Hardocp Fall 2012 GPU and Driver Comparison Roundup

[raghu78]

hardocp has a catalyst 12.11 beta vs detonator 310.33 beta comparison

http://www.hardocp.com/article/2012/11/12/fall_2012_gpu_driver_comparison_roundup/3

HD 7970 Ghz is the fastest single GPU without any contention. the GTX 660 Ti is demolished by the HD 7950 Boost . HD 7870 almost matches GTX 660 Ti. Nvidia though must be readying the GK110 based Nvidia GTX 780 to emphatically take back the single GPU crown.

Nvidia releases 310.54 beta. But comparisons are shown against 306.89 whql. so its unclear whats the improvement from 310.33 beta and is it enough to put GTX 680 on competitive footing with HD 7970 Ghz.

http://www.geforce.com/whats-new/articles/nvidia-geforce-310-54-beta-drivers-released

 

[aaksheytalwar]

Thanks for sharing.

 

[Shmee]

Nothing real new, IMO, especially when Ocing is accounted for.

 

[aaksheytalwar]

Important findings:
There are instances where the difference between a 670 and 680 are 15-25% hint: look at min fps. Also, both oc so that diff is always maintained.

Second, the 7970 ghz is easily 20-30%+ faster than a 7950 boost. And 7970s oc higher as well. So rest assured an oced 7970 will be about 15-20% faster than an oced 7950 in times when it really counts, ESP min fps.

 

[Remobz]

Thanks as well.

 

[toyota]

Important findings:
There are instances where the difference between a 670 and 680 are 15-25% hint: look at min fps. Also, both oc so that diff is always maintained.

Second, the 7970 ghz is easily 20-30%+ faster than a 7950 boost. And 7970s oc higher as well. So rest assured an oced 7970 will be about 15-20% faster than an oced 7950 in times when it really counts, ESP min fps.

um the 670 is clocked much lower so the difference is NOT maintained when both the 670 and 680 are oced. there typically less than a 5% overall difference between a 670 and 680 at the same clocks and a 680 has no better chance of overclocking any further than a 670 either.

 

[raghu78]

Important findings:
There are instances where the difference between a 670 and 680 are 15-25% hint: look at min fps. Also, both oc so that diff is always maintained.

Second, the 7970 ghz is easily 20-30%+ faster than a 7950 boost. And 7970s oc higher as well. So rest assured an oced 7970 will be about 15-20% faster than an oced 7950 in times when it really counts, ESP min fps.

the clock difference between HD 7950 Boost and HD 7970 Ghz is more than the clock difference between GTX 680 and GTX 670. the HD 7950 boost is not running at 925 Mhz consistently at stock but closer to 850 Mhz . But if you max out power control to +20% it will run at 925 Mhz consistently and the gap with HD 7970 Ghz will reduce. At the same clocks HD 7950 is 3 - 5% slower than HD 7970. this has been proven by many users on ocn and even these forums.

From these benchmarks its clear that
HD 7970 (1.2 Ghz) = HD 7950(1.25 Ghz) > GTX 680(1.35 Ghz)

 

[chimaxi83]

Good to see AMD in a definitive overall lead :thumbsup:

 

[aaksheytalwar]

The bechmarks marks members have done are avg fps not min fps. Avg fps of the better card is just 5-10% higher. But min fps are 10-20% higher assuming both are oced to their respective max.

 

[Skurge]

Pretty embarrassing for nV here, although I wouldn't be so clear with a wider range of games, but this pretty much shows how overprices Nv's top 3 cards are. Well actually seeing how poorly the 650/640 perform. It's seems only the 650Ti and the 660 aren't terribly priced.

 

[Termie]

The bechmarks marks members have done are avg fps not min fps. Avg fps of the better card is just 5-10% higher. But min fps are 10-20% higher assuming both are oced to their respective max.

Sorry, but first, you're wrong, and second HardOCP's methodology is highly questionable. The 680 only beats the 670 by more than 10% in minimums in two tests (BF3 MP and W2), and even then, the results of the 660Ti, which are either equal to or greater than the 670 in a majority of games, show that the testing was flawed. Furthermore, the average 670 and 680 both pick up about 150Mhz in boost when overclocked, so the score is even there.

Battlefield SP Apples-to-Apples:
GTX680 Min: 43
GTX670 Min: 44
GTX660Ti Min: 36

Battlefield MP Apples-to-Apples:
GTX680 Min: 46
GTX670 Min: 38
GTX660Ti Min: 37

Sleeping Dogs Apples-to-Apples:
GTX680 Min: 25
GTX670 Min: 23
GTX660Ti Min: 21

Max Payne 3 Apples-to-Apples:
GTX680 Min: 16
GTX670 Min: 15
GTX660Ti Min: 22

Witcher 2 Apples-to-Apples:
GTX680 Min: 19
GTX670 Min: 14
GTX660Ti Min: 18


By the way, the most important takeaway I see in this entire test is the revival of the HD7870. I am so amused by everyone talking about how everyone with a $200 budget should "stretch" to buy the 7950, when even the "boost" model is only about 15% faster than the much cheaper 7870. And of course, the 660Ti should be discontinued.

 

[aaksheytalwar]

Hardocp uses different settings for different cards. Thus the 660ti / 7950 boost seem faster than they really are.

