GTX 680 vs. Radeon HD 7970 - Multi-Display Showdown @ [H]

This smoothness thing bothers me. Especially in summary when recommending the slower framerate configuration due to it. In practice it has always been the higher fps the smoother so I'm not computing here. Some further analysis would be nice.
 
VMaxx said:
This smoothness thing bothers me. Especially in summary when recommending the slower framerate configuration due to it. In practice it has always been the higher fps the smoother so I'm not computing here. Some further analysis would be nice.
Click to expand...

Old school 3dfx thinking, fps is everything. We've learned long ago, that FPS isn't everything. The whole basis for our gameplay evaluation of the experience that a card delivers, which we started in 2003, is because we have learned that fps is not everything, gameplay has evolved past this as the primary measurement for a better gameplay experience.
 
Today's look is solely squared off between the AMD Radeon HD 7970 GHz Edition and GeForce GTX 680 in both single-card form and CrossFireX and SLI.
Click to expand...

In the Red Camp, you can get overclocked XFX Radeon HD 7970 GHz Edition for $419 after $30 MIR right now, making CrossFireX a $840 investment, and you'll get better raw framerate performance than shown here today with the reference clocked 7970 cards used in our evaluation.
Click to expand...

Could we get some clarification on exactly which AMD cards were reviewed as in the review the introduction and conclusion contradict each other.
 
Dont get it , gtx 680 feels smoother ??????

Sleeping dogs
Radeon 7970 ghz edition - 5760x1200 High Quality - 16XAF, gtx 680 sli not playabel at theese settings but feel smoother ????

Max payne - same settings, so here i can understand that gtx feels smother, even if radeon 7970 has higher avarage fps

Battelfield 3 singel player
Radeon 7970 ghz edition - 5760x1200 4xMSAA - 16AF, gtx 680 sli not playabel at theese settings but feels smoother ????


Battelfield 3 multi player
Radeon 7970 ghz edition - 5760x1200 2xMSAA - 16AF, gtx 680 sli not playabel at theese settings but feels smoother ????

what am i missing since gtx680 is better according to the testers ?????, when gtx 680 sli not are playabel at same setting in 3 of 4 games ???
 
I would love to see something about Radeon Pro though! [H] is about best bang for buck is it not, I wouldn't give a crap if I had to install some software and got a better experience.

Also was the micro stutter apparent in the higher FPS apples to apples? If you dropped from the 4x MSAA to the 680's 2xMSAA was it just as smooth etc.

This has opened a can of worms now :p
 
osse said:
Dont get it , gtx 680 feels smoother ??????

Sleeping dogs
Radeon 7970 ghz edition - 5760x1200 High Quality - 16XAF, gtx 680 sli not playabel at theese settings but feel smoother ????

Max payne - same settings, so here i can understand that gtx feels smother, even if radeon 7970 has higher avarage fps

Battelfield 3 singel player
Radeon 7970 ghz edition - 5760x1200 4xMSAA - 16AF, gtx 680 sli not playabel at theese settings but feels smoother ????


Battelfield 3 multi player
Radeon 7970 ghz edition - 5760x1200 2xMSAA - 16AF, gtx 680 sli not playabel at theese settings but feels smoother ????

what am i missing since gtx680 is better according to the testers ?????, when gtx 680 sli not are playabel at same setting in 3 of 4 games ???
Click to expand...

You have been here 3 years and still don't understand their review methodology? Either you are a paid troll or intellectually dishonest. Which is it?
 
yes i understand the methodlogy and i been reading this for lot more than 3 years, trolling, i state a fact that gtx 680 sli is not playabel according to the testers in 3 of 4 games in this review with same settings as 77970, gtx 680 sli has to lower settings , and when u lower settings u get higher framrates. if they stated that on same settings gtx sli was smother i would have understand it. What is trolling abaut that ?
 
