Fall 2012 GPU and Driver Comparison Roundup @ [H]

Great review! I have been really happy with my reference Galaxy 680 since May 2012. So while AMD has caught up (or even passed Nvidia) in many regards, I have been enjoying a great level of performance for 6-7 months. Plus - back then, I don't think the GHz versions were out.

One thing that is important to with the Nvidia drivers is the adaptive v-sync which really works great with my monitor. Even so, if I were in the market today for a video card today - the 7970GHz would be a definite contender.
 
Wish they had included a <$200 card in the review. Also maybe the 7750 or 7770 and GTX 650 near the $100 point as well. It'd be nice to know if the drivers are improving the lower tier cards as well.
 
Ajiki-kun said:
Damn. All the results of how much better the 7950/7970 keeps getting is a knife being driven deeper into my gut. If I didn't have so many driver problems with the AMD cards, I would never have ditched them.

I hope nVidia gets decisively and conclusively destroyed very quickly in sales, and soon, so that nVidia finally drops prices on Kepler cards for the first time. Let me at least SLI my 660Ti ._.
Click to expand...

How do you think I feel with SLI 660ti? I agree though, that the 7970 is just mocking me with these newer driver updates. I never dreamed that they would finally release stable drivers.

I suppose you can't rely on hell freezing over though. Glad they got their act together.
 
saitei said:
How do you think I feel with SLI 660ti? I agree though, that the 7970 is just mocking me with these newer driver updates. I never dreamed that they would finally release stable drivers.

I suppose you can't rely on hell freezing over though. Glad they got their act together.
Click to expand...
Ouch. I'm sorry to hear that man. I hope for both our sakes nVidia at least continues to release performance-improving drivers in the future.

The thing is, I'm hearing the bugs that I ditched 79xx for are still around, so they still haven't even released stable drivers. I couldn't even jump ship if I wanted to. To me, AMD cards and their owners are like the French soldier in Monty Python and the Holy Grail.
 
The pure numbers look good for AMD, but you can't deny the smoothness advantage that Nvidia has in the multi-card arena. You couldn't pay me enough to go back to the chop of Crossfire.
 
Dogboy714 said:
The pure numbers look good for AMD, but you can't deny the smoothness advantage that Nvidia has in the multi-card arena. You couldn't pay me enough to go back to the chop of Crossfire.
Click to expand...

Absolutely. Though for the vast majority multi GPU is not a consideration.
 
I hopped on the 680s at launch and don't regret my decision given SLI vs CFX smoothness, but damn ATI has really turned around the single GPU crown. Impressive work with the drivers.
 
the article is about the comparison of the old and new drivers, but shows no comparison. Its nice to see the numbers of the new cards. But i thought you were going to see if the claimed percentage numbers were acurate. I'll have to look at older reviews and compar numbers to get my own percentages. It would be nice to see the comparisons in the review of how the real world outcomes of these new driver updates effected these cards.

Nice review showing the new numbers of these cards though.
 
raghu78 said:
there is a mistake in page 6. In apples to apples comparison for HD 7970 and GTX 680 the fraps graph is incorrect. the settings are mentioned as 1920 x 1080 8x MSAA but the fraps graph of 2560 x 1600 2x MSAA is shown.

http://hardocp.com/article/2012/11/12/fall_2012_gpu_driver_comparison_roundup/6

please correct this

Just the thumbnail was borked, it is fixed now, thanks - Brent
Click to expand...

raghu78 said:
in page 7 the apples to apples comparison is confusing. are all the cards using the same settings. if so then the fraps graph is showing ubersampling for the hd 7970 ghz, gtx 680, gtx 670 and 4x MSAA for hd 7950 boost, GTX 660 Ti, HD 7870.

This has been fixed, Thanks - Brent
Click to expand...

