 1600x1200 8xAA/16xAF:
HD3870 : 43,8
8800GT : 36,4
HD3850 : 36,9

Rainbow Six Vegas ? 1600x1200
HD3870 : 46,3
8800GT : 40,2
HD3850 : 39,7

Stalker 1600x1200 1xAA/16xAF:
HD3870 : 31,5
8800GT : 19,6
HD3850 : 14,0

DX10 benchmark

Bioshock 1600x1200 1xAA/16xAF:
HD3870 : 37,0
8800GT : 42,8
HD3850 : 31,8

Call of Juarez 1600x1200 4xAA/16xAF:
HD3870 : 16,8
8800GT : 10,3
HD3850 : 6,3

Company of Heroes 1600x1200 8xAA/16xAF:

HD3870 : 20,1
8800GT : 16,3
HD3850 : 14,6

Crysis 1600x1200 4xAA/16xAF
HD3870 : 10,6
8800GT : 7,6
HD3850 : 4,3

Lost Planet 1600x1200 8xAA/16xAF Snow:
HD3870 : 18,8
8800GT : 16,3
HD3850 : 7,8

World in conflict 1600x1200 4xAA/16xAF:
HD3870 : 28,4
8800GT : 37,5
HD3850 : 23,9


http://www.computerbase.de/art...adeon_hd_3850_rv670/2/

When you turn on HQ and AA , 8800GT ain't so destroyer driverheaven make it out to be.

The 3870 is faster in every 8xAA test but one. And faster overall. Even the 3850 is right up there.

For the people claiming there is no need for AA above 1600x1200, or for 8xAA, you dont have a clue what other people like, or want.

 

[imported_Scoop]

Originally posted by: Azn
Maybe if you had a CRT but LCD is more noticeable if you don't add some kind of AA even high resolutions.

I have an LCD, actually what I meant to say in my post was that if you're actually gaming and not looking at the scenery 90% of the time, you don't see a difference. AA is the last, ok second to last after VSync, option that I turn on in every game.

 

[imported_Scoop]

Originally posted by: Ackmed
From another thread, nicely typed up, so Ill just paste it.

Originally posted by: tuteja1986
DX 9 benchmark

Anno 1701 1600x1200 4xAA/16xAF
HD3870 : 35,7
8800GT : 31,3
HD3850 : 30,6

Clive Barker's Jericho 1600x1200 8xAA/16xAF:
HD3870 : 15,6
8800GT : 5,6
HD3850 : 1,3

Colin McRae Dirt 1600x1200 4xAA/16xAF
HD3870 : 15,8
8800GT : 4,5
HD3850 : 4,3

Company of heroes 1600x1200 8xAA/16xAF
8800GT : 64,2
HD3870 : 58,7
HD3850 : 18,6

F.E.A.R 1600x1200 8xAA/16xAF:
HD3870 : 40
8800GT : 40
HD3850 : 33


Gothic 3 1600x1200 16xAF:
8800GT : 37,6
HD3870 : 20,8
HD3850 : 18,0

Oblivion 1600x1200 8xAA/16xAF:
HD3870 : 39,7
8800GT : 35,7
HD3850 : 34,3

Prey 1600x1200 8xAA/16xAF:
HD3870 : 43,8
8800GT : 36,4
HD3850 : 36,9

Rainbow Six Vegas ? 1600x1200
HD3870 : 46,3
8800GT : 40,2
HD3850 : 39,7

Stalker 1600x1200 1xAA/16xAF:
HD3870 : 31,5
8800GT : 19,6
HD3850 : 14,0

DX10 benchmark

Bioshock 1600x1200 1xAA/16xAF:
HD3870 : 37,0
8800GT : 42,8
HD3850 : 31,8

Call of Juarez 1600x1200 4xAA/16xAF:
HD3870 : 16,8
8800GT : 10,3
HD3850 : 6,3

Company of Heroes 1600x1200 8xAA/16xAF:

HD3870 : 20,1
8800GT : 16,3
HD3850 : 14,6

Crysis 1600x1200 4xAA/16xAF
HD3870 : 10,6
8800GT : 7,6
HD3850 : 4,3

Lost Planet 1600x1200 8xAA/16xAF Snow:
HD3870 : 18,8
8800GT : 16,3
HD3850 : 7,8

World in conflict 1600x1200 4xAA/16xAF:
HD3870 : 28,4
8800GT : 37,5
HD3850 : 23,9


http://www.computerbase.de/art...adeon_hd_3850_rv670/2/

When you turn on HQ and AA , 8800GT ain't so destroyer driverheaven make it out to be.

