GTX 580 v HD 5970 2GB performance face-off

[qubit]

@EastCoasthandle & mdsx1950: I hate you!

Those bloody graphs of there's are confusing to look at, so I hadn't looked at them very closely before. But I did now!

So, generally speaking, it does appear that the lowest minimums on the 580 & 5970 are about the same, with the 5970 otherwise having higher performance. Therefore, it's all win for the 5970 here.

Where the 5970 falls down, is the latter part of the F1 2010 test though. The 580 is generally better there, without such huge drops. I reckon that loss is more likely due to having two GPUs working together though, rather than it be a brand issue. Remember, one 580 GPU is more powerful than one 5870 GPU and we have two 5870s in tandem here, so it's not really an apples to apples comparison. And dual GPU setups always have issues.

Would a GTX 295 have done any better under those conditions? I dunno. I say a 295, because it has weaker GPUs that are closer in performance to the 5870, making it a better comparison to the 5970.

Regarding the smoother gameplay comments in the performance summary, it's certainly possible that the 580 still played smoother in overall feel and the benches don't really show this. I've not tested this myself so I can't say. I can only relate their comments to my own experience between the 4870 and 8800 GTX.

EDIT: yes, a 5970 with 2GB per GPU would be very interesting. Bring it on!

 

[cadaveca]

EDIT: yes, a 5970 with 2GB per GPU would be very interesting. Bring it on!

They are already avaialble for sale, although @ like $1200.

The thing that is missed here, is that it takes a certain performance level before 1.5GB+ of vram actually shows it's strength, and those apps were chosen specifically to highlight that. This review doesn't cater to anyone, except high-end users, and as ECH mentions, merely highlights where 2GB of vram is needed. We are very close to the point that yes, a gpu needs 2GB.

I'll wait for 6970 before making final judgements, though.

 

[EastCoasthandle]

Well you won't need a 5970 4Gig version. All a reviewer would need to do is lower the resolution to 1920 IMO.

 

[cadaveca]

Yes, but I have a 2560x1600 monitor. 1920x1080/1200 numbers are of no use to me, unless they are using 3 of them.

1920x1080 is far from the high-end, now. There's no need for it in a high-end review. 5760x1080 is the new high-end.

 

[erocker]

At max DX11 settings, 4xMSAA, Very High textures are cra***er then High textures.

How's that for a simple

No idea. I guess we need to know what consoles use for graphics settings in relation to the PC.

 

[EastCoasthandle]

Yes, but I have a 2560x1600 monitor. 1920x1080/1200 numbers are of no use to me, unless they are using 3 of them.

1920x1080 is far from the high-end, now. There's no need for it in a high-end review.

So what? 1920 is a valid resolution for enthusiast. That resolution wasn't suggested just for you . It was a suggestion to implicate that we need more then just 1 resolution when looking at review results. Therefore, 1920 added to the 2560 results they already have would IMO bring a very clear picture as to how one should interpret each card. Multi monitor card reviews should always remain separate from single monitor reviews.

 

[cadaveca]

So what? 1920 is a valid resolution for enthusiast. That resolution wasn't suggested just for you . It was a suggestion to implicate that we need more then just 1 resolution when looking at review results. Therefore, 1920 added to the 2560 results they already have would IMO bring a very clear picture as to how one should interpret each card. Multi monitor card reviews should always remain separate from single monitor reviews.

LoL. As if I am pompous enough to think those reviews cater to me, and me alone. get off your high horse there, bub. Fact of the matter is, in this thread alone, I am not the only 2560x1600 user.

1920x1080 doesn't reveal where the true performance deficit is. the review isn't there to cater to everyone..it's there to highlight a very obvious fault, and in a way it cannot be disputed.

Fact of the matter is, high-end cards are overkill for 1920x1080. Mid-range cards are more than capable to meet the needs of this res, so I fail to see how 1920x1080 has ANY revelance to high-end cards...you show cards as they are intended to be used, not as they are being used, in mis-matched, poorly designed, self-built boxes.

