980 vs 970 RESULTS [update]

[redzo]

I have known this, 970 has never been on my upgrade cards.
GTX 670 could have easily been OC'ed to 680 performance level in games. Nvidia got smarter this time. Intel like.

There is also a bigger price gap this time. gtx980 turns performance/$ into ashes, less bang for your $ if go the fastest single gpu around(980) vs gtx670/680 days.

 

[nitromullet]

a 980 is not 15% faster than a 970. The correct representation is, 980 is 15fps better than a 970. This is especially true when there are scenes that's really pushing the cards. When 970 is rendering at an unplayble 15fps. The 980 can render at 30fps.

You own data simply doesn't back up this claim.

During a particular graphical demanding scene in metro redux
970: 16~19fps oc 1500
980: 22~24fps oc 1480

During a particular graphical demanding scene in battlefield 4 with resolutational scale tune to 200%

970: 11~15ps oc 1500
980: 16~18fps oc 1480

 

[BonzaiDuck]

Like I said, Tom's Hardware had reviewed a spate or earlier-gen cards more than a year ago -- maybe ten or more.

This review of September 2014 just compares a GTX 970 and GTX 980 with a 770 and some Radeons:

http://www.tomshardware.com/reviews/nvidia-geforce-gtx-980-970-maxwell,3941.html

ADDENDUM: Note the comparison differences in a chart for 1920x1080 versus what would seem is a "4K" resolution. So my personal "consumer perspective:" I've got a (now old from 2009) Hanns-G 28" HD monitor with a single dead pixel. I really cannot fret about a single dead pixel. When -- anytime soon -- will I go goo-gah and spendthrift to buy a 4K monitor? Not likely.

So the question remains: How does 2x GTX-970 versus 1x GTX-980 compare? For 4K, also?

 

[Lepton87]

As much as this thread sounds like a troll thread, 15fps better? Really? There's indeed something wrong with the 970.
http://forums.anandtech.com/showthread.php?t=2416150
http://www.techpowerup.com/209205/g...caps-video-memory-usage-to-3-3-gb-report.html*

Or there might be, I jumped to the conclusions too early. the 970 indeed seems to choke earlier than the 980.But why?

ps. I'm glad that I didn't "upgrade" my Titans, 6GB to 3.3GB would be a very significant downgrade.... And with watercooling heat isn't an issue, 37C at most fully loaded overvolted and overclocked, stock cooling couldn't handle that even with the fan at 100%.

 

[BonzaiDuck]

As much as this thread sounds like a troll thread, 15fps better? Really? There's indeed something wrong with the 970.
http://forums.anandtech.com/showthread.php?t=2416150
http://www.techpowerup.com/209205/g...caps-video-memory-usage-to-3-3-gb-report.html

ps. I'm glad that I didn't "upgrade" my Titans, 6GB to 3.3GB would be a very significant downgrade.... And with watercooling heat isn't an issue, 37C at most fully loaded overvolted and overclocked, stock cooling couldn't handle that even with the fan at 100%.

Yup. Most certainly. But it still depends on "starting point:" Upgrading from what to what, and for what by whom?

But I'd have about four days short of a month to go on an RMA window. The TechReport article was only posted two days ago!! What do they mean by "have an update as soon as possible"? Firmware? Driver? Or the whole hardware enchilada?

 

[ShintaiDK]

As much as this thread sounds like a troll thread, 15fps better? Really? There's indeed something wrong with the 970.
http://forums.anandtech.com/showthread.php?t=2416150
http://www.techpowerup.com/209205/g...caps-video-memory-usage-to-3-3-gb-report.html

ps. I'm glad that I didn't "upgrade" my Titans, 6GB to 3.3GB would be a very significant downgrade.... And with watercooling heat isn't an issue, 37C at most fully loaded overvolted and overclocked, stock cooling couldn't handle that even with the fan at 100%.

Is your links supposed to have any relevance to a 15fps drop?

Also we got multiple members with GTX970 here already tested above 3.5GB without any issues.

And unless you wanted to free up money. I dont see a reason for your sidegrade.

 

[Lepton87]

Is your links supposed to have any relevance to a 15fps drop?

Also we got multiple members with GTX970 here already tested above 3.5GB without any issues.

And unless you wanted to free up money. I dont see a reason for your sidegrade.

The 970 should not have any bigger drops in performance than the 980 and if it does I'm investigating why. Not being able to address the full memory space could be a reason, I'm not saying it is.

