Originally posted by: josh6079
Because Anandtech is inept at providing realistic benches when it comes to Oblivion. When the R5x series emerged with the ability to do HDR+AA, that feature went unsung on Anandtech's benches for the sole reason that nVidia couldn't do it as well. Now that nVidia has finally been able to do it since November, they still have yet to bench Oblivion the way most enthusiasts are playing it.I wonder why Derek hasn't tested Oblivion with AA enabled.
I made a previous thread concerning this issue. Anandtech just can't be bothered. They need to provide some better GPU benchmarks, such as the ones found at Hard, Rage3D, Xbit, etc.
Originally posted by: munisgtm
Originally posted by: josh6079
Because Anandtech is inept at providing realistic benches when it comes to Oblivion. When the R5x series emerged with the ability to do HDR+AA, that feature went unsung on Anandtech's benches for the sole reason that nVidia couldn't do it as well. Now that nVidia has finally been able to do it since November, they still have yet to bench Oblivion the way most enthusiasts are playing it.I wonder why Derek hasn't tested Oblivion with AA enabled.
I made a previous thread concerning this issue. Anandtech just can't be bothered. They need to provide some better GPU benchmarks, such as the ones found at Hard, Rage3D, Xbit, etc.
agreed . I think AT should do GPU reviews like Hard's or Bit-tech's
Originally posted by: thilan29
Originally posted by: Remedy
I would like to see how the GTS performs in SLI though.
It would still only have a 320mb frame buffer correct? So then you would still be bandwidth limited?
Originally posted by: chizow
Originally posted by: munisgtm
Originally posted by: josh6079
Because Anandtech is inept at providing realistic benches when it comes to Oblivion. When the R5x series emerged with the ability to do HDR+AA, that feature went unsung on Anandtech's benches for the sole reason that nVidia couldn't do it as well. Now that nVidia has finally been able to do it since November, they still have yet to bench Oblivion the way most enthusiasts are playing it.I wonder why Derek hasn't tested Oblivion with AA enabled.
I made a previous thread concerning this issue. Anandtech just can't be bothered. They need to provide some better GPU benchmarks, such as the ones found at Hard, Rage3D, Xbit, etc.
agreed . I think AT should do GPU reviews like Hard's or Bit-tech's
Well, I think AT does a good job keeping their reviews consistent from one to another, but ya sometimes that means sacrificing detail. They're typically a high-level overview and don't get into as many specific tests as some of the other sites. Also, AT gives outstanding technical background and market insight on new GPUs which I haven't seen on most other sites.
But to emphasize the point, this Polish site review uses RivaTuner in a way I've never seen before in a benchmark. Maybe I just haven't read enough on other sites, but it shows local vs. non-local video memory usage at each resolution/AA setting. It clearly illustrates why the 320MB GTS is getting slaughtered at higher resolutions. As you increase the resolution and enable AA, local memory is getting pegged @300MB and the rest overflows slamming your system/page file resulting in massive performance hits. Its a shame nVidia didn't see this coming, as the damage has been done and I doubt many if any sites will revisit their tests even if a driver optimization fixes this.
Originally posted by: munisgtm
I agree that AT gives great technical background and stuff but if adds additional things like Hardocp , that would be one hell of a review .
Originally posted by: munisgtm
Originally posted by: chizow
Originally posted by: munisgtm
Originally posted by: josh6079
Because Anandtech is inept at providing realistic benches when it comes to Oblivion. When the R5x series emerged with the ability to do HDR+AA, that feature went unsung on Anandtech's benches for the sole reason that nVidia couldn't do it as well. Now that nVidia has finally been able to do it since November, they still have yet to bench Oblivion the way most enthusiasts are playing it.I wonder why Derek hasn't tested Oblivion with AA enabled.
I made a previous thread concerning this issue. Anandtech just can't be bothered. They need to provide some better GPU benchmarks, such as the ones found at Hard, Rage3D, Xbit, etc.
agreed . I think AT should do GPU reviews like Hard's or Bit-tech's
Well, I think AT does a good job keeping their reviews consistent from one to another, but ya sometimes that means sacrificing detail. They're typically a high-level overview and don't get into as many specific tests as some of the other sites. Also, AT gives outstanding technical background and market insight on new GPUs which I haven't seen on most other sites.
But to emphasize the point, this Polish site review uses RivaTuner in a way I've never seen before in a benchmark. Maybe I just haven't read enough on other sites, but it shows local vs. non-local video memory usage at each resolution/AA setting. It clearly illustrates why the 320MB GTS is getting slaughtered at higher resolutions. As you increase the resolution and enable AA, local memory is getting pegged @300MB and the rest overflows slamming your system/page file resulting in massive performance hits. Its a shame nVidia didn't see this coming, as the damage has been done and I doubt many if any sites will revisit their tests even if a driver optimization fixes this.