This tends to happen when I trade sleep for doing video card reviews. Had a tight turn around on this one and the real job has been keeping me rather busy. Many of the graphs were made above 35,000ft this time around...

raghu78 said:
sleeping dogs comparison is just baffling. GTX 680 is shown to run at 2560 x 1600 High AA at 30.3 fps in the sleeping dogs game performance review done earlier. in this review GTX 680 runs at 2560 x 1600 High AA at 51.5 fps. the changes between the 2 reviews are the game version and drivers. but how can that amount to 70% performance increase.

http://www.hardocp.com/article/2012/11/12/fall_2012_gpu_driver_comparison_roundup/5

http://www.hardocp.com/article/2012/10/02/sleeping_dogs_gameplay_performance_iq_review/5
Click to expand...

I'll need to double check this one - it does seem a bit high (and I was being careful doing reality checks back to Brent's IQ for Sleeping Dogs). I won't be home until next week though...

wabbitseason said:
Thanks for doing this. That being said, the following quote from the conclusion of the article is possibly the most grudging, evasive, heavily-qualified acknowledgement of the 7970's superiority:

"For those that must have the absolute highest frame rates (even if it makes no difference to the overall game play experience), the Radeon HD 7970 seems to pull ahead of the GTX 680."


The mid-range conclusion, however, tells it like it is. Bravo.
Click to expand...

I'd say that what I said about the 7970 is rather accurate. If you took a "Pepsi challenge" at the same graphics settings on a 30" screen (and no FRAPS counter), you wouldn't be able to tell me whether there's a 7970 or 680 under the hood. That is the experience that I had with the two cards, and given that they're within a couple of Hamiltons of each other in price, its really a push in my mind. That comment was more of a snipe at gamers that insist on frame rate being king and NOT intended to play down the card's performance.

The conclusion would have been far more favorable to AMD on the high end if the comparison was to a standard clocked 7970 or a factory overclocked card, which are all pricing much lower than the cheapest GHz edition cards (most of which being capable of hitting GHz edition speeds - pricing has been linked a few times earlier in this thread).

baldrik said:
Thanks David!
Click to expand...

You're welcome.

Warrior said:
the article is about the comparison of the old and new drivers, but shows no comparison. Its nice to see the numbers of the new cards. But i thought you were going to see if the claimed percentage numbers were acurate. I'll have to look at older reviews and compar numbers to get my own percentages. It would be nice to see the comparisons in the review of how the real world outcomes of these new driver updates effected these cards.

Nice review showing the new numbers of these cards though.
Click to expand...

It would have been great to have had the time to do a before and after comparison, but the amount of time that went into this evaluation was getting up there given the number of cards included. Claimed percentages can be great, but if you get a 10% gain on a 40FPS average, you're probably not going to be able to increase any graphics settings anyway.....

Ninja edit - incase you're curious, about recorded 80 game run throughs went into this review, ranging from 5-10 minutes each (most are close to 10 minutes other than BF3 Multiplayer), plus additional time to play at various settings to hone in on the right one when it is close.
 
Last edited:
raghu78 said:
sleeping dogs comparison is just baffling. GTX 680 is shown to run at 2560 x 1600 High AA at 30.3 fps in the sleeping dogs game performance review done earlier. in this review GTX 680 runs at 2560 x 1600 High AA at 51.5 fps. the changes between the 2 reviews are the game version and drivers. but how can that amount to 70% performance increase.

http://www.hardocp.com/article/2012/11/12/fall_2012_gpu_driver_comparison_roundup/5

http://www.hardocp.com/article/2012/10/02/sleeping_dogs_gameplay_performance_iq_review/5
Click to expand...

As you mentioned, the game version and driver version are both very different between the two. The game has had 3 or 4 patches itself between our gamplay eval and today's article, plus the driver is a lot newer as well from 306.23 to 310.33 is a very big driver difference in the world of drivers, it can all add up to the difference you see.
 
Just wanted to say thanks for a great article.
I'm going to be in the market shortly, and this was *exactly* the kind of write-up I was looking for.
Much appreciated.
 
Thanks for the article guys, really helpful comparison. It sounds like you're all very busy and it's great you took the time to do a head-to-head comparison of the updates that are out right now.
 
AuxNuke said:
Great review, thanks for your time and effort.

Any chance for some X-fire vs SLI results? I've never had an issue with the single card performance of the HD 7970 or GTX 680 however, the multi-gpu performance of the 7970 was never a "smooth" experience. While the frame rates would normally dictate a great experience, X-fire on the 79XX series always seemed choppy, skipping, or stuttering. The method of testing video cards here at the [H] is perfect for this test as it takes into consideration the playable experience of the game and not just frame rates.
Click to expand...