And given the limited number of tests, 2 out of 5 for intensive games is like 40%. I am not saying it would be in 40% games but at least a dozen games among the 100+ releases out there.

 

[Termie]

Hardocp uses different settings for different cards. Thus the 660ti / 7950 boost seem faster than they really are.

And given the limited number of tests, 2 out of 5 for intensive games is like 40%. I am not saying it would be in 40% games but at least a dozen games among the 100+ releases out there.

I quoted apples-to-apples for that reason. The settings are the same, the results are inconsistent at best.

 

[raghu78]

HardOCP's methodology is highly questionable. The 680 only beats the 670 by more than 10% in minimums in two tests (BF3 MP and W2), and even then, the results of the 660Ti, which are either equal to or greater than the 670 in a majority of games, show that the testing was flawed. Furthermore, the average 670 and 680 both pick up about 150Mhz in boost when overclocked, so the score is even there.

Battlefield SP Apples-to-Apples:
GTX680 Min: 43
GTX670 Min: 44
GTX660Ti Min: 36

Battlefield MP Apples-to-Apples:
GTX680 Min: 46
GTX670 Min: 38
GTX660Ti Min: 37

Sleeping Dogs Apples-to-Apples:
GTX680 Min: 25
GTX670 Min: 23
GTX660Ti Min: 21

Max Payne 3 Apples-to-Apples:
GTX680 Min: 16
GTX670 Min: 15
GTX660Ti Min: 22

Witcher 2 Apples-to-Apples:
GTX680 Min: 19
GTX670 Min: 14
GTX660Ti Min: 18

By the way, the most important takeaway I see in this entire test is the revival of the HD7870. I am so amused by everyone talking about how everyone with a $200 budget should "stretch" to buy the 7950, when even the "boost" model is only about 15% faster than the much cheaper 7870. And of course, the 660Ti should be discontinued.

Please read carefully. in Max Payne 3 the apples to apples comparison of HD 7970 Ghz, GTX 680 and GTX 670 are at 1920 x 1080 8x MSAA while the HD 7950 boost, GTX 660 Ti and HD 7870 use 1920 x 1080 4x MSAA. In Witcher 2 the apples to apples settings comparison is not clear. i have posted a question in hardocp forums if the 2 sets of cards are using same settings or different settings in Witcher 2.

also when overclocked HD 7950 gains more from overclocking than HD 7870. the average HD 7950 overclock of 1.15 Ghz would be 25 - 30% faster than HD 7870 (1.25 Ghz). the price difference is also around 25%. but perf never increases linearly with price when you go up the product stack. so there is no doubt that HD 7950 is a better buy.

 

[RussianSensation]

Pretty embarrassing for nV here, although I wouldn't be so clear with a wider range of games, but this pretty much shows how overprices Nv's top 3 cards are.

Legion Hardware has a review with Cats 12.11 vs. 310.33 Betas. HD7950 Boost and HD7970 Ghz edition are clear price/performance winners.

HD7970 Ghz wins in 10 games, ties in 2 games (trade blows depending on resolution) and loses in 2 games against GTX680:
http://www.legionhardware.com/artic...hz_edition_7950_iceq_xsup2_boost_clock,4.html

GTX680 is now even slower in BF3:

Good to see HardOCP finally recognize that HD7950 is far superior to 660Ti and has been from the day 660Ti was released. With Sapphire HD7950 with 3 games going for $280, I can't even imagine buying a GTX660Ti in the US.

I find it most ironic that so many of us agreed that HD7000 series was overpriced until price drops started to trickle in May-June time-frame but even now many people still do not want to admit that GTX600 series has been overpriced since June. Not that any of this matters since we know people will still buy NV even if AMD cards cost less, overclocker better and/or are faster for the same price and come with a bunch of free games. JHH has mastered the marketing / brand value game.

 

[raghu78]

Legion Hardware has a review with Cats 12.11 vs. 310.33 Betas. HD7950 Boost and HD7970 Ghz edition are clear price/performance winners.

HD7970 Ghz wins in 10 games, ties in 2 games (trade blows depending on resolution) and loses in 2 games against GTX680:
http://www.legionhardware.com/artic...hz_edition_7950_iceq_xsup2_boost_clock,4.html

GTX680 is now even slower in BF3:

Good to see HardOCP finally recognize that HD7950 is far superior to 660Ti and has been from the day 660Ti was released. With Sapphire HD7950 with 3 games going for $280, I can't even imagine buying a GTX660Ti in the US.

I find it most ironic that so many of us agreed that HD7000 series was overpriced until price drops started to trickle in May-June time-frame but even now many people still do not want to admit that GTX600 series has been overpriced since June. Not that any of this matters since we know people will still buy NV even if AMD cards cost less, overclocker better and/or are faster for the same price and come with a bunch of free games. JHH has mastered the marketing / brand value game.

AMD has to improve how the market perceives their products. AMD's marketing and product promotion is nowhere near Nvidia's. They have to match Nvidia in terms of driver robustness which I feel they are yet to achieve. Also launch day performance and stability for the latest AAA titles needs to improve, especially for Crossfire. After 12.11 Nvidia's performance gap has widened considerably. but looks like 310.54 beta seems to be giving a nice performance boost in quite a few games, with BF3 benefiting significantly. the competition is good for the consumer. :thumbsup: