• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

MSI 7970 vs Evga 680 The real bench results T&B

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

TURN & BURN

Member
Joined
Jan 19, 2006
Location
Florida
Hi All,

Well me and a friend dont beleive everything we read including web reviews on new product, so we did some real time testing under a level playing field between the 7970 and the 680.

It was no suprise to Me as to the outcome however my friend did expect better.

Lets start with the field!!! or system specs:

T&B
Asus Rampage 4 Formula bios 1202
8320 @ stock 38x100.3 turbo mode
GSkill RipjawZ 2133 16GB CL9-11-10-28 @ 1600MHz
MSI 7970 Clocked to 1099/1502 2 MHz high on the GPU to the 680 but as close as I can get it memory clocks same making the cards equal in speed respect.
OCZ Vertex 3 120GB windows 7 64bit

Abeeftec
Asus Rampage 4 Extreme bios 1202
8320 @ stock 38x100.3 turbo mode
GSkill RipjawZ 2133 16GB CL9-11-10-28 @ 1600MHz
Evga GTX 680 1097/1502
2x OCZ 120GB raid 0 windows 7 64 bit

Bench programs, we used HWbot video card bench programs and cine.

06 Marks
3D Marks Vantage
11 Marks
Unique haven Extreme D11
Cine-Bun 11.5+

Now for the not so startling results from the 680:rofl: which is rated to have 10% and them some better power than the 7970.

First are the cards under load with clocks!!

MSI 7970 Load test clocks
CPUZ_speed_VALD.JPG


Evga GTX 680 load test clocks
CLOCK_CHECK_680.jpg


As you can well see the cards are set to run at the same GPU and MEMORY speeds so there is no advantage in this respect providing a much more true and stable reading on all testing!!

I will have Abeeftec also put up his point of view on this as it was a joint venture and he was expecting the 680 to BLOW away the 7970!

Note in every screen shot to follow we loaded 3 CPUZ's showing
Mobo
CPU
Memory

Also loaded is GPUZ on all screenshots!

NOW FOR THE FUN READY>>>>>>>>>>>>>>

First up lets do simple Cinebench MSI 7970 will be first in all benching picks.

CINE-BUNS MSI 7970 OpenGL
AR4F_7970_8320_stock_CinaBun.JPG


CINE-BUNS Evga GTX 680 OpenGL
AR4X_680_3820_Stock_CineB.jpg


This was rather a blow out we didn't expect this much of a difference at all very suprised with the MSI 79.86 to the Evga's 53.40

Now things get closer on the next few but!!!!!!

06 MARKS MSI 7970
AR4F_7970_3820_stock_06M.JPG


06 MARKS Evga GTX 680
AR4X_680_3820_Stock_06m.jpg


On 06 marks we were looking at the MSI 7970 28264 vs Evga 27374 still a rather LARGE difference seems its not getting any better for the 680!

We move on now lets see VANTAGE up!!

3D Marks Vantage MSI 7970
AR4F_7970_8320_stock_VAN.JPG


3D Marks Vantage Evga 680
AR4X_680_3820_Stock_Vant.jpg


Still I'm not seeing the 10%+ out of the 680?? MSI 7970 GPU score 35511 vs Evga 680 GPU score 33476 OUCH?:bang head

So far I have to say give me the 7970 :clap:

OK some UH extreme again 7970 :D

UH Extreme MSI 7970
AR4F_7970_3820_stock_UHX.JPG


UH Extreme Evga 680
AR4X_680_3820_Stock_Heav.jpg


Once again proper testing at same card speeds shows the 7970 beating the 680 :screwy: what can I say..... MSI= 2138.752 Evga= 1837.98 maybe a bad 680?? I dont think so especially since my 7970 been clocked to near death for the last 2 months!

Finaly the Evga shines but to little to late?

11 Marks MSI 7970
AR4F_7970_8320_stock_11M.JPG


11 Marks Evga 680
AR4X_680_3820_Stock_m11.jpg


Here the Evga 680 SHINED horay!!!:thup:

Now I will have Abeeftec chime in on his thoughts as we did this beacuse he expected alot more from the 680. We may also be doing more benching between the cards soon to add to this post.
 
Last edited:
Interesting results... though I'm pretty sure you can't compare clock speed due to different architecture. They're both powerhouse cards, no doubt about it.
 
nice. always funny how whatever site wants to look better they can make it look that way.
 