 

[Spaceman Spiff]

I don't think a res drop is even necessary. Just go down to 4xaa and these 1g memory limits may disappear. But even if you want 8xaa these cards are only getting 40fps like I said before. So is it wise for a highend user to jump on a 580 now or wait for 69xx results and price wars? Is 40fps acceptable to a highend user?

 

[cadaveca]

Is 40fps acceptable to a highend user?

Most likely, not. High-end users don't spend exponential amounts of money for miniscule gains, just to get 40FPS. The review should be taken for what it's worth...showing where high-end cards need more vram, and where they don't.

 

[nesco1801]

Yes, but I have a 2560x1600 monitor. 1920x1080/1200 numbers are of no use to me, unless they are using 3 of them.

And what do u run on 2560x1600 with a single 5870?

Donkey Kong and Tetris?

 

[qubit]

I don't think a res drop is even necessary. Just go down to 4xaa and these 1g memory limits may disappear. But even if you want 8xaa these cards are only getting 40fps like I said before. So is it wise for a highend user to jump on a 580 now or wait for 69xx results and price wars? Is 40fps acceptable to a highend user?

No, definitely not. I'd adjust the game and driver settings so that I get a solid 60fps with vsync lock ie no dropped frames at least 95% of the time. Judders really ruin a game.

 

[cadaveca]

And what do u run on 2560x1600 with a single 5870?

Donkey Kong and Tetris?

Pretty much anything, except a few titles. I actually had 2x5870, and just recently downgraded in anticipation of the 6970/6990. I also have 3x 1920x1080 monitors, and am using a single 5870 on one of those...well, actually, Im using 2 monitors, but not in 3D

 

[EastCoasthandle]

LoL. As if I am pompous enough to think those reviews cater to me, and me alone. get off your high horse there, bub. Fact of the matter is, in this thread alone, I am not the only 2560x1600 user.

1920x1080 doesn't reveal where the true performance deficit is. the review isn't there to cater to everyone..it's there to highlight a very obvious fault, and in a way it cannot be disputed.

Fact of the matter is, high-end cards are overkill for 1920x1080. Mid-range cards are more than capable to meet the needs of this res, so I fail to see how 1920x1080 has ANY revelance to high-end cards...you show cards as they are intended to be used, not as they are being used, in mis-matched, poorly designed, self-built boxes.

Based on your reply to my post I simply took you of your high horse . Per your own admission there should be more then just 1 resolution in a review. So the fact remains that if 1920 resolution was included and not just 2560 and multi monitor resolutions (per your previous post) we would indeed see a much clearer picture to the 3 cards tested.

What has been disputed and repudiated is that the review isn't a performance review but a vram review. Specially when a 5870 with just 2 Gigs of ram is just as good as a 580 which is suppose to be a next gen card in F1 2010 . Although it was found in just 1 game the other game which shows similar resolution and IQ provided a measly 6 FPS improvement is very minuscule for a next gen card. While the IQ is other games didn't use 8xAA and the Civilization review should have been omitted do to the 0-1 FPS mins on all cards.

 

[cadaveca]

What has been disputed and repudiated is that that review isn't a performance review but a vram review. Specially when a 5870 with just 2 Gigs of ram is just as good as a 580 which is suppose to be a next gen card in F1 2010 .

Sure, we both agree on that, for sure..

No, definitely not. I'd adjust the game and driver settings so that I get a solid 60fps with vsync lock ie no dropped frames at least 95% of the time. Judders really ruin a game.

Well, and to me, this review kinda highlights why dropping those details is nessecary. It's not cut and dry, but neither is the performance.

 

[nesco1801]

Pretty much anything, except a few titles. I actually had 2x5870, and just recently downgraded in anticipation of the 6970/6990. I also have 3x 1920x1080 monitors, and am using a single 5870 on one of those...well, actually, Im using 2, but not in 3D

But remember that those few titles are the sole reason for the whole industry still moving at all.