The 3870 is faster in every 8xAA test but one. And faster overall. Even the 3850 is right up there.

For the people claiming there is no need for AA above 1600x1200, or for 8xAA, you dont have a clue what other people like, or want.

WOW, those are some great frame rates to play your games with huh?

 

[Ackmed]

Originally posted by: Scoop

WOW, those are some great frame rates to play your games with huh?

No, most of them wouldnt be playable to me with 8xAA. However, those are all new games. You do realize people play games that are over a year old, right? With "older" games, 8xAA is very playable, and looks stellar. I still play BF2 because I enjoy it so much. The game engine for it, benefits greatly from AA to me, and the more the better.

The fact is, the 8800 cards take a nosedive when AA is cranked up. Which probably doesnt matter to most people, as its harder to get playable frames at 8xAA. But as I said, people still play older games, that you can get playable frames. Its too bad reviews only show new games, as if thats all we play.

Once again, you dont know what other people see, and what they dont. You can notice jaggies easily when playing, and not "looking at the scenery". Some games show jaggies more than others. Dont act as if your opinion, is the only one that matters.

 

[dreddfunk]

It is always good to add new data to any analysis, but the data here hardly back up such an inflammatory title to the original post. Clearly, the 3870 has a little better legs at high AA according to this one article. Still, the difference is hardly huge. At some point, every video card 'tanks'. The 3870 may be a better option for gamers with certain specific requirements. While this is good to know, it doesn't justify this statement:

Sadly whether intentional or not, whether Nvidia bought all these wins or they just got them out of the pure ignorance of hardware reviewers it is happening.

Implying that NVIDIA possibly 'bought' reviewers off doesn't make sense to me at all. If the reviews on the 8800GT somehow excluded very, very common settings and resolutions, your point might have some validity. The omission of 8xAA in testing just isn't very earth-shattering. In fact, very few reviews run tests at higher AA settings for a number of reasons. First, both NVIDIA and ATI implement higher AA modes differently, which makes testing--and its focus on framerates--problematic. Second, while performance in older games is important and interesting, it simply takes a back seat to testing on current games, games that usually are demanding enough to render testing at AA levels above 4x virtually irrelevant (the games become unplayable).

@Ackmed - yes,I agree without reservation that nobody should try and tell another person what they should need or want. Yet I hope you would agree that 8xAA is hardly a mainstream setting, which is the underlying point that others are trying to make.

 

[hooflung]

8xAA is hardly a mainstream setting because cards have traditionally tanked. The 3970 could be the first card on the market that fills this gap and also show what the HD 3k hardware can actually is capable of. I'd like to know if I buy a video card that it can run tomorrow's games at acceptable rates and TODAYS games and YESTERDAYS games faster than anything on the market at the highest of settings.

 

[Munky]

8xAA is not a blur, it uses the same method as 4xAA only with more samples. However, I find that the 4xAA setting is good enough to get rid of the jaggies, I'm not too concerned about how a card performs with 8xAA.

 

[SniperDaws]

Originally posted by: Scoop

Originally posted by: Ackmed
From another thread, nicely typed up, so Ill just paste it.

Originally posted by: tuteja1986
DX 9 benchmark

Anno 1701 1600x1200 4xAA/16xAF
HD3870 : 35,7
8800GT : 31,3
HD3850 : 30,6

Clive Barker's Jericho 1600x1200 8xAA/16xAF:
HD3870 : 15,6
8800GT : 5,6
HD3850 : 1,3