 

[hunkeelin]

First, the whole darn thread is interesting. But Hunkeelin's discussion of binning processors especially caught my attention. Offhand, I can't think of a single reason why it is not true. Hold that thought.

On the other side of the same equation is "product differentiation" -- a classic example being the automobile model-lines and even "brands" among GM subsidiaries. Product differentiation is a deliberate attempt to capture extra profits -- similar to "consumer surplus" seen by a monopolist. You sell "more stuff" at multiple price-points addressing various consumer market-segments and therefore you earn more profit with a greater aggregate customer-base than at a single price-point. "Cadillac plus Chevrolet" is better than either "Cadillac" or "Chevrolet" separately.

So I'm guessing this may actually reconcile with the observation about binning defective product. Intel had done this with the Pentium 2, producing a "260," a "300," etc. They rolled all the processors off the assembly-line, then deliberately disabled a pre-calculated feature making this or that production-unit a lesser or greater processor. This then led to someone in the Philippines who was able to re-enable the feature, and "Voila" -- you had a form of CPU-counterfeiting.

They could actually do it both ways: test the product for certain types of defects, and disable enough more of them to meet marketing department assessment of profitability among us "great-unwashed GTX 970" consumers.

BY THE WAY. Did anyone mention a performance comparison between 2x GTX 970 in SLI against a single 980? Perhaps even with judiciously chosen overclocks between the two configurations?

That would actually be an important topic of interest, since 2x GTX 970 is just a tad more than 1x 980 in price. So it actually concerns a "bang-for-buck" ratio of value in making those consumer choices.

970 sli is 10~13fps over 980 depending on the particular scene
980 sli is 10~15 fps over 970 depending on the particular scene

There you go

 

[BonzaiDuck]

Is your links supposed to have any relevance to a 15fps drop?

Also we got multiple members with GTX970 here already tested above 3.5GB without any issues.

And unless you wanted to free up money. I dont see a reason for your sidegrade.

Incredible, the number of hits, posts and responses in just a couple days on that thread.

Umm . . umm . . UM-mmmm!! The plot thickens. Some of those posts from forum veterans I know well enough for their tech-knowledge seem to imply -- if anything -- and if there is a "flaw" -- it is correctible through either firmware or drivers.

What's your take on that, ShintaiDK?

 

[hunkeelin]

You own data simply doesn't back up this claim.

It depends on the scene. On that particular scene that's the result. I am by no means bias to 980. However, you are right, most of the time during my test the 980 is around 10~12fps over 970. Not a consistent 15fps. I apologize.

 

[hunkeelin]

The underlying thing is. When you game with a 980 it's "smoother". While 970 choke at some scenes. A 980 will give you a smoother gaming experience.

 

[nitromullet]

It depends on the scene. On that particular scene that's the result. I am by no means bias to 980. However, you are right, most of the time during my test the 980 is around 10~12fps over 970. Not a consistent 15fps. I apologize.

The maximum single card frame rate delta between the 970 and 980 from your own data is 6 fps.

During a particular graphical demanding scene in metro redux
970: 16~19fps oc 1500
980: 22~24fps oc 1480

During a particular graphical demanding scene in battlefield 4 with resolutational scale tune to 200%

970: 11~15ps oc 1500
980: 16~18fps oc 1480

The interpretation of your own data is simply incorrect.

 

[Carfax83]

The interpretation of your own data is simply incorrect.

Not only that, but the test itself was contrived. 16x MSAA is utterly useless for modern titles, and overtly harmful to performance.

The 980 is faster yes, we all know that. But the delta between it and the 970 is smaller than the delta between the 780 and the 770, because they have the same amount of ROPS and the same memory bus.

Only at reference clock speeds is the performance gap between the two substantial, and only because NVidia wanted to make the 980 their undisputed halo product.

 

[RussianSensation]

Conclusion:
The message I want to deliver is that reviews in google are a mis representation. A percentage performance scale is not how you judge a gpu. a 980 is not 15% faster than a 970. The correct representation is, 980 is 15fps better than a 970. This is especially true when there are scenes that's really pushing the cards. When 970 is rendering at an unplayble 15fps. The 980 can render at 30fps. That's a difference.

You didn't provide enough data to back up your claims that a single 980 can render something at 30 fps when in the same scene a 970 drops to just 15 fps. Even if this is true in 1 game, you need to provide hard data for 10-20 games before we take such a claim seriously. I find it hard to believe that a GTX980 SLI could maintain 30 fps when GTX970 SLI would only be 15 fps in the same scene.