Brent_Justice said:
We don't use any third party utilities to change the out-of-box experience between NVIDIA and AMD GPU performance/gameplay experience. We feel it is up to AMD to implement these features to improve its platform, if it so chooses, if it doesn't, well then we are left with what we are left with as the experiences between both. If we use a third party utility for AMD, then we'd have to use one for NVIDIA, and then its just not fair.
Click to expand...

Fair enough, but would it be possible to explore this in its own separate article? Perhaps then we can learn a bit more about what goes on behind the scenes of multi-gpu rendering. Just a thought. :)
 
If FPS isn't everything, perhaps it is time to convert benchmarks over to something more meaningful that can mathematically demonstrate "smoothness".

I might suggest using a similar methodology to other sites - frame latency. Basically measure the average amount of time it takes to render frames and then call out the number of frames outside the standard deviation and compare that way. This method would more clearly show cards that render inconsistently and therefore would "feel" less smooth even at lower average frame rates. Basically, it would just put the science behind your observation while also pushing forward the [H] leadership in benchmarking methodology.

It might also start to force vendors into writing drivers that focus on more consistent rendering instead of inconsistent "surges" that result in a worse gaming experience. It seems like right now, AMD might be "cheating" for lack of a better word, by masking issues through higher FPS averages. Cheating might be too strong a word (since really either company could wind up doing this) - maybe it is just laziness or the wrong approach to optimization.
 
IMO it should count for a bit more that you can crank up the settings further on the 7970s than the 680s... this whole thing seems like a draw, with only your personal preferences making the decision for you.

And, to beat my favorite dead horse, I still think that some form of testing should be done to ground the comments about "smoothness".

I think it's hard to argue from the data that the 680s are preferable... sure SLI may be smoother at a given framerate, but you could just turn down the 7970 CF settings to 680 SLI levels, and compensate for the "smoothness" decrease by pushing framerates even higher.
 
For those interested in the "smoothness" TechReport had an article last year that sort of explains the smoothness or lack of when using Multi-GPU setups.

Apologise in advance if link to another site is against the rules.
 
I recall reading at one point earlier this year that [H] testers were hoping to obtain software that could be used to subjectively test "smoothness." I'm guessing either it's still in the works or that it turned out not to be feasible for whatever reason.
 
osse said:
yes i understand the methodlogy and i been reading this for lot more than 3 years, trolling, i state a fact that gtx 680 sli is not playabel according to the testers in 3 of 4 games in this review with same settings as 77970, gtx 680 sli has to lower settings , and when u lower settings u get higher framrates. if they stated that on same settings gtx sli was smother i would have understand it. What is trolling abaut that ?
Click to expand...

Exactly, and well said.
If anyone doesn't understand this logic then they should learn what they are talking about before throwing the "T" word around.
 
Phishy714 said:
Reading does help!

He reviewed the GHZ edition card, which is in fact $419 AR. The $359 card you called out is not a GHZ edition card. And no - you cannot assume that its the same thing because overclocking was not taken into consideration at all in this test. You also cannot guarantee that EVERY non-ghz 7970 will be able to OC to ghz levels. The "boost" mentioned in the description does not make it a GHZ card. Its not branded as such for either for it doesn't even boast a 1ghz core clock from stock.

If you take OC into consideration, its very likely that AMD will pull further ahead in raw FPS numbers, but sadly, this wasn't tested.

Reading rocks!
Click to expand...

Yes reading, do tell :rolleyes:

" you can get overclocked XFX Radeon HD 7970 GHz Edition for $419 after $30 MIR right now, making CrossFireX a $840 investment, and you'll get better raw framerate performance than shown here today with the reference clocked 7970 cards used in our evaluation."

But not to be too hard on ya, I think the [H] should seriously clarify what the hell they are saying here. They seem to contradict what card and what speeds were used.

Also I think we can guarantee that an older non-GHz card can hit GHz levels. It is a modest clock and anyone with a brand card in a decent case will hit it.
 