Colin McRae Dirt 1600x1200 4xAA/16xAF
HD3870 : 15,8
8800GT : 4,5
HD3850 : 4,3

Company of heroes 1600x1200 8xAA/16xAF
8800GT : 64,2
HD3870 : 58,7
HD3850 : 18,6

F.E.A.R 1600x1200 8xAA/16xAF:
HD3870 : 40
8800GT : 40
HD3850 : 33


Gothic 3 1600x1200 16xAF:
8800GT : 37,6
HD3870 : 20,8
HD3850 : 18,0

Oblivion 1600x1200 8xAA/16xAF:
HD3870 : 39,7
8800GT : 35,7
HD3850 : 34,3

Prey 1600x1200 8xAA/16xAF:
HD3870 : 43,8
8800GT : 36,4
HD3850 : 36,9

Rainbow Six Vegas ? 1600x1200
HD3870 : 46,3
8800GT : 40,2
HD3850 : 39,7

Stalker 1600x1200 1xAA/16xAF:
HD3870 : 31,5
8800GT : 19,6
HD3850 : 14,0

DX10 benchmark

Bioshock 1600x1200 1xAA/16xAF:
HD3870 : 37,0
8800GT : 42,8
HD3850 : 31,8

Call of Juarez 1600x1200 4xAA/16xAF:
HD3870 : 16,8
8800GT : 10,3
HD3850 : 6,3

Company of Heroes 1600x1200 8xAA/16xAF:

HD3870 : 20,1
8800GT : 16,3
HD3850 : 14,6

Crysis 1600x1200 4xAA/16xAF
HD3870 : 10,6
8800GT : 7,6
HD3850 : 4,3

Lost Planet 1600x1200 8xAA/16xAF Snow:
HD3870 : 18,8
8800GT : 16,3
HD3850 : 7,8

World in conflict 1600x1200 4xAA/16xAF:
HD3870 : 28,4
8800GT : 37,5
HD3850 : 23,9


http://www.computerbase.de/art...adeon_hd_3850_rv670/2/

When you turn on HQ and AA , 8800GT ain't so destroyer driverheaven make it out to be.

The 3870 is faster in every 8xAA test but one. And faster overall. Even the 3850 is right up there.

For the people claiming there is no need for AA above 1600x1200, or for 8xAA, you dont have a clue what other people like, or want.

WOW, those are some great frame rates to play your games with huh?

The 8800GT in those benchies is the 256mb version NOT the 512mb and also the website that did the test were bias to ATI so those benchies are what i call BOLLOCKS.

http://forums.anandtech.com/me...=2127438&enterthread=y
Here is the original thread

 

[imported_Scoop]

Dear Ackmed,

Ok, my bad, I'm just used to thinking that what I say is my opinion, not seeing it necessary to stress this separately.

 

[aldamon]

Originally posted by: munky
However, I find that the 4xAA setting is good enough to get rid of the jaggies, I'm not too concerned about how a card performs with 8xAA.

I agree.

 

[praesto]

I've never applied 8xAA to my games. I can barely notice a difference in jaggies, going from 4x to 8xAA. Therefore, these numbers do not matter to me. Just as Ackmed pointed out, one guys preference doesn't necessairly apply to other people, and I agree completely with him.

According to the results at computerbase, the 8800gt exceeds vs the 3870 with 4xAA enabled. So for me, the 8800GT must be the best purchase.

This review: http://firingsquad.com/hardwar...0_performance_preview/

also tells me that the 8800gt is far superior. Except in bioshock actually. Firingsquad, however, did not use the forced AA methods. Probably because the 3870 fares like the 2900xt did with zero optimizations when AA was forced in bioshock.

The very same conclusion can be drawn from this review: http://www.techspot.com/review/76-asus-radeon-hd-3870/

and this one: http://vr-zone.com/articles/AMD_Radeon_HD3850_and_HD3870_-_AMD's_salvation?/5392-14.html

aaaand this one: http://www.hardwarecanucks.com...rformance-preview.html

I can't bother digging up more. Some of reviews even show worse numbers for the 38700 than the ones I've chosen, but I have a feeling that some people would get very upset if I included those.

 

[Ackmed]

Originally posted by: SniperDaws


The 8800GT in those benchies is the 256mb version NOT the 512mb and also the website that did the test were bias to ATI so those benchies are what i call BOLLOCKS.

http://forums.anandtech.com/me...=2127438&enterthread=y
Here is the original thread

I know where the thread is.. I got the numbers from there.

