GeForce GTX 580 vs. Radeon HD 5970 2GB Performance @ [H]

hackztor said:
at 1920 these cards probably get 60+. look at normal review cards for that. People paying 500 for a card probably have something a little bigger than a 24-27 inch monitor (150-300 dollars for monitors these sizes). Ive seen my gtx580 blast any game maxed out at 1920. Even at 2560x1600 the frames are acceptable with maxed settings. Go look at the normal review for these cards and you can see your 1920 res. The point of this article was to show where the ati card chokes with the less ram even though it gets higher fps sometimes.
Click to expand...

I guarantee that sales of the 480/580/5870/5970/6970 far outstrip sales of 30 inch monitors. Plenty of people buy $500 cards and play on 24 inch monitors - because as the previous poster pointed out, even with a $500 card you still can't run good framerates with high AA and all settings turned up on some games. So the 1920 benchmarks would be useful - although I understand [H]'s position in not providing them based on their testing philosophy (not enough time / extrapolations can be made for max playable settings, etc).
 
but it would of been interesting if it choked at 1920 at high AA (personally i never get an 5970 as 2x5870 or 2x5850 tend to be faster for some reason and provide more consistent frame rates)

not that i would buy an ATI card as drivers are to unpredictable and so is the installer (was longer but more of an rant but facts about the drivers and CAT are quite true)

i agree with Forceman comment
 
Great review, &...

The GeForce GTX 580 allows a consistently higher level of the gameplay experience compared to the Radeon HD 5970. We were able to game at higher settings with the GTX 580 than we were with the Radeon HD 5970. The most important factor, beyond framerates, is the visual quality and experience returned by the product. The GeForce GTX 580 allows a more immersive, smoother, and consistent quality of gameplay.
Click to expand...

...I'd say that's pretty impressive on nVidia's/the 580's part. High-end single GPU solution is always the way I go, and for the 580 to beat-out the 5970 in such a way as to warrant the comment quoted above, is rather impressive.

I recently moved from the 480 to the 580 and have noticed better results all around with noise, temps and performance. It differs per-game, but there's still an improvement in performance that I'm noticing and liking, and the noise and temps of the 580 are a massive improvement over the 480 from what I've experienced thus far.

kcthebrewer said:
1920x1200 does not stress these cards (except for possibly Metro 2033).
Click to expand...

Generally, but depends on the game and how much you push the settings. But, that's what's nice... 1920x1200 widescreen is a great size res, and the most common at this point, and with the kind of horsepower of something like the 580, you have the luxury of getting full-out eye candy without sacrificing performance.
 
Great review, thanks for taking into account the vram limitation choking ATI cards.

Since this is [H], I hope requesting an over-clocked comparison or two for future reviews isn't out of line :)
 
I have a 5970 and I'm most likely upgrading to 580 SLI soon.

One of the important things to factor into your buying decision is that a single 5970 is your performance ceiling. You cannot add more and improve Eyefinity frames. In games that don't play nicely with Crossfire - and despite what AMD's marketing department would like you to believe, there are a great many of them - you are running at single Radeon 5850 speed.

A 580 gives you more headroom with SLI and you still have near 5970 performance when you are forced to disable SLI.

Of course this could change when ATI brings out the next 5870 equivalent but it would have to wreck the 580 for me to consider Crossfire over SLI after a year of misery with Eyefinity + Crossfire and crap drivers.

Obviously, it goes without saying that price is not factored into this argument. I'm simply looking at performance ceilings and upgrade paths for 1-3 screen, high end setups.
 
Thanks for the comparison. It's interesting, but certainly not apples to apples. I don't like to run multiple cards. As your article a few weeks ago pointed out the venders need to stay current on the drivers/game profiles to make the cards work up to their potential. Frankly with a $400 - $500 dollar investment (per card), I don’t trust either of them to keep up with the latest software. I have ran multiple GPUs a few times, but always prefer a strong single GPU over a pair of lessors.
 