I agree that AT gives great technical background and stuff but if adds additional things like Hardocp , that would be one hell of a review .
Originally posted by: chizowBut to emphasize the point, this Polish site review uses RivaTuner in a way I've never seen before in a benchmark. Maybe I just haven't read enough on other sites, but it shows local vs. non-local video memory usage at each resolution/AA setting. It clearly illustrates why the 320MB GTS is getting slaughtered at higher resolutions. As you increase the resolution and enable AA, local memory is getting pegged @300MB and the rest overflows slamming your system/page file resulting in massive performance hits. Its a shame nVidia didn't see this coming, as the damage has been done and I doubt many if any sites will revisit their tests even if a driver optimization fixes this.
Originally posted by: CaiNaM
Originally posted by: chizowBut to emphasize the point, this Polish site review uses RivaTuner in a way I've never seen before in a benchmark. Maybe I just haven't read enough on other sites, but it shows local vs. non-local video memory usage at each resolution/AA setting. It clearly illustrates why the 320MB GTS is getting slaughtered at higher resolutions. As you increase the resolution and enable AA, local memory is getting pegged @300MB and the rest overflows slamming your system/page file resulting in massive performance hits. Its a shame nVidia didn't see this coming, as the damage has been done and I doubt many if any sites will revisit their tests even if a driver optimization fixes this.
i'm not sure. if you look at the call of juarez benchmarks, it shows the same thing happening (the card's non-local use of memory increasing), yet the 320mb version still continues to lead, and hold the same 10% fps advantage.
this would seem to indicate the explanation is not so simple...
Originally posted by: chizow
Not sure what you're referring to, but I see a big fat donut in place of #s for Call of Juarez 1920 w/ 4x AA for the 320 GTS.
This is a flawed comparison because the difference between a 640 MB version and a 320 MB version is the ability to run 4xAA at the same resolution.I'd rather run 1920x1200 with no AA and no AF and get 60+ frames instead of 1280x1024 with 10,055X AA and 45,485X AF with < 30fps.
Not only is this false but if it's borderline unplayable on 640 MB varient then it's certainly unplayable on a 320 MB version. The perfomance difference between the two is quite large at 4xAA and would likely be even larger at higher AA levels.Fact of the matter is that anything over 1920 w/ AA is borderline unplayable on anything short of an 8800GTX in most current games
Again this is totally false - I run plenty of new games at 1920x1440 with 16xAF and 4xAA on my 8800 GTS (640 MB). If I had a 320 MB GTS I'd have to run 2xAA or even no AA and this would have a huge impact on IQ.Simply put, gaming at high resolutions and AA/AF cranked up simply don't mix well with newer games
Its not a flawed comparison because I'm not comparing my preferences to any cards. I'm just pointing out there is less need to run ANY AA at higher resolutions, so how a card performs at high resolutions + AA isn't as significant as how it performs at higher resolutions w/out AA.Originally posted by: BFG10K
This is a flawed comparison because the difference between a 640 MB version and a 320 MB version is the ability to run 4xAA at the same resolution.I'd rather run 1920x1200 with no AA and no AF and get 60+ frames instead of 1280x1024 with 10,055X AA and 45,485X AF with < 30fps.
There's very little difference until 16x12/10 in a few games and 19x14/12 on everything else, and yes, 2560 with 4x AA in today's games are borderline unplayable. Unless you've changed your tone from 3 years ago and consider < 30 fps playable? So you can either split hairs between 20 fps and 13 fps at 2560 with 4x AA or you can acknowledge they both run like pigs, turn off AA and then split hairs between 45 and 60 fps. Just curious what newer games you're running at 1920+ w/ 4x AA? Its certainly not Oblivion or R6: Vegas. The next generation of games and their offspring (UT3, Crysis) are all going to stress the GPU similarly, if not moreso.Not only is this false but if it's borderline unplayable on 640 MB varient then it's certainly unplayable on a 320 MB version. The perfomance difference between the two is quite large at 4xAA and would likely be even larger at higher AA levels.Fact of the matter is that anything over 1920 w/ AA is borderline unplayable on anything short of an 8800GTX in most current games
Again this is totally false - I run plenty of new games at 1920x1440 with 16xAF and 4xAA on my 8800 GTS (640 MB). If I had a 320 MB GTS I'd have to run 2xAA or even no AA and this would have a huge impact on IQ.[/quote]Simply put, gaming at high resolutions and AA/AF cranked up simply don't mix well with newer games
All the more reason why I'd want a card that was already handling the bare minimum AA for high resolutions. (aka - The 640 MB GTS)The next generation of games and their offspring (UT3, Crysis) are all going to stress the GPU similarly, if not moreso.