I just want to let people know 980's price is justified. Spilling out nonsense like 980 is only 15% better for 50~60% of the price is not right.

It is right because that's the actual performance difference on average. The performance difference in SLI is less because SLI doesn't scale 100%.

@ 1440P
GTX970 SLI provides 90% of the performance of 980 SLI for 40% less cost.
Alternatively, that's the same as stating that GTX980 SLI is only 11% faster than GTX970 SLI for 67% more money ($1100 vs. $660).

If 980 SLI is worth it for you over 970 SLI, then that's perfectly fine because we don't know your income level, and how many hours you play game a day. Since you have a 5960X ($1000 CPU), 32GB of DDR4 3000mhz ($600-900) and Shure SE 535 ($450-500 headphones), you don't come off as someone who is strapped on money to not blow spend $400 extra on 980 SLI. For the rest of us, it's a bad choice. It doesn't change hard data that 970 SLI is the price/performance choice and that 980 SLI will hardly provide a much better overall gaming experience with just an 11% advantage on average.

Using some random test with 16xMSAA isn't proof of anything. No one here would play Metro at 16xAA, so it's a pointless test. At least use DSR or super-sampling.

Also, I have to note, during scene's where it's really pushing the cards. The average fps between oc(1500mhz) and stock(1316mhz) is minimal. 1~2fps increase at best. For both 970 and 980.

This contradicts every review out there. You can overclock a GTX970 to match or pass a reference 980 in performance, but a GTX980 OC will still beat a GTX970 OC by the same 15-20% range in games that allow for the 980's shader and texture performance to shine. The performance increase on a 970/980 @ 1.5Ghz is A LOT more than 1-2 fps in games over stock.

Your comment that 970 and 980 hardly benefit from 1.5Ghz overclocking doesn't align with professional data. Almost 30 fps gain from a stock 980!

Also, clearly you can overclock a 970 to beat a stock 980, but no one ever made a claim that a 970 OC can match a 980 OC.

So the question remains: How does 2x GTX-970 versus 1x GTX-980 compare? For 4K, also?

It would also be a question if no one tested this. I am not sure if you haven't read any reviews since Sept 26th.

GTX970 SLI = $660
R9 295X2 = $660
GTX980 SLI = $1100 for 8-14% more performance at 4K over the other 2 setups (comparing stock vs. stock not OC vs. OC) Sure, you can overclock 980 more than an R9 295X2 in % terms but the advantage over the 970 SLI will remain nearly fixed.

Think about this, someone spends $660-700 on R9 295X2/970 SLI. Well when GM200 or Pascal drops, they'll get way more performance and use that $ saved while getting a setup 40-50% faster than a 980 SLI. If you can easily afford the best of the best every new gen, go for it. Otherwise, 980 SLI is simply a waste of $ vs. a 295X2/970 SLI.

If you look hard enough, you can find R9 295X2 beating 970 SLI at 4K by 38%.

Alternatively, you can find rare cases where 980 SLI beats 970 SLI by an unusually large amount.

So I guess then if all you do is play those games 24/7, then sure maybe there will be a 20% difference between 980 SLI and 970 SLI in some odd title.

 

[BonzaiDuck]

. . . .
It would also be a question if no one tested this. I am not sure if you haven't read any reviews since Sept 26th.

GTX970 SLI = $660
R9 295X2 = $660
GTX980 SLI = $1100 for 8-14% more performance at 4K over the other 2 setups (comparing stock vs. stock not OC vs. OC) Sure, you can overclock 980 more than an R9 295X2 in % terms but the advantage over the 970 SLI will remain nearly fixed.

Think about this, someone spends $660-700 on R9 295X2/970 SLI. Well when GM200 or Pascal drops, they'll get way more performance and use that $ saved while getting a setup 40-50% faster than a 980 SLI. If you can easily afford the best of the best every new gen, go for it. Otherwise, 980 SLI is simply a waste of $ vs. a 295X2/970 SLI.

If you look hard enough, you can find R9 295X2 beating 970 SLI at 4K by 38%.

Alternatively, you can find rare cases where 980 SLI beats 970 SLI by an unusually large amount.

So I guess then if all you do is play those games 24/7, then sure maybe there will be a 20% difference between 980 SLI and 970 SLI in some odd title.

How much one "wants to spend," and how much "one CAN spend" are related questions but still different. To me, it would seem more cost-effective to buy the two 970's at that $600-$700 price range -- better than a single 980. And as you say, you don't really gain diddly-squat for the 980 SLI option. You only spent another $600 bucks.