Kyle:

If you get the chance to get your hands on the 4GB 5970, please re run some of the tests. I know that this is a good showing for Nvidia, but it does show some weak spots. But I do admit this is the first Green card I would buy in a long time. But I would like to see what the 2 GB per card would let this thing do. It was already shown to make a huge impact on hi-res games that used MSAA and such. If one in a reasonable range can be found, I think this story should be revisted.

At least they have one thing right.... actually Kyle and his team tend to get a very many of their things correct and their testing standards I find are better than most. We still can do Triple or more monitors from one card.... some 5870's can support 6. If Nvidia could dodge that bullet I would consider it, but thats tooo much money for this persons account this month... maybe I will contribute to the economy in Dec.

Even thought I admittedly lean red, the 460 and the 580 have caught my eyes... but I know that these products are in response to the card I have been using for nearly a year. The real big one (question) is the 6900 series going to just be a refresh??? Or are they going to break out some of their new technology???

Its exciting times...

I haven't read the forums because I know it will be the same Green this and Red this... and what it should be is GODDAMN isn't competition GREAT.... Team Green should bow its head to team Red and say thank you for pushing Nivida up from the rubble of the horrid card to respond to ATi's boggling and card killing 9700. Then it was ATi's turn after their horrid 2xxx series of cards that literally were trash... I think those cards proved that they leaked more current than they used....

This is what a good fight is about, but waiting to see what you competitor has done and responding a year later, and it takes the 580 card to finally really deal a true hit to ATi(AMD).... And yet, if that is all Nvidia has... what if the 69xx comes back and puts the 580 in the back seat for another year??? It may not get back seated but it may not be near the top...

I think what give AMD a leg up is they have a great CPU division that had the fastest proc's for a long time.... They can take lessons learned and move on... NVidia is a one big die company and it seems that ATi wants to be more modular..... and yet powerful... and it seems to be working.

I'm 100% sure the tests would have swung around had they tested with the 4 GB 5970's
 
This sounds like a design flaw. The 5970 lacks the memory capacity to make its extra speed useful. Easily fixed. Bump it up to 3 GB.
 
[QUOTEThis sounds like a design flaw. The 5970 lacks the memory capacity to make its extra speed useful. Easily fixed. Bump it up to 3 GB.][/QUOTE]

Its already been done. There are 5970's with 4GB, 2 to each core
 
IceEmQuick said:
I'm 100% sure the tests would have swung around had they tested with the 4 GB 5970's
Click to expand...

Probably - but the 4GB edition is not in the same price bracket as a 580. The cheapest 4GB 5970 on Newegg is a Black Edition XFX priced at $1,199! Terrible value at this point.
 
v0rtic3s said:
Generally, but depends on the game and how much you push the settings. But, that's what's nice... 1920x1200 widescreen is a great size res, and the most common at this point, and with the kind of horsepower of something like the 580, you have the luxury of getting full-out eye candy without sacrificing performance.
Click to expand...

What are you referring to full-out eye candy? It can play pretty much anything (again Metro 2033 is an exception) at 8x AA at 2560x1600 with all in-game settings maxed.

As from the results and personal experience (I am one of the outliers that uses a 30"), if your monitor only supports 1920x1200 this video card is probably not a good purchase as there are many other cards that can play at that resolution with all quality settings enabled.
 
Forceman said:
I guarantee that sales of the 480/580/5870/5970/6970 far outstrip sales of 30 inch monitors. Plenty of people buy $500 cards and play on 24 inch monitors - because as the previous poster pointed out, even with a $500 card you still can't run good framerates with high AA and all settings turned up on some games. So the 1920 benchmarks would be useful - although I understand [H]'s position in not providing them based on their testing philosophy (not enough time / extrapolations can be made for max playable settings, etc).
Click to expand...

I've had a 27" 1920x1200 monitor for awhile, now....;) I don't think it is entirely necessary to run at higher resolutions, especially when using FSAA. It seems to me that 8X FSAA @2560x1534 is a very limited IQ proposition. The *whole point* to FSAA is that it allows you simulate the IQ of much higher resolutions while actually running in lower resolutions.