06 Marks = DX9: 7970 +3%
3D Marks Vantage = DX10: 7970 +4%
Unique haven Extreme was tested in DX9: 7970 +6%
11 Marks = DX11: 680 +5%

CINE: what can i say but wow...

None of this translates into Gaming, mostly not DX10 /11 these days..

what we want is games tested without nVidia looking over the shoulder :)
 
Last edited:
I appreciate the testing, but keeping them at the same clock speed isn't a very good test, I think you have to go in ratio for the card (stock vs stock, 10% vs 10% OC) etc, unless you are testing maximum overclock on each card, which can be luck of the draw if you get a good/bad overclocking card.

Overall it looks like the 7970 may be a much better card for benching, as Bobnova stated otherwise we would see a lot more results being posted on hwbot with the 680(s)
 
I appreciate the testing, but keeping them at the same clock speed isn't a very good test, I think you have to go in ratio for the card (stock vs stock, 10% vs 10% OC) etc, unless you are testing maximum overclock on each card, which can be luck of the draw if you get a good/bad overclocking card.

I understand what your saying but the fact is the test would be flawed to test at there stock speeds as there is a clear advantage to the card with higher clocks!

If the cards are matched to GPU & Memory speeds then it becomes which card is better by fact of the production and drivers ect.

They both have the same die size and running them at the same clocks gives a even play ground.

Anything else would be a waste as the higher clocked card will beat the lower card evertime!

This test was just to show if the cards were indeed clocked the same, which would pove to be better overall and thats not to say there was any great crushing defeat.

We will also try sli vs crossfire if we can to see how the scaling is between cards.

But the main reason for this post was to show that with the same clocks which card prevails nothing more.

Thanks for the input as we expected it.

T&B
 
The higher clocked card would only (should only) be faster if comparing the same architecture. It's for the same reason that a higher clocked AMD chip loses to Intel even with the same die size/etc.
 
@ TURN & BURN

I tend to agree with that, in the same way that a lot of CPU's are the same but simply clocked higher,- i did not get the Phenom II 1100T because its exactly the same as a 1090T just factory set a little higher.... i can easily do that myself and save the extra money.

Yes the GPU's are not even the same brand and for now the 680 is cheaper (in the USA, but not the UK)

Yes if they run at the same clock speed's and are as good as eachother clock for clock, or one is better... then for me that changes everything.

Then it depends on which gets the most out of your games at there 24/7 stable clocks.
 
06 Marks = DX9: 7970 +3%
3D Marks Vantage = DX10: 7970 +4%
Unique haven Extreme was tested in DX9: 7970 +6%
11 Marks = DX11: 680 +5%

CINE: what can i say but wow...

None of this translates into Gaming, mostly not DX10 /11 these days..

what we want is games tested without nVidia looking over the shoulder :)

I coulden't agree more, we weren't looking to compare gaming but just simple benching, we know we will get alot of flack on this as was posted already about the clocks being the same, but we feel it makes it a good way to see how they both run at matched speeds.

This is in no way a bashing post for either AMD or Evga just a experiment of the cards in benching.

In fact they are very closely matched in most respects except cine & 11 marks where we got some oppsite ends of the spectrum.

SO everyone please enjoy post what you may.
 
I coulden't agree more, we weren't looking to compare gaming but just simple benching, we know we will get alot of flack on this as was posted already about the clocks being the same, but we feel it makes it a good way to see how they both run at matched speeds.

This is in no way a bashing post for either AMD or Evga just a experiment of the cards in benching.

In fact they are very closely matched in most respects except cine & 11 marks where we got some oppsite ends of the spectrum.

SO everyone please enjoy post what you may.


Yeah i understand that :)
 
The higher clocked card would only (should only) be faster if comparing the same architecture. It's for the same reason that a higher clocked AMD chip loses to Intel even with the same die size/etc.

I think thats what we are looking at is the architecture! does AMD or Evga offer more for the same clock speed who architecture seems to have the edge overall?

I hope this better helps understand what we were after here and helped provide some insite.
 
Last edited:
But you can't just equate clocks, because you haven't taken into consideration the actual architecture. You could draw from this that the GTX680 is much more efficient because it has 75% of the SP's as the HD7970 and still beats it in most (game) benchmarks, and is only behind in your testing overall not including Cinebench (see below).

IMO I would think the most useful clock for each card would be their average of their maximum potential (which you can't do with just 1 card obviously). Also in the results the GTX680 might have been downclocking, especially on the heavier benchmarks - since the power target is 100%. Even at 132% my card downclocks from the maximum boost when overclocked, it doesn't constantly stay at the peak clock.