Again, if it is only going to get worse, wouldn't you want the card with the greater IQ cushion?And yes, the 640MB 8800GTS runs 1900 acceptably in today's games, but even the more taxing games give it problems with AA turned on. Its only going to get worst, not better, so don't get used to those AA settings for too long.
If that resolution equals the real estate of the side of a house, I'd be complaining too. It's not just a matter of resolution, but screen size that determines how bad the aliasing actually is.I'm sure you'll be complaining about jaggies even when you're running 12 billion x 10 billion resolution.....
All the more reason to buy a card based on how it performs w/out AA at high resolutions instead of worrying about how it runs compared to a more expensive card that barely runs AA at high resolutions today. Neither card is running AA in tomorrow's games....which leaves them both with AA off, at which point the 320 performs similarly to the 640. Again, AA is a luxury you can afford when running current and older games...expecting AA at the highest resolutions for future games or even worst, making purchasing decisions with the expectation they'll run the highest resolutions w/ AA for future games is foolish.Originally posted by: josh6079
All the more reason why I'd want a card that was already handling the bare minimum AA for high resolutions. (aka - The 640 MB GTS)
Not when they both run about the same at settings where frame rate is still acceptable. You can have your 25 fps 4xAA at 1920 in Crysis, I'll be satisfied running at 60 fps and no AA....same as the guy with the 320MB getting 58 fps.Again, if it is only going to get worse, wouldn't you want the card with the greater IQ cushion?
Of course, but I'm sure a 12 billion x 10 billion would fill the side of a mansion without any problems considering a 30" widescreen running 2560 is only 4 million pixels and looks better than a 19x12 with 16x AA.that resolution equals the real estate of the side of a house, I'd be complaining too. It's not just a matter of resolution, but screen size that determines how bad the aliasing actually is.
What? You've seen benchmarks of Crysis, Quake Wars, Alan Wake, etc. comparing an 8800GTS 640 MB and 320? Are you sure you're not seeing the 420 edition?Neither card is running AA in tomorrow's games....
They don't run about the same, that's the point. If they did nVidia would be shooting themselves in the foot and might as well discontinue the 640 MB GTS.Not when they both run about the same at settings where frame rate is still acceptable.
While we're grabbing numbers out of our arses, can you give me the frame rate for the both GTS versions for Alan Wake? Quake Wars?You can have your 25 fps 4xAA at 1920 in Crysis, I'll be satisfied running at 60 fps and no AA....same as the guy with the 320MB getting 58 fps.
...a 30" widescreen running 2560 is only 4 million pixels and looks better than a 19x12 with 16x AA.
I wouldn't need to see any benchmarks to know newer games built on game engines that bring the 640MB GTS to its knees with AA are only going to make AA unplayable. Queue the R6: Vegas benchmarks......Originally posted by: josh6079
What? You've seen benchmarks of Crysis, Quake Wars, Alan Wake, etc. comparing an 8800GTS 640 MB and 320? Are you sure you're not seeing the 420 edition?
Um, they certainly do. In most every review I've seen (the exception being Q4 Ultra mode), the 320MB and 640MB run within a few frames of each other w/out AA up to 1920. Again, you can sacrifice FPS for IQ, but I'll notice a massive drop in frame rates in a 20 FPS slideshow far before I notice a jaggy on a rope or leaf off in the distance.They don't run about the same, that's the point. If they did nVidia would be shooting themselves in the foot and might as well discontinue the 640 MB GTS.
That's only a valid comparison if the 640MB is running AA as fast as the 320MB is without, so unless the 640MB is getting a 0-5% performance hit with AA enabled, your comparison is flawed.Only in some benchmark that doesn't currently exist do they "run about the same", or in a game without enabling any kind of AA at which point the image quality is inferior.
Once again.....queue up the R6: Vegas benchmarks. UT3 Engine....not sure what else you're looking for to gauge future performance. Maybe if someone made a game today based on an innovative next generation game engine that'd give us the results we're looking for. Oh wait....let me pull some more numbers out of my arse :roll: But ya, you can have your 25 fps with 4xAA @ 1920 in Crysis, I'll be content with my inferior image quality running 60 fps with no AA....just like the guy with the 320MB GTS.While we're grabbing numbers out of our arses, can you give me the frame rate for the both GTS versions for Alan Wake? Quake Wars?You can have your 25 fps 4xAA at 1920 in Crysis, I'll be satisfied running at 60 fps and no AA....same as the guy with the 320MB getting 58 fps.