I actually think this was a very good, very smart comparison. It had not occurred to me to make this comparison, in fact (I am definitely aging and the quantum-state gluons are flowing nowhere near as fast as they used to...;)). Good article, though--I looked at the title for 2-3 minutes before I realized this wasn't actually a *6970* article--which I kept thinking was way ahead of NDA release!...;)
 
Last edited:
I find this review pretty bias to be honest, the 5970 clearly proves it's the better card by winning the majority of the benchmarks even if you give the slower 580 a win in F1 despite it being slower...and call Civ5 undetermined due to the clearly erratic behaviour of ALL of the cards in that game...the 5970 still clearly wins the 3/5 other benchmarks without question.

What is most disturbing is that the 5970 was clearly marketed as an overclockers card, they put on a HSF rated for 400W of thermal output on a 300W TDP card, and released official overvolt tools. The marketing for the card was all overclocking orientated as well.

I'd wager that most gamers buying this card know full well that it's 5870 hardware downclocked and that actually 5870 speeds and above are possible pushing this card significantly past it's stock speeds, decent 5870 overclocking speeds can be achieved by people willing to voltmod the card, even on stock cooling, we're talking towards 1GHz GPU and 5GHz Memory.

So stock frame rates it wins.
Overclocking potential it clearly wins, about 25% increase compared to the 580's 8-10%
Price it wins.

Plus, you can always hook up two GTX 580 cards in SLI that will surely blow the 5970 away
Click to expand...

You neglect to mention that you can put 2 5970s in QuadFire or the more sweet spot of a 5970 and 5870 for TriFire.

Like all the other recent reviews of the 580vs5970 have confirmed, such as anandtech, the 5970 is clearly the faster card, it wins the majority of the benchmarks in average frame rate and minimum frame rate. Anyone after a more balanced view of the performance differences would do well to read the 580 reviews where it's compared to the 5970 also, try and ignore conclusions that do not follow from the numbers.
 
PrincessFrosty said:
I find this review pretty bias to be honest, the 5970 clearly proves it's the better card by winning the majority of the benchmarks even if you give the slower 580 a win in F1 despite it being slower...and call Civ5 undetermined due to the clearly erratic behaviour of ALL of the cards in that game...the 5970 still clearly wins the 3/5 other benchmarks without question.
Click to expand...

If you had read the article, in Civ5 you will see that the game play experience other than the loading pauses the 5970 gave a worse experience to the end user at those settings.

Framerate being higher doesn't always mean a better experience if it is stuttering or laggy.

I don't see how the article is biased, there is proof that there are issues with the 5970 2GB at 8xAA @ 2560x1600.

It would be like if someone showed you the stats of a baseball player with runners in scoring position and saying it is biased.
 
If nothing else ATI or Nvidia if your running maxed out you still need two of what ever it is HD5970 or GTX580.
 
While It would be nice to see a review where both the 5970 and 580 are overclocked, the 5970 is still going to run out of memory at 8x MSAA in F1 2010 and Civilization V.
It makes me wonder if it is going to run out of memory in Eyefinity resolutions for these games also.
I think it would probably be fine with lower AA settings up to 3x 1920x1200 but I wonder if it could even run these games at 3x 2550x1600.
 
PrincessFrosty said:
You neglect to mention that you can put 2 5970s in QuadFire or the more sweet spot of a 5970 and 5870 for TriFire.
Click to expand...

And you seem to forget that dual-gpu scaling is already poor enough, and anything beyond dual-gpu scaling is worse. I haven't seen any performance comparisons between GTX 580 SLI + HD 5970 QuadFire, but from the performance reviews I have seen of QuadFire I'd be pretty confident in SLI GTX 580s outperforming QuadFire HD 5970s.
 
kcthebrewer said:
If you had read the article, in Civ5 you will see that the game play experience other than the loading pauses the 5970 gave a worse experience to the end user at those settings.

Framerate being higher doesn't always mean a better experience if it is stuttering or laggy.

I don't see how the article is biased, there is proof that there are issues with the 5970 2GB at 8xAA @ 2560x1600.
Click to expand...