Oh. Also. truth is, I might have read more reviews on the issue -- on the 9x0 NVidia models. I thought I'd read "enough" though, for not being sure of wanting SLI, for looking at both 970 and 980, and for just "looking for a particular make and model."

No loss to me. I'm not going for 4K and two or three monitors. It's like the old Stones song: "Ya can't . . . always get . . . what ya WA-ant! . . . . But ya get whatcha nee-eed!" If you only want what you need, it should wag your Willie well enough . . .

 

[RussianSensation]

How much one "wants to spend," and how much "one CAN spend" are related questions but still different. To me, it would seem more cost-effective to buy the two 970's at that $600-$700 price range -- better than a single 980. And as you say, you don't really gain diddly-squat for the 980 SLI option. You only spent another $600 bucks.

That has been my recommendation since 970 came out, assuming sufficient PSU and chassis. There are certain members for whom the 980 made more sense. If you have a miniITX like the Raven RVZ01/upcoming RVZ02, it's just not possible to have a 970 SLI setup HTPC in your living room in such a case. I think NV would sell quite a bit of GTX990 dual-chip cards for that reason. But if GM200 is close enough, maybe that's why they skipped such a card entirely.

 

[BonzaiDuck]

That has been my recommendation since 970 came out, assuming sufficient PSU and chassis. There are certain members for whom the 980 made more sense. If you have a miniITX like the Raven RVZ01/upcoming RVZ02, it's just not possible to have a 970 SLI setup HTPC in your living room in such a case. I think NV would sell quite a bit of GTX990 dual-chip cards for that reason. But if GM200 is close enough, maybe that's why they skipped such a card entirely.

Yo! Is your sig as updated as it can be? The way it is, I feel "less alone." But a lot of people are still rocking Sandy systems. Just wondered.

I'll have to decide whether to order another 970. Again -- want versus need, and the possibility that I'll actually be overpowered for any "need" I have.

So far, it just seems like a great card, despite being "second tier." With a slight adjustment with AfterBurner to the fan profile, it doesn't seem to go above the mid-50's C with gaming that pushes GPU usage to 90% or a bit more. The default setting actually powers down the fans at a lower temperature threshold, so I found the card in a sort of "gray area" so ve-ry slow-ly dropping the temperature but still in the 50s some minutes after gaming stopped.

For power, the AfterBurner monitor shows it never climbing above 80W. I've yet to overclock it, but that, too, will happen.

 

[RussianSensation]

Yo! Is your sig as updated as it can be? The way it is, I feel "less alone." But a lot of people are still rocking Sandy systems. Just wondered.

It is for now. I managed to snag a new Asrock Extreme4 X99 and 16GB G.Skill DDR4-2666 for $338, but have no access to a MicroCenter until I go back to the US in March. I am not sure if I'll flip these parts for some profits and wait for Skylake-K/BW-E (or maybe flip them and buy the same parts at MicroCenter again which would make the upgrade to an X99+5820K platform even less expensive). Sounds more e-peen than anything for games over the 4790K but I haven't played with new CPU parts in 4 years and buying yet another 4-core CPU seems boring since my Q6600 from August 2007! :'(

I would like an X99-A board with USB 3.1 but I don't think they are out yet, and not in my price range:
http://www.tomshardware.com/reviews/usb-3.1-performance-benchmark,4037-2.html

I'll have to decide whether to order another 970. Again -- want versus need, and the possibility that I'll actually be overpowered for any "need" I have.

Ya, that's a good point. If a single 970 is good enough, you can pick up a discounted one for $200 in 12-15 months and save $130 instead of paying extra now to futureproof for 12-15 months from now. I am thinking of dumping my 7970s before they completely tank in value. With 960 and 285 2GB going for $200 and bombing in modern games that use > 2GB of VRAM, there is still a chance to sell 7970s for something.

Not sure how 380X/GM200 will stack up against $500 R9 290s. If 380X/GM200 are $600 and are nearly as fast as R9 290s, it might be worth getting without the hassle of CF. However, I have a feeling there will be some big clearance sales on R9 290/290X cards "no one wants". It'll be hard to pass up on stop-gap R9 290Xs at $400 if an R9 380X/GM200 are $650 or more. But then if an $400 R9 380 is just 15% within a 390, I'll be tempted to get 2 of those and well that's $800. Can't win with GPUs, can you

Really hoping for some clarity and market launch on R9 300 series so that price wars on 970/980 can begin. :awe:

 

[BonzaiDuck]

. . .