Yep, its 2x as many pixels in an area that's only what? @30% larger than a 24". Better pixel density and lower pitch eliminate the need for AA and the image looks sharper as well. But ya its something you need to see I guess. I have and as great as it looks, I know that running that resolution simply isn't realistic unless you plan to live on the bleeding edge 24/7/365 while forking out $600-1000 every 6 months.Let me get this straight. You're claiming that a 30" panel running at a native 2560x1600 resolution without AA looks better than a 1920x1200 panel with 16xAA?
That's the basis for your whole argument? A crappy port that is only one game out of many and doesn't even have working AA?Queue the R6: Vegas benchmarks......
There's simply no use to discussing image quality differences with you then.Yep, its 2x as many pixels in an area that's only what? @30% larger than a 24". Better pixel density and lower pitch eliminate the need for AA and the image looks sharper as well.
Yeah, they do run similarly, without AA. When you look at more than just one crappy ported game, you find several major titles that allow the 640 MB to have playable frames and noticeably better IQ.Um, they certainly do. In most every review I've seen (the exception being Q4 Ultra mode), the 320MB and 640MB run within a few frames of each other w/out AA up to 1920.
I also would like to know where you're getting your information that Crysis will run at 25 fps with 4xAA on a 640 MB GTS, since the UT3 engine isn't what Crysis will be using.UT3 Engine....not sure what else you're looking for to gauge future performance.
That's only a valid comparison if the 640MB is running AA as fast as the 320MB is without
We know. No one is saying that you can't be content without using AA for your games. What I take issue with is when you're saying AA in general isn't as beneficial as a high resolution and that Crysis is only going to get 25 fps with 4x AA and a 640 MB GTS just because R6: Vegas is struggling - a completely different game ported from the 360 and using a completely different engine.But ya, you can have your 25 fps with 4xAA @ 1920 in Crysis, I'll be content with my inferior image quality running 60 fps with no AA....just like the guy with the 320MB GTS.
Of course it's flawed, we're talking about a comparison at the same resolution where one card can run 4xAA but the other can't due to performance reasons.Its not a flawed comparison
I disagree. If anything I would argue that AA is even more important at high resolutions because aliasing stands out even more when it's being compared to an otherwise sharp picture.I'm just pointing out there is less need to run ANY AA at higher resolutions, so how a card performs at high resolutions + AA isn't as significant as how it performs at higher resolutions w/out AA.
The differences look pretty big to me.There's very little difference until 16x12/10 in a few games and 19x14/12 on everything else,
Again with your resolution fixation. Nobody was talking about 2560x1600 per-se, we're talking about resolutions where 4xAA is possible on the 640 but not on the 320. These might be 2560x1600 but they might not be.2560 with 4x AA in today's games are borderline unplayable.
The difference in many cases is the difference between playable and not playable, like in AT's Fear score for example: 51 FPS vs 37 FPS.So you can either split hairs between 20 fps and 13 fps at 2560 with 4x AA or you can acknowledge they both run like pigs, turn off AA and then split hairs between 45 and 60 fps.
Fear, Doom 3, Quake 4, Condemned, Call of Juarez (SM 2.0), Serious Sam 2 and Far Cry to name a few.Just curious what newer games you're running at 1920+ w/ 4x AA?
Right, and when that happens what card do you suppose will run out of steam first, the 320 or the 640?The next generation of games and their offspring (UT3, Crysis) are all going to stress the GPU similarly, if not moreso.
That's when you have the option to turn off AA. You don't have the option on the 320 MB if you're already running AA disabled in current games so what are you going to do then?And yes, the 640MB 8800GTS runs 1900 acceptably in today's games, but even the more taxing games give it problems with AA turned on.
As long as you have at least a minimum of 4xAA then dropping the resolution has far less impact on image quality than it does when AA is disabled. Without AA jagged edges explode exponentially when the resolution is dropped.Its only going to get worst, not better, so don't get used to those AA settings for too long.
At 1920x1440 the difference between 0xAA, 2xAA and 4xAA is as plain as day.As for huge impact lol....I'm sure you'll be complaining about jaggies even when you're running 12 billion x 10 billion resolution.....
Originally posted by: keysplayr2003
A good point to bring up would be what is the max resolution a given person can play at? What are his/her monitors limits? I am in this category. I'm looking to buy a 22" widescreen who's native res is 1680x1050. The GTS 320 seems to do pretty well at 1600x1200 or lower with AA enabled, so I have to ask myself, is a GTS 640 needed? Or is a GTS 320 enough for my needs. Factor in price differences as well. Tough decision so far for me. Not really sure which way to go.
Originally posted by: Skyguy
So, just to switch gears.........
Are we all in agreement that at 1280x1024 the 320mb card will more than suffice, even with AA on?