I discounted F1 and Civ5 and assumed that for whatever reason that [H] subjective opinion about lower but more stable frame rates being "better". The 5970 still clearly beats the 580 in the 3 other tests.

The article is bias because even the worst interpretation of the numbers towards the 5970 still show it winning the majority of the benchmarks,(clearly winning 3 out of 5, and beating the 580 in the other 2 in terms of frame rate but reviewer subjectively marking it down) and this is also the case on other review sites, the cards trade blows but generally speaking the 5970 wins in way more places both in minimum and average frame rate.

Yet the review comes out in favour of the 580, which is more expensive no less, the only justification doesn't line up with the results from the benchmarks, there was no complaints about the other 3 games about how the 5970's frame rate was clearly higher, so how can a more expensive part, winning less benchmarks still be favoured?

5*General said:
And you seem to forget that dual-gpu scaling is already poor enough, and anything beyond dual-gpu scaling is worse. I haven't seen any performance comparisons between GTX 580 SLI + HD 5970 QuadFire, but from the performance reviews I have seen of QuadFire I'd be pretty confident in SLI GTX 580s outperforming QuadFire HD 5970s.
Click to expand...

Actually TriFire scales incredibly well, people with a 5970 @ 5870 speeds + a 5870 are getting very good scaling in a wide range of games, like everything else multi-gpu there are games that favour multi-gpu and those that don't. QuadFire does tend to see more limited returns for spend but again it depends on the game, some games do scale quite well with 4 GPUs. Anyway the reviewer boasts SLI as if it's unique to the 580 and discounts the 5970 again for no reason, however it was intended, it still appears really bias to me.
 
WD [h] - I agree single GPU is the way to go, this is good info regarding cards now priced similarly. I don't see this as a biased view, just true info - Kyle, correct me if I'm wrong has been using a 5970 for ages - says a lot...I've been down SLI road 3 times, no more 4 me, don't like the lag/stuttering...take it all for what it's worth..
 
So using your own personal experience, can run the games in the test that Brent did and you didn't experience the issues that he brought up?

You have the card and the monitor to do it. If you would like to provide useful feedback using a stock clocked 5970 instead of just saying [H] is biased and wrong, I am sure everyone would appreciate it.

Brent's gameplay evaluation is that 8x AA @ 2560x1600 on a 5970 2GB has limitations providing a suboptimal gaming experience and that the 580 GTX doesn't have this issue.
 
Sorry if this is a stupid question, but why is RAM an issue with the 5970, when it has 2 gigs of RAM as opposed to GTX 580's 1.5 gigs? Can the 5970 only use half of the RAM at one time?
 
addictedto60fps said:
Sorry if this is a stupid question, but why is RAM an issue with the 5970, when it has 2 gigs of RAM as opposed to GTX 580's 1.5 gigs? Can the 5970 only use half of the RAM at one time?
Click to expand...

The 5970 has 1GB of RAM per GPU (and it has two GPUs). In order for Crossfire (and SLI works the same way) to work, the VRAM on each has to be mirrored. So while there is 2GB total on the card, it's effectively 2x1GB with each 1GB partition storing the same data.
 
kcthebrewer said:
So using your own personal experience, can run the games in the test that Brent did and you didn't experience the issues that he brought up?

You have the card and the monitor to do it. If you would like to provide useful feedback using a stock clocked 5970 instead of just saying [H] is biased and wrong, I am sure everyone would appreciate it.

Brent's gameplay evaluation is that 8x AA @ 2560x1600 on a 5970 2GB has limitations providing a suboptimal gaming experience and that the 580 GTX doesn't have this issue.
Click to expand...

He's recommending a slower, more expensive card over a cheaper, faster card. In the benchmarking the 5970 shows it is clearly more capable, they mark down certain benchmarks based on their own subjective opinion, the numbers clearly show the 5970 faster.

We can either trust the benchmarks or we can't, if they're posting numbers and graphs that show the 5970 is clearly faster, why should we ignore their numbers and take their subjective opinion on the matter? That's exactly why we have performance measurements like these, because they're not subjective.