There's a software called Radeon Pro, which completely eliminates the stuttering.
 
how about GTX 670 vs HD 7950 B ?whos win?
from what i can see, they are pretty similiar??

and can u add more benchmark game & benchmark synthetic pls.
it's only 4 game lol ...but nice article.

nvidia just release new beta driver 310.54 lol ...will u re-benchmark again?
 
zipper71 said:
how about GTX 670 vs HD 7950 B ?whos win?
from what i can see, they are pretty similiar??

and can u add more benchmark game & benchmark synthetic pls.
it's only 4 game lol ...but nice article.

nvidia just release new beta driver 310.54 lol ...will u re-benchmark again?
Click to expand...

That driver is for Black Ops II and AC3 performance. We will use it in our look at both games, and all reviews going forward.
 
Rizen said:
I hopped on the 680s at launch and don't regret my decision given SLI vs CFX smoothness, but damn ATI has really turned around the single GPU crown. Impressive work with the drivers.
Click to expand...

I have to say Im happy I went green this round. With the CF issues and the reality that it took nearly 10 months for decent drivers on the reds, its a shame that the cards hit their driver potential near the end of the cycle.
Back when I purchased my cards the 7970 could hardly trade blows with the gtx 670, and now its up there with the 680.
 
N3M3515 said:
There's a software called Radeon Pro, which completely eliminates the stuttering.
Click to expand...

Looks like if you and Brent get bored (fat chance of that) and do a Multi card review you can make it even rediculously longer adding in some runs with RadeonPro thown in on some of those games where the graph says it's playable but your eyes say differently.
 
Single-Card Eyefinity/NV Surround and SLI/CrossFireX Eyefinity/NV Surround card/driver roundup has been scheduled :)
 
When I read the header on the front page, I was expecting something completely different from what the article actually covered! I was expecting to see benchmarks of old drivers vs new drivers.
 
AParsh335i said:
The research I have done says otherwise - the extra memory available on the HD7970/HD7950 and bigger memory interfaces tend to help a lot at the larger Surround/Eyefinity resolutions.

A typical Eyefinity today is 5760x1080 or 5760x1200, and the closest comparison to those resolutions in the tests presented in this article would be the 2560x1600. As you can see in the graphs, 2560x1600 is where AMD had it's largest leads.

Just as a simple example - Max Payne 3 shows HD7970 52fps avg and GTX 680 47fps avg. Now look at 1920x1080 and you'll see that Nvidia takes the lead back, HD7970 at 30fps and GTX 680 at 33fps (note: fps dropped overall because MSAA was changed significantly higher). I wish there was more data to present from this specific test, but only a few games were tested so its hard to say my "theory" is 100% true.

Anyone else have info on this? Also I am referring to only 1 card for eyefinity/surround - the original post also mentions multicard, can anyone else tell us from personal experience?
Click to expand...

I am talking about head to head multicard set-ups.
Not single card EyeFinity.
I have a GTX 670 SLi machine sitting right next to a HD 7970 Crossfire machine.

Generally, game for game, the nvidia drivers are superior to the AMD drivers in simplicity, ease of use and competancy across all games......nvidia also tends to release drivers in advance of the release of a game, not weeks after.

I'm just giving my opinion, based only on my personal experience.
I am not disputing the results of this article, which are very complete.
 
magoo said:
I am talking about head to head multicard set-ups.
Not single card EyeFinity.
I have a GTX 670 SLi machine sitting right next to a HD 7970 Crossfire machine.

Generally, game for game, the nvidia drivers are superior to the AMD drivers in simplicity, ease of use and competancy across all games......nvidia also tends to release drivers in advance of the release of a game, not weeks after.

I'm just giving my opinion, based only on my personal experience.
I am not disputing the results of this article, which are very complete.
Click to expand...

depends on the game honestly.. if its got a nvidia branding/splash screen on it, yes they get driver updates a couple weeks in advance otherwise they do a generic update that at least makes the game use SLI then actually release the correct profile a few weeks later. now if only nvidia could actually release their drivers with a change log that isn't insanely misleading forcing you to read through the 30 page PDF file to see what was actually changed vs what was changed 6 drivers before it i'd be happy.
 