Tests bias towards ATi? I didnt know a dozen or so popular games were all bias towards ATi. Thanks for clearing that up...

praesto, yes we all know the 8800gt is the faster overall card. This isnt new news. Both cards have pros/cons, but the performance "crown" goes to the gt. If all you care about is speed, its the obvious choice. And most people dont use 8xAA, because its unplayable, they cant tell the difference, or any other reason they have. This thread is about its 8xAA performance, and that it tanks when applied, for whatever reason. Its pretty easy to just acknowledge this fact, and move on. No need to storm in and try to defend anything.

 

[JACKDRUID]

Originally posted by: Ackmed

Originally posted by: SniperDaws


The 8800GT in those benchies is the 256mb version NOT the 512mb and also the website that did the test were bias to ATI so those benchies are what i call BOLLOCKS.

http://forums.anandtech.com/me...=2127438&enterthread=y
Here is the original thread

I know where the thread is.. I got the numbers from there.

Tests bias towards ATi? I didnt know a dozen or so popular games were all bias towards ATi. Thanks for clearing that up...

... la la la

you missed the point .. let me highlight for you.

 

[ginfest]

Originally posted by: JACKDRUID

Originally posted by: Ackmed

Originally posted by: SniperDaws


The 8800GT in those benchies is the 256mb version NOT the 512mb and also the website that did the test were bias to ATI so those benchies are what i call BOLLOCKS.

http://forums.anandtech.com/me...=2127438&enterthread=y
Here is the original thread

I know where the thread is.. I got the numbers from there.

Tests bias towards ATi? I didnt know a dozen or so popular games were all bias towards ATi. Thanks for clearing that up...

... la la la

you missed the point .. let me highlight for you.

Note to self: if you are going to "pimp" a certain brand, use fair and accurate testing methods

 

[jeffw2767602]

AMD and ATI fanboys are really having a rough year, aren't they?

 

[Distant]

First off to the guy who said they are using the 8800GT 256mb card get real the 256 version wasn't even out when this review came out, if you look at the link here you can clearly see they are testing with the 8800GT 512MB not the 256.


And I find it hilarious how some of you are saying 8xAA just doesn't matter, what a joke! There is a difference between 4 and 8 and both nvidia and ATI can be tested at 8x in the same way so can you really tell me that what cards can do on a 30 inch monitor is more important then 8x AA tests on lower resolutions?

Many people play older games still, hell probably most of us are still playing one of valves games on steam.

If you want to talk about fanboys I think blatantly stating 8xAA doesn't matter, because it makes your card maker (nvidia) look worse then all the current glowing reviews, yes all cards will fall at some point, but if you really don't think that anyones going to care that their new 8800GT they bought is going to tank when they try to run a game at 8xAA your the one who needs to take the goggles off.

I bet you think high AA tests are irrelevant for sli/crossfire comparisons as well?

 

[Quiksilver]

Originally posted by: Ackmed
From another thread, nicely typed up, so Ill just paste it.

Originally posted by: tuteja1986

---Snipped wall of text---

http://www.computerbase.de/art...adeon_hd_3850_rv670/2/

When you turn on HQ and AA , 8800GT ain't so destroyer driverheaven make it out to be.

The 3870 is faster in every 8xAA test but one. And faster overall. Even the 3850 is right up there.

For the people claiming there is no need for AA above 1600x1200, or for 8xAA, you dont have a clue what other people like, or want.

I wouldn't exactly consider those playable frame rates on either card at that point anyway. So you would have to lower AA anyway(or use 16X CSAA), at which point the 8800GT remains on top.

 

[Cookie Monster]

You know this is where CSAA comes in.

Why use your traditional 8xMSAA that kills badnwidth/framebuffer and have a very large impact on performance, when infact you are better off using 16xCSAA that has a similiar performance hit that of a 4xAA while providing a better overall IQ.
Theres a reason CSAA is in nV hardware. Its because of situations like these where your card lacks bandwidth and framebuffer.


edit - An investigation would be nice. If they are actually applying 8xAA, if its a driver problem, or other factors influencing such performance nose dive.

 

[JACKDRUID]

Originally posted by: Distant

If you want to talk about fanboys I think blatantly stating 8xAA doesn't matter, because it makes your card maker (nvidia) look worse then all the current glowing reviews

why should we trust this unknown site when it's number deviate so much from all other reviews done by much better/well-know/trusted sites?