It's been established which card is faster by all the reviews of the 580 that include the 5970 speeds, to recommend a card which is both slower and more expensive is just crazy, how they can come to that conclusion based on their own review numbers is beyond me.
 
PrincessFrosty said:
how they can come to that conclusion based on their own review numbers is beyond me.
Click to expand...

It's called having a brain and using it.

"Performance Numbers" don't mean jack diddle when they're going up against "End-User Experience".

If they wanted to really make things one-sided they could have picked a slew of games that don't use Crossfire, and we could have gotten a bunch of pretty graphs of the 580 murdering the 5970.

Would that have made you happy?
 
An obvious conclusion of Vram (texture) thrashing.

What did you expect @ 2560x1600 + 8xAA?

Why wasn't 4gb 5970 tested or in the mix?
 
Last edited:
PrincessFrosty said:
He's recommending a slower, more expensive card over a cheaper, faster card. In the benchmarking the 5970 shows it is clearly more capable, they mark down certain benchmarks based on their own subjective opinion, the numbers clearly show the 5970 faster.

We can either trust the benchmarks or we can't, if they're posting numbers and graphs that show the 5970 is clearly faster, why should we ignore their numbers and take their subjective opinion on the matter? That's exactly why we have performance measurements like these, because they're not subjective.

It's been established which card is faster by all the reviews of the 580 that include the 5970 speeds, to recommend a card which is both slower and more expensive is just crazy, how they can come to that conclusion based on their own review numbers is beyond me.
Click to expand...

You obviously never experienced texture thrashing. It is when your video card runs out of memory and start paging memory back and forth doing 2x the work just to render the image.

A benchmark is nothing more than numbers at different points of a graph.
Without dissecting the numbers a 60fps average frame rate can be useless when the card deeps into teens frequently.

This is why Brent showed plot graph showing where the big dip was @ 8xaa.

1289507336SzLWKc8bvR_1_1_l.gif


As you can see @ 8x AA there are sections of the game where frame rates plummet to unplayable scenario on the 5970 while 2gb 5870 and GTX580 does not.

@ 4x AA 5970 doesn't have a problem.

1289507336SzLWKc8bvR_1_2_l.gif
 
Quick question

Is 8x AA really that important at 2560x1600? Is it that much better than 4x or even 2x?

I run at 1600x1200 and find 2x provides more than enough smoothing for me.

Appreciate its a good test for reviews/benchmarks but in real world? Seems a waste of processing power to me.

Maybe would make an interesting article on levels of AA and ultra hi-res? Comparisons of the quality of effect acheived and the hit in performance to reach it.
 
daglesj said:
Quick question

Is 8x AA really that important at 2560x1600? Is it that much better than 4x or even 2x?

I run at 1600x1200 and find 2x provides more than enough smoothing for me.

Appreciate its a good test for reviews/benchmarks but in real world? Seems a waste of processing power to me.

Maybe would make an interesting article on levels of AA and ultra hi-res? Comparisons of the quality of effect acheived and the hit in performance to reach it.
Click to expand...

Personally, I don't use AA at all with 2650x1600. Never seems to need it IMO.
However on my old 24", it was helpful with image quality.
 
Vagrant_Zero said:
It's called having a brain and using it.

"Performance Numbers" don't mean jack diddle when they're going up against "End-User Experience".

If they wanted to really make things one-sided they could have picked a slew of games that don't use Crossfire, and we could have gotten a bunch of pretty graphs of the 580 murdering the 5970.

Would that have made you happy?
Click to expand...

Is that so? Then why even include frame rate numbers at all? "End-user Experience" is subjective and subject to bias and numerous other human mistakes, even if your intentions are genuine/pure

The more games you test the more pronounced the 5970s win becomes because the 580 actually only beats the 5970 hands down in a couple of games that we've seen, review sites like anandtech using a much larger number of games for testing show the 5970 win in most games and a handful of draws and losses for the 5970 vs the 580. If anything [H] small set of games for testing has encouraged their bias view, clearly the card is faster overall but there are some circumstances where struggles, these could very well just be bugs or bad management from the game which could be cleared up in patches.