People who talk about Radeon Pro should note that afaik AMD still default CCC texture quality to (quality) instead of (high quality) so you get the direction they take their cards.
 
Semantics said:
People who talk about Radeon Pro should note that afaik AMD still default CCC texture quality to (quality) instead of (high quality) so you get the direction they take their cards.
Click to expand...

The [H] has done extensive tests of AA levels and texture quality, and oddly have never pointed out that Radeon cards have had poor or lesser texture quality.
 
Just do a max playable settings and then a smooth feel settings . It'd be nice to know at what point xfire feel smooth to get perspective on the issue.
 
funkydmunky said:
The [H] has done extensive tests of AA levels and texture quality, and oddly have never pointed out that Radeon cards have had poor or lesser texture quality.
Click to expand...
Depends who you talk to about said tests, short of the good old 2D quality test it's all just a shit storm. And...

Yes in still images it's quite hard to tell the difference but their choice of stock performance optimizations not per game profiles to tailored performance leads to more noticeable in motion characteristic of their textures, ie the shimmer, crawling and noise while in motion.
http://hardocp.com/article/2011/12/22/amd_radeon_hd_7970_video_card_review/
Though texture filtering was great with the Radeon HD 6000 series, AMD has sought to make it even better with the Radeon HD 7000 series. There was an issue with the Radeon HD 6000 series that sometimes produced a shimmering or crawling effect in the distance on far out textures on certain planes. AMD has improved filter kernel weights that should reduce this shimmering or crawling seen on textures. After having using the new Radeon HD 7970, we can state that AMD has indeed improved and fixed this issue.


In our testing, we found Skyrim to show this issue the most on rocky roads as we are traveling throughout the world on the Radeon HD 6970. We did see this shimmering and crawling. When we installed the Radeon HD 7970 it was gone, and the texture quality was better. It isn't visible in all games, and Skyrim seemed to be the worst offender, in fact, the only offender, we experienced in our testing. AMD has certainly tackled this issue, and texture quality should now be near perfect with the Radeon HD 7970.
Click to expand...
Part of which is due to AMD has always been slightly more saturated and sharper filtering and nvidia goes for a softer look, part of it has to do with default texture optimizations.

The point of it was not to say 7970 is shit quality to 680, maybe such a statement would be easy if it was a 5xxx or 6xxx card but the point was...

AMD seems push the edge in trying to balance IQ and get higher performance a bit more then nvidia. Anyways a lot of IQ is hard for avg user to even notice, sorta like phsyx wont miss it if never had it. So short of two systems side by side and testing you don't pick out any differences based on what you can remember a couple days back.
 
one thing i see right at the beginning of this test is that the card compared is the reference gtx 680 (2gb vram) against the ghz 7970 (6gb vram) when they should be comparing the Classified GTX 680 (4gb vram) instead. i would say the gtx680 will still come out on top but far better comparison
 
Semantics said:
So short of two systems side by side and testing you don't pick out any differences based on what you can remember a couple days back.
Click to expand...

[H] just made the point of micro-stuttering based on side-by-side, all be it at lesser GFX settings on the Nvidia side. There was nothing about texture quality. Your long winded post doesn't add anything to the reality of the situation. If you aren't commenting on this review, please do not attempt to derail it based on your own agenda. The review didn't at any time mention anything about either card having lower GFX quality, except for lower AA resolutions for the 680 to maintain playability at high resolutions matching the 7970. Not once did the [H] mention texture quality.
 
genkifd said:
one thing i see right at the beginning of this test is that the card compared is the reference gtx 680 (2gb vram) against the ghz 7970 (6gb vram) when they should be comparing the Classified GTX 680 (4gb vram) instead. i would say the gtx680 will still come out on top but far better comparison
Click to expand...
I think you misread something it mentions no where that it's using sapphire's card it's just a generic GTX680 to a generic 7970 Ghz edition, which is 2gb vs 3gb, and in a platform of a very large resolution plus AA which does make a difference. But it's not unfair in the aspect of 2 generic top of the line cards competing in the same tests.
 
genkifd said:
one thing i see right at the beginning of this test is that the card compared is the reference gtx 680 (2gb vram) against the ghz 7970 (6gb vram) when they should be comparing the Classified GTX 680 (4gb vram) instead. i would say the gtx680 will still come out on top but far better comparison
Click to expand...