The GTX680 I think is gimped for DP, which may explain the low Cinebench score (to make the Quadro cards seem more worthwhile).
 
The GTX680 I think is gimped for DP, which may explain the low Cinebench score (to make the Quadro cards seem more worthwhile).

I do remember hearing rumors about that a while ago. Struck me as a very nvidia thing to do.
 
nice benches and comparison... but totally flawed methodology. I am not hating on you, just trying to help you do better next time. :thup:

You cannot overclock a 7970 and compare it to a stock 680. That is no way a fair and equal representation of the 680. clock for clock comparison cannot be made across architectures and companies (intel vs AMD)

As has been said by others, good comparisons are "stock vs stock" "stock+10% vs stock+10%" and "average OC vs average OC" (this would require a large sample size to get an accurate average since it is hit and miss how well cards OC)
 
nice benches and comparison... but totally flawed methodology. I am not hating on you, just trying to help you do better next time. :thup:

You cannot overclock a 7970 and compare it to a stock 680. That is no way a fair and equal representation of the 680. clock for clock comparison cannot be made across architectures and companies (intel vs AMD)

As has been said by others, good comparisons are "stock vs stock" "stock+10% vs stock+10%" and "average OC vs average OC" (this would require a large sample size to get an accurate average since it is hit and miss how well cards OC)

I think that's just playing on semantics.

There are none reference 7970 running at these clock speeds out of the box, even CCC will take it to 1250.

Whose to say nVidia have not overclocked there reference cards already? when exactly does a clock become an overclock?

The point here is how they compare at the same clock.
 
nice benches and comparison... but totally flawed methodology. I am not hating on you, just trying to help you do better next time. :thup:

You cannot overclock a 7970 and compare it to a stock 680. That is no way a fair and equal representation of the 680. clock for clock comparison cannot be made across architectures and companies (intel vs AMD)

As has been said by others, good comparisons are "stock vs stock" "stock+10% vs stock+10%" and "average OC vs average OC" (this would require a large sample size to get an accurate average since it is hit and miss how well cards OC)

If you are comparing clock vs. clock of course you can.

The 680 has turbo built-in to ramp up higher clocks which you could technically call overclocking.
 
I think that's just playing on semantics.

There are none reference 7970 running at these clock speeds out of the box, even CCC will take it to 1250.

Whose to say nVidia have not overclocked there reference cards already? when exactly does a clock become an overclock?

The point here is how they compare at the same clock.
Nvidia has stated their ('there' is not possessive - sorry pet peeve of mine) reference clock speed is 1006Mhz. PERIOD. It boosts to 1058Mhz. You dont see the 2600K being listed at 3.8Ghz on the box since it will turbo to 3.8Ghz, right? Err wait you know AMD better... You dont see the FX-8150 listed at 4.2Ghz, right? Thet GTX680's reference clock is 1006Mhz with a boost feature. That out of the way, it becomes an overclock when its over the reference/boost speeds...a.k.a - above reference speeds.

The only thing the clock for clock comparisons show are the efficiency of the architecture. You can IF IF IF and BUT BUT BUT all day, but in order to fit in AMD's own power thresholds, and likely noise, the 7970's are reference clocks are what they are regardless if they can overclock to the moon.





If you are comparing clock vs. clock of course you can.

The 680 has turbo built-in to ramp up higher clocks which you could technically call overclocking.
 
Nvidia has stated their ('there' is not possessive - sorry pet peeve of mine) reference clock speed is 1006Mhz. PERIOD. It boosts to 1058Mhz. You dont see the 2600K being listed at 3.8Ghz on the box since it will turbo to 3.8Ghz, right? Err wait you know AMD better... You dont see the FX-8150 listed at 4.2Ghz, right? Thet GTX680's reference clock is 1006Mhz with a boost feature. That out of the way, it becomes an overclock when its over the reference/boost speeds...a.k.a - above reference speeds.

The only thing the clock for clock comparisons show are the efficiency of the architecture. You can IF IF IF and BUT BUT BUT all day, but in order to fit in AMD's own power thresholds, and likely noise, the 7970's are reference clocks are what they are regardless if they can overclock to the moon.

What are you dribbling on about? again your ranting about something completely off subject.

AMD / nVidia set there clocks and we set our own, to see how they compare there is perfectly valid.
Its really very simple.
You need to learn to relax a little.
 
Back