Great review... thanks. At this point nvidia needs to review their pricing. Could be a big win for consumers for the holidays.
 
David_Schroth said:
I'd say that what I said about the 7970 is rather accurate. If you took a "Pepsi challenge" at the same graphics settings on a 30" screen (and no FRAPS counter), you wouldn't be able to tell me whether there's a 7970 or 680 under the hood. That is the experience that I had with the two cards, and given that they're within a couple of Hamiltons of each other in price, its really a push in my mind. That comment was more of a snipe at gamers that insist on frame rate being king and NOT intended to play down the card's performance.

The conclusion would have been far more favorable to AMD on the high end if the comparison was to a standard clocked 7970 or a factory overclocked card, which are all pricing much lower than the cheapest GHz edition cards (most of which being capable of hitting GHz edition speeds - pricing has been linked a few times earlier in this thread).
Click to expand...


Understood on the first bolded portion.

The second bolded portion is absolutely true. If you had used a 7970 instead of a GHz edition this would have been a turkey-shoot.

Regardless these comparisons were sorely needed on [H] and I'm glad you took the time to do them.
 
Great review Dave.

Brent_Justice said:
Single-Card Eyefinity/NV Surround and SLI/CrossFireX Eyefinity/NV Surround card/driver roundup has been scheduled :)
Click to expand...

Please make sure you add RadeonPro (linked in the toms hardware review earlier in this thread).
 
Brent_Justice said:
Single-Card Eyefinity/NV Surround and SLI/CrossFireX Eyefinity/NV Surround card/driver roundup has been scheduled :)
wand3r3r said:
Great review Dave.

Please make sure you add RadeonPro (linked in the toms hardware review earlier in this thread).
Click to expand...
Click to expand...

Yes I would also like to see a test where RadeonPro and its dynamic v-sync is used to eliminate the stuttering on the Radeons.

I know its third-party software and not "hardware" based like nvidia's Adaptive v-sync is but behaves the same way and for the most part, completely eliminates stuttering :cool:

I have been using that program on my 7970cf and 5870cf systems and it has been a lot smoother sailing than without it.

Ps. I know it's my second post so please take it easy one me :eek:
 
Osjur said:
Yes I would also like to see a test where RadeonPro and its dynamic v-sync is used to eliminate the stuttering on the Radeons.

I know its third-party software and not "hardware" based like nvidia's Adaptive v-sync is but behaves the same way and for the most part, completely eliminates stuttering :cool:

I have been using that program on my 7970cf and 5870cf systems and it has been a lot smoother sailing than without it.
Click to expand...

+1

Please include RadeonPro in your testing. Sounds like just what AMD has been lacking.
 
Just like to add, AMD release 12.11 Beta 7 drivers, with even more performance increases I guess, with BF3 being 10-15% faster on top of beta 6 drivers?

For your eyefinity/surround review you should use the new ones!
 
DASHlT said:
Just like to add, AMD release 12.11 Beta 7 drivers, with even more performance increases I guess, with BF3 being 10-15% faster on top of beta 6 drivers?

For your eyefinity/surround review you should use the new ones!
Click to expand...

Pretty sure they were the same figures re-listed as it's still the same driver, just different beta/rc
 
Digital Viper-X- said:
Pretty sure they were the same figures re-listed as it's still the same driver, just different beta/rc
Click to expand...

http://support.amd.com/us/kbarticles/Pages/AMDCatalyst1211betadriver.aspx

Look how the webpage puts it.

It's like its added on the original performance they got.

Now I am not saying you aren't right....Just the way it shows there to me. Looks like even more performance.

First part shows bf3 25%, then with beta 7 says another 10-15%?....yea im not 100% sure either

Ok now i feel stupid. It even says "More performace"

So wow 35-40% more performance for BF3 from 12.11 drivers now!!!!!!!!!!!!!!!!! holy crap
 
You must log in or register to reply here.
Back
Top