Your classification is n00bie for good reason.
 
maxius said:
me as well... i am interested in learning more about this personal observation of smoothness is it a scientific observation? or feelings and if their is any nvidia tech involved behind the scenes that make it smoother and exactly what it does.
Click to expand...

It means how big is the framerate variation (this number, "variation" and "standard deviation", you can obtain through a statistic formula in excel menu). You can feel that easily with regular vsync: big steps from 60-30-15 fps wich give the impression of stutter.

But this is no more an issue for a long time: using AMD drivers vsync + frame limiter/cap you have 0 variation.
You can do frame limiter/cap with OSD tools like afterburner/dxtory.

Or do all with everything with one profile (dynamic frame cap + vsync): http://www.radeonpro.info/
 
Last edited:
Smoothness is not a problem anymore for years now. You just need to frame cap + vsync.
Simracers make that a rule(simulators even have the option in the graphic settings menu).
The HOCP graphics are the best for that: showing how much the framrate goes below a margin
 
Another gold star for Nvidia frame metering, I for one can always see the frame time issues on CFX.
 
Semantics said:
Yes in still images it's quite hard to tell the difference but their choice of stock performance optimizations not per game profiles to tailored performance leads to more noticeable in motion characteristic of their textures, ie the shimmer, crawling and noise while in motion.
Click to expand...

IIRC it was due to LOD bias. Anyway, texture quality and optimizations have been a non issue for quite some time now, and certainly in this review and discussion.
 
genkifd said:
one thing i see right at the beginning of this test is that the card compared is the reference gtx 680 (2gb vram) against the ghz 7970 (6gb vram) when they should be comparing the Classified GTX 680 (4gb vram) instead. i would say the gtx680 will still come out on top but far better comparison
Click to expand...

GHz Edition 7970 is 3GB, same as standard 7970
 
Semantics said:
I think you misread something it mentions no where that it's using sapphire's card it's just a generic GTX680 to a generic 7970 Ghz edition, which is 2gb vs 3gb, and in a platform of a very large resolution plus AA which does make a difference. But it's not unfair in the aspect of 2 generic top of the line cards competing in the same tests.
Click to expand...

i thought the ghz edition only came in 6gb vram sorry guys. then its fair
 
Brent_Justice said:
We don't use any third party utilities to change the out-of-box experience between NVIDIA and AMD GPU performance/gameplay experience. We feel it is up to AMD to implement these features to improve its platform, if it so chooses, if it doesn't, well then we are left with what we are left with as the experiences between both. If we use a third party utility for AMD, then we'd have to use one for NVIDIA, and then its just not fair.
Click to expand...

IMO RadeonPro (and NV Inspector) should be included in your reviews.

I understand that you want to focus on the "out-of-the-box" experience because that is what both AMD and NV have to offer themselves. While this seems correct in terms of "corporate fairness" it somehow leads to a bit unhandy conclusions.

I am pretty sure that most of the high-end pc-gamers with multi-gpu setups which tend to overclock, undervolt or mod their hardware are using such third-party tools for their every-day use.

That leaves this "out-of-the-box" experience as the only focus at least a little incomplete and could very well be extended by including tools like RadeonPro.

In the past you have focused well on that useful, practical side of pc-gaming in your reviews but excluding tools like RadeonPro from the evaluations seems not reflecting the every-day gaming-use of multi-gpu-systems nowadays.

To my mind both AMD and NV should officially support great tools like RadeonPro an NV Inspector. Modern multi-gpu gaming has obviously reached a point of complexity in which those companies are dependent on the community for extending the gaming-experience.
 