 

[cadaveca]

But remember that those few titles are the sole reason for the whole industry still moving at all.

I'm not the one complaining about any game's performance. I'm the one upgarding my machines to meet those performance needs, instead of complaining about it.

Maybe you should consider that.

 

[nesco1801]

yeah, and poor me has to tweak and sh**

 

[cadaveca]

yeah, and poor me has to tweak and sh**

I don't know why you need to tweak..I've yet to see any game that actually does a poor job of customizing settings based on installed hardware. I mean sure, you might not be playing at you monitor's natural resolution, but the need to tweak anything comes from you adjusting things away from those settings the app has when it first launches.

Even GT4 plays decently on default settings. Metro2033..well...

Oh, and when I say you, I don't mean you specifically, nesco.

 

[DrPepper]

Lower the graphics settings down to a console's graphics settings (720p as well) and it should do much better. It's simple really.

Also the console versions are all DX9 with all their hardware resources dedicated to the game. My 8800GT is leagues faster than any console in DX9 never mind a GTX5xx and 5xxx.

 

[nesco1801]

Well tbh Metro 2033 was the only game not played at max.
And sometimes in Crysis I set objects to Gamer and sometimes not

So that's how I "tweak"


PS

and there's this Black Shark that's killing me So yeah, mostly Eastern block is reason to upgrade

EDIT:link

 

[erocker]

Also the console versions are all DX9 with all their hardware resources dedicated to the game. My 8800GT is leagues faster than any console in DX9 never mind a GTX5xx and 5xxx.

All settings maxed out, except resolution set at 1280x720 (a bit better than on console) my setup has no issues at all running Metro 2033. Also take into consideration 30 fps is all that is required for a console.

 

[Bundy]

Most likely, not. High-end users don't spend exponential amounts of money for miniscule gains, just to get 40FPS. The review should be taken for what it's worth...showing where high-end cards need more vram, and where they don't.

The review is worth very little IMO.

They do not compare cards for gaming experience, instead they compare how badly each fails at settings that would not be used by a gamer. Of the 5 charts, only 2 are useful for comparison.

As an indicative buying guide, the review shows that people who want to run with high settings should get higher memory cards. As a comparison of performance, the review is fail.

 

[cadaveca]

instead they compare how badly each fails at settings that would not be used by a gamer.

Huh? Maybe your defination of gamer is flawed. The average PC gamer is 34 years old, and has some money to burn(hence buying games in the first place). I use "high-end" stuff, becuase I've been playing games for near on 30 years now. I might have been a "gamer", before you were born.

As an indicative buying guide, the review shows that people who want to run with high settings should get higher memory cards. As a comparison of performance, the review is fail.

I agree. And to me, this is important, as many people are downplaying the need for more than 1GB on videocards. This highlights some situations where 1GB doesn't suffice.

More importantly, as an Eyefinity user, I am more than aware of this need for larger framebuffers, so it's always good to see things I've mentioned before, shown again by other people, in other situations. So my opinion on the whole thing is a bit biased. To me, it's useful.

 

[Bundy]

Huh? Maybe your defination of gamer is flawed. The average PC gamer is 34 years old, and has some money to burn(hence buying games in the first place). I use "high-end" stuff, becuase I've been playing games for near on 30 years now. I might have been a "gamer", before you were born.

Uh? I'll explain. Gamers don't play at 30 fps, irrespective of their hardware. The review should have compared at 4x AA.

BTW - I am no spring chicken either

 

[Black Panther]

The way I see it is that it appears that the GTX580 is more 'future-proof' than the 5970.

What I can say is that I game on 2560x1440 and barely see any difference between 2xAA and 4xAA let alone 8xAA....

Currently I'm using 8xAA on Fallout New Vegas, but that's only because I don't see any drop in fps between it and 4xAA (and FONV is a Dx9 game anyway so one can't compare Dx11 effects here)...