Ya, that's a good point. If a single 970 is good enough, you can pick up a discounted one for $200 in 12-15 months and save $130 instead of paying extra now to futureproof for 12-15 months from now. I am thinking of dumping my 7970s before they completely tank in value. With 960 and 285 2GB going for $200 and bombing in modern games that use > 2GB of VRAM, there is still a chance to sell 7970s for something.

Not sure how 380X/GM200 will stack up against $500 R9 290s. If 380X/GM200 are $600 and are nearly as fast as R9 290s, it might be worth getting without the hassle of CF. However, I have a feeling there will be some big clearance sales on R9 290/290X cards "no one wants". It'll be hard to pass up on stop-gap R9 290Xs at $400 if an R9 380X/GM200 are $650 or more. But then if an $400 R9 380 is just 15% within a 390, I'll be tempted to get 2 of those and well that's $800. Can't win with GPUs, can you

Really hoping for some clarity and market launch on R9 300 series so that price wars on 970/980 can begin. :awe:

You're spot-on about that, but it becomes problematic if the resellers suddenly show "out-of-stock" and you have to look elsewhere. I picked up a 780 GTX ASUS "Direct-CU" card exactly a year ago, and that happened. I'm glad for it, though, because these new cards seem like a better bet. Of course, what comes around if I wait long enough -- it could repeat the same cycle in my personal decisions.

Graphics cards can eat your wallet if you let them!

I'd used ATI Radeon cards way back when. I could wonder if the great "Intel vs NVidia" conflict of some six years ago might have influenced me, but it didn't. Chipsets and vid cards are two different things. I just haven't had any incentive to "cross over" to "Cross Fire.".

 

[5150Joker]

If I wasn't coming from 2 x Titans, I'd have picked up 2 x 970s. The performance difference is obviously going to be there, especially OC vs OC but not $550 worth of difference. If you are building a brand new system and just feel like splurging on top end stuff then go for it but if you're on a budget, 970 is the way to go. Next upgrade I'll probably go for the 2nd tier GM200 consumer cards - you get 90% of the performance of the top card at nearly half the price if current trends keep up.

 

[flexy]

I am sorry, your test was a waste of time...it doesn't tell us anything.

We KNOW that a GTX 980 is faster than a 970.

"Particular demanding scene" doesn't mean anything. The current hot topic is performance decrease of 970 when it uses more than 3.2G. You didn't even specify what videomemory in your scene is allocated.

Your statement "15fps better" is better than "20%" I pure nonsense. Sorry, it is. If I'd see a review using terms like "XYZ FPS better" I would stop reading right away.

You say it counts to measure FPS in a particular demanding scene and you give an example where the cards barely get to 16 or 20 FPS.

Except: NO ONE plays a game at settings where they only get 16 FPS.

In the same way as it is not relevant to look at benchmarks where one benches and compares cards, say, running at 4K resolution and ultra-settings etc.. OF COURSE the "better" card will have an advantage there...but it's meaningless if you do NOT play in those high resolutions. Why should I care that a card can only get 20FPS in 4k resolution when I play at 1920 where the same card may well be beyond 100FPS.

A "smaller" card can be almost equal in terms of performance at "normal" settings where the "better" card only THEN shows an edge in extreme situations, like the test you did which is so demanding it brings any card to its knees.

Or differently: You could play a game at extreme settings and extreme resolutions and AA where a card with 4GB (or 6, or 8) GB has a big advantage, say, it's 50% faster than a card with only 2GB. However it can be that the two cards at a normal resolution like 1920 are doing equally fine since in the lower resolution it doesn't need 4GB memory so the card with only 2GB wouldn't lose out due to lack of video memory. In this lower resolution the performance could be about the same. You can therefore not make a general statement about the card's performance like you do based on an extreme condition.

 

[hunkeelin]

I am sorry, your test was a waste of time...it doesn't tell us anything.

We KNOW that a GTX 980 is faster than a 970.

"Particular demanding scene" doesn't mean anything. The current hot topic is performance decrease of 970 when it uses more than 3.2G. You didn't even specify what videomemory in your scene is allocated.

Your statement "15fps better" is better than "20%" I pure nonsense. Sorry, it is. If I'd see a review using terms like "XYZ FPS better" I would stop reading right away.

You say it counts to measure FPS in a particular demanding scene and you give an example where the cards barely get to 16 or 20 FPS.