H.A.M.mer said:
IMO RadeonPro (and NV Inspector) should be included in your reviews.
Click to expand...

This, and also a max OC to max OC comparison would also be interesting.

To be honest I think only a small proportion of the people who buy GTX 680's and HD 7970's care about out of the box experience, after all if you spend $800+ on a GPU setup then surely you would download a small piece of software to get the most out of it.

Great review though, looks like AMD have got their act together and the CF scaling on some of the games is remarkable.
 
Amazing review, as always, [H]ard|Lords. Thank you for taking the time needed for this.
 
So I went back and had a look at the graphs and I think I see the problem. Especially in Sleeping Dogs, the AMD graph is all over the place. Wild swings up and down where the Nvidia graph is relatively smooth.

BF3 and Max Payne look to be about the same with the 2 cards but BF3 multiplayer has AMD showing these wild swings again.

I would think wild swings in frame rates like the ones shown in those graphs would be noticeable so I guess Brent and the boys were on to something after all. Does make me wonder how many games AMD cards would do this on. Since the AMD cards don't have a problem with BF3 but do on BF3 MP, maybe its just in games that are really CPU intensive (I'm assuming Sleeping Dogs is).
 
Sharps97 said:
I might suggest using a similar methodology to other sites - frame latency. Basically measure the average amount of time it takes to render frames and then call out the number of frames outside the standard deviation and compare that way. This method would more clearly show cards that render inconsistently and therefore would "feel" less smooth even at lower average frame rates. Basically, it would just put the science behind your observation while also pushing forward the [H] leadership in benchmarking methodology.
Click to expand...

This comes up with every SLI/CrossfireX review that we do. In short, Kyle and Brent have been working with NVIDIA and AMD for a while to get the right (hardware) tools in order to properly "put science behind" our observations. The "other" sites are measuring frame times at the box and NOT the observed frames at the monitor.
 
"Overall, it's hard to call one a winner..." LOL you just can't help it :D

So if "framerates are not everything" why not save bandwidth not showing the Max FPS in you graphs ?

So if "framerates are not everything" why not test always with Vsync ON? ...maybe Crossfire would the "smoother" this way........
 
Well I would be interested in the GTX680 4gb. I have this card and do all my gaming at 5760x1200 Batman AC and BF3, Guild Wars2. I feel these games are very playable at higher settings.I have been contemplating another card for this setup. Just wondering if the extra gig helps. Very interesting read.
 
David_Schroth said:
This comes up with every SLI/CrossfireX review that we do. In short, Kyle and Brent have been working with NVIDIA and AMD for a while to get the right (hardware) tools in order to properly "put science behind" our observations. The "other" sites are measuring frame times at the box and NOT the observed frames at the monitor.
Click to expand...

It's still a step in the right direction. Throwing your hands up in the air and saying "oh well we can't measure this perfectly so we won't try" ALWAYS sounds like a cop-out to me.
 
Actually, changing graphs based on monitor refresh rates might not be a bad idea...

Have 2 graphs; 60 & 120, and see which cards give more consistent, and therefore smoother, results based on these caps. The lows become especially important for smoothness results.
 
"Out-of-the-box" or no, this is an extreme overclocking hardware enthusiast site, isn't that right? How is it remotely [H]ard to exclude a freely-available tool which addresses the exact problem you are lamenting? By that line of reasoning we shouldn't overclock these cards at all, in the interest of a fair comparison..
 
My question relates a little more to the subjective experience of Nvida VS AMD. For desktop and single monitor gaming (IE Starcraft2, Dota2, etc) was it easier to do this on Nvidia or AMD? I currently have a 5850 and the process of switching back and forth is incredibly time consuming. Is the Nvidia solution any better?

EDIT: Ideally I wouldn't have to switch profiles in order to play single monitor games. I was under the impression that was possible with NVSurround.
 
You must log in or register to reply here.
Back
Top