Raw power wise the 5970 wins out, even at stock, completely and totally disregarding the whole overclocking scenario it's still more powerful. If you can find situations that eat away vRAM like crazy such as 8xAA at 2560x1600 then sure you'll demonstrate a card that with more memory "wins", the same way you can prove that Nvidia wins if you throw ridiculous amounts of tessellation at the card such as unigine.

daglesj said:
Quick question

Is 8x AA really that important at 2560x1600? Is it that much better than 4x or even 2x?

I run at 1600x1200 and find 2x provides more than enough smoothing for me.

Appreciate its a good test for reviews/benchmarks but in real world? Seems a waste of processing power to me.

Maybe would make an interesting article on levels of AA and ultra hi-res? Comparisons of the quality of effect acheived and the hit in performance to reach it.
Click to expand...

I was going to bring this up as well, but didn't bother because it's a hard point to defend, but since someone else has I might as well put in my $0.02. Purely my opinion of course but I own a 2560x1600 panel, it has one of the highest PPIs of any standard sized PC monitor out there, as such AA is less necessary at native resolution, it's still nice to have but I find that 2x is enough to remove the annoyances of jaggies and jaggie crawling, and anything higher than 4xAA is not not necessary. I'd be willing to bet that in a double blind study of 4xMSAA vs 8xMSAA that most people couldn't tell the difference except in some very specific circumstance. On a moving scene in the game it is of little to no benefit, especially weighted against the performance drop.

That's another thing that frustrates me about [H] reviews, they aim for very low frame rates a lot of the time, 30fps is not a good target fps, you should be looking for 60 average at least to maintain fluid gameplay, that's a whole other story though.


MrXROK said:
You obviously never experienced texture thrashing. It is when your video card runs out of memory and start paging memory back and forth doing 2x the work just to render the image.

A benchmark is nothing more than numbers at different points of a graph.
Without dissecting the numbers a 60fps average frame rate can be useless when the card deeps into teens frequently.

This is why Brent showed plot graph showing where the big dip was @ 8xaa.

As you can see @ 8x AA there are sections of the game where frame rates plummet to unplayable scenario on the 5970 while 2gb 5870 and GTX580 does not.

@ 4x AA 5970 doesn't have a problem.
Click to expand...

I have experienced it, I did frequently with my 4870 512mb crossfire setup in more modern games (at the time) like GTA IV and DX11 Age of Conan. I understand texture thrashing.

There is a single anomalous dip below 30fps in F1at that setting, for almost the entire run through the frame rate is higher than the 580, for all we know this is a bug with F1 or an anomaly with a specific track or configuration, they hand a "win" to the 580, clearly the slower card of the 2.

Even if we grant the F1 and Civ5 are just flat out failings of the card and give the win to the 580 that's still only 2 wins vs 3, if they want to recommend a more expensive card over a cheaper one the least they could do is have it be the clear winner in the majority of the benchmarks rather than 2 rather "questionable" wins.
 
Frosty, you do realize that raw FPS aren't everything don't you? For example, you could have a multi GPU setup cranking out high max FPS in games, and high(er) average FPS, but crappy low FPS. It could also (depending on the game) be jerky, and stutter.

Now lets say you have a powerful single GPU. Sure the max FPS aren't as high (70 v.s. 90 or whatever), and the average is maybe 1-2 FPS lower, but the minimum is 4-5 higher and there is zero stutter or jerking between frame renders. What setup would you rather play on? As a current 5870 owner, the 580 is looking kinda good right now, and I can kiss the horrid driver support goodbye.
 
PrincessFrosty said:
Is that so? Then why even include frame rate numbers at all? "End-user Experience" is subjective and subject to bias and numerous other human mistakes, even if your intentions are genuine/pure

That's another thing that frustrates me about [H] reviews, they aim for very low frame rates a lot of the time, 30fps is not a good target fps, you should be looking for 60 average at least to maintain fluid gameplay, that's a whole other story though.
Click to expand...

I agree with you on both accounts.

I'd never turn up the eye candy if I was getting 50-60 FPS. 30 FPS is just too uncomfortable to game at in my opinion, but I do think others feel the same way as well.