Except: NO ONE plays a game at settings where they only get 16 FPS.

In the same way as it is not relevant to look at benchmarks where one benches and compares cards, say, running at 4K resolution and ultra-settings etc.. OF COURSE the "better" card will have an advantage there...but it's meaningless if you do NOT play in those high resolutions. Why should I care that a card can only get 20FPS in 4k resolution when I play at 1920 where the same card may well be beyond 100FPS.

A "smaller" card can be almost equal in terms of performance at "normal" settings where the "better" card only THEN shows an edge in extreme situations, like the test you did which is so demanding it brings any card to its knees.

Or differently: You could play a game at extreme settings and extreme resolutions and AA where a card with 4GB (or 6, or 8) GB has a big advantage, say, it's 50% faster than a card with only 2GB. However it can be that the two cards at a normal resolution like 1920 are doing equally fine since in the lower resolution it doesn't need 4GB memory so the card with only 2GB wouldn't lose out due to lack of video memory. In this lower resolution the performance could be about the same. You can therefore not make a general statement about the card's performance like you do based on an extreme condition.

Battle field, resolutional scale up to 200%. On the third mission. The raining scene where you are facing the huge sky scraper. 970sli is at 25fps while 980 is at 40. I use fraps and I never see fps drop below 40. I dont' think using percentage to represent the strength of gpu is correct. I may not be very thorough with my review and take screen shots etc... However, the 980 is general smoother and during scenes that matters for gamer, the 980 provide a better gaming experience. Also for metro, x16/x4 doesn't really matter. The fps doesn't change that much, what matters in metro is when you turn up the ssaa up to x4.

I get pissed off during gaming when I experience lag. The 970sli do give me those bad experience. However, 980sli never gave me those headache.

I am not here to convince anyone to buy the 980, but I assume whoever buys two 970 in sli; they are hoping to play games smoothly while turning everything to the max. The problem is, the 970 sli do not provide that experience. The 980sli do. (All test are done in 1080p, i'm not even going into the realm of 1440p and 4k yet; The current gen gpu cannot handle 1440p and 4k at max settings)

 

[kasakka]

Battle field, resolutional scale up to 200%. On the third mission. The raining scene where you are facing the huge sky scraper. 970sli is at 25fps while 980 is at 40.

I get pissed off during gaming when I experience lag. The 970sli do give me those bad experience. However, 980sli never gave me those headache.

I am not here to convince anyone to buy the 980, but I assume whoever buys two 970 in sli; they are hoping to play games smoothly while turning everything to the max. The problem is, the 970 sli do not provide that experience. The 980sli do. (All test are done in 1080p, i'm not even going into the realm of 1440p and 4k yet; The current gen gpu cannot handle 1440p and 4k at max settings)

Isn't it more that you run with unrealistic settings? Downscaled 4K resolutions and on top of that high level of antialiasing is surely going to put any setup to its knees but not provide much of an image quality difference over something more reasonable. This whole "maxing everything" is just plain stupid as nowadays most games don't gain much from turning everything to 11. Just lower FPS with high VRAM use (from uncompressed textures and high AA) with image quality improvements only noticeable in screenshots.

 

[hunkeelin]

Isn't it more that you run with unrealistic settings? Downscaled 4K resolutions and on top of that high level of antialiasing is surely going to put any setup to its knees but not provide much of an image quality difference over something more reasonable. This whole "maxing everything" is just plain stupid as nowadays most games don't gain much from turning everything to 11. Just lower FPS with high VRAM use (from uncompressed textures and high AA) with image quality improvements only noticeable in screenshots.

You are partially correct. However, there's a difference in b4 100% resolution scale vs 150%. I can tell a difference between 150% resolution's scale vs 200%. A lot of people might say it's not worth though.

For metro; there is a noticeable amount of difference when ssaa is off compare to ssaa is 2. There's also a slightly noticeable difference between ssaa x2 vs ssaa x4 if you really look into detail. Especially when you are in tunnels.

For someone who really pay attention to every single detail then I think the price of 980 is justified. For any hardcore gamer, experiencing lag is extremely frustrating given they spend over 2k in their rig. The price/performance is that bad for 980 which everybody claims. As I said, a 10~13fps increase in performance is not equals to 15% improve in performance. 60(970sli) vs 70(980sli) means nothing. however, 25~27fps(970sli) vs 38~41fps(980sli) means something.

 

[Final8ty]

In these test the GPU grunt became the bottleneck more than anything else so would mask most other issues.