Concluding the 580 the winner I also thought was a bit odd, although the information provided (without the subjective opinion) as really useful! It doesn't look like to me that the 580 really won at anything especially considering price. Then again, maybe Kyle is right, maybe gaming on the 580 does somehow provide a better 'experience'. If that's the case, then maybe even minimum FPS numbers mean squat. Maybe it's something intangible.

Not sure...
 
According to the review, the 580 is the best one in resolutions above 4x 2560x1600. At resolutions 4x 2560x1600 and below, it seems that the 5970 is the clear winner.
[H] judges (and always have) gameplay experience at highest possible resolution, so I can't see that they have done something wrong or differently here?

If I were to use 8X AA with 2560x1600, I would get best gameplay experience with the 580. If I were to use 4X AA with 2560x1600, I would get best gameplay experience with the 5970 and most likely I would get the best gameplay experience with a 5970 with all resolutions below 2560x1600 with 8X AA. :)
 
Tamlin_WSGF said:
According to the review, the 580 is the best one in resolutions above 4x 2560x1600. At resolutions 4x 2560x1600 and below, it seems that the 5970 is the clear winner.
[H] judges (and always have) gameplay experience at highest possible resolution, so I can't see that they have done something wrong or differently here?

If I were to use 8X AA with 2560x1600, I would get best gameplay experience with the 580. If I were to use 4X AA with 2560x1600, I would get best gameplay experience with the 5970 and most likely I would get the best gameplay experience with a 5970 with all resolutions below 2560x1600 with 8X AA. :)
Click to expand...

The 5970 generates higher frame rates. The problem is that the gaming experience isn't as consistent as it is with single GPU cards. In other words the frame rates on the 5970 are likely to fluctuate much more often than they will on the GeForce GTX 580.
 
Dan_D said:
The 5970 generates higher frame rates. The problem is that the gaming experience isn't as consistent as it is with single GPU cards. In other words the frame rates on the 5970 are likely to fluctuate much more often than they will on the GeForce GTX 580.
Click to expand...

I understood it as it was only when it hit a memory limit:

4X 2560x1600 no memory limit:

Now that we are not bound by memory limitations, you can see how much smoother, and less erratic, the performance line is for the Radeon HD 5970. There are no wild swings up and down in performance, it is more consistent. Along with that, performance differences have increased and the Radeon HD 5970 is now 25% faster than the GeForce GTX 580.
Click to expand...

1289507336SzLWKc8bvR_1_2.gif


8X 2560x1600:
At this setting, with 8X MSAA enabled, the Radeon HD 5970 was very inconsistent in framerates. You can see this clearly on the graph, the GTX 580 and HD 5870 2GB lines are more smoother compared to the Radeon HD 5970. The Radeon HD 5970 experiences wild fluctuations in performance, and erratic behavior. This is because we are running into VRAM limits with 1GB per GPU in this game at 8X MSAA at 2560x1600.
Click to expand...
1289507336SzLWKc8bvR_1_1.gif


Meaning, for 4X AA 2560x1600 (and probably below), get a 5970, for 8X AA 2560x1600 (and probably above) get a 580?
 
So how many 2560x1600 gamers we got here compared to 1920*1200 and 1680*1050? lol

Pointless comparision, 1920*1200 results or don't bother imo.
 
daglesj said:
Quick question

Is 8x AA really that important at 2560x1600? Is it that much better than 4x or even 2x?

I run at 1600x1200 and find 2x provides more than enough smoothing for me.

Appreciate its a good test for reviews/benchmarks but in real world? Seems a waste of processing power to me.

Maybe would make an interesting article on levels of AA and ultra hi-res? Comparisons of the quality of effect acheived and the hit in performance to reach it.
Click to expand...

It depends on the user and game I think. A racing game I would rather 16x SSAA. With a FPSI can live with 2x AA or none depending on the game.

If you are getting a card for 2560x1600 might as well get a card that doesn't texture thrash with AA. That's what 4gb 5970 is for.
 
You must log in or register to reply here.
Back
Top