EVGA GTX 680 Classified 4GB Smashes 2 GHz GPU Clock Barrier

[btarunr]

EVGA's GeForce GTX 680 4GB, in the skilled hands of Vince "k|ngp|n" Lucido, smashed the 2 GHz barrier for GPU clock, setting a new clock speed world record for the GTX 680. The card achieved a GPU clock speed of 2,002 MHz, with a core voltage of 1.212V; and memory clock speed of 7.406 GHz. The card went on to score P16472 in 3DMark 11 (performance preset). The rest of the test-bench included Intel Core i7-3960X six-core processor clocked at 5.53 GHz, EVGA X79 Classified motherboard, 16 GB of quad-channel DDR3-2400 MHz G.Skill RipjawsZ memory, and EVGA NEX 1500W PSU. The GPU was cooled by k|ngp|n Tek-9 FAT 6.0 LN2 evaporator, while the CPU was cooled by k|ngp|n Dragon F1 Dark LN2 evaporator. Find the ORB page here.



View at TechPowerUp Main Site

 

[Assimilator]

That is extremely impressive. Has the 7970 breached the 2GHz barrier yet?

 

[1c3d0g]

LOVE this classified version's huge single fan.

 

[Quake]

EVGA's GeForce GTX 680 4GB, in the skilled hands of Vince “k|ngp|n” Lucido, smashed the 2 GHz barrier for GPU clock, setting a new clock speed world record for the GTX 680. The card achieved a GPU clock speed of 2,002 MHz, with a core voltage of 1.212V; and memory clock speed of 7.406 GHz. The card went on to score P16472 in 3DMark 11 (performance preset). The rest of the test-bench included Intel Core i7-3960X six-core processor clocked at 5.53 GHz, EVGA X79 Classified motherboard, 16 GB of quad-channel DDR3-2400 MHz G.Skill RipjawsZ memory, and EVGA NEX 1500W PSU. The GPU was cooled by k|ngp|n Tek-9 FAT 6.0 LN2 evaporator, while the CPU was cooled by k|ngp|n Dragon F1 Dark LN2 evaporator

Anyone who ever OCed any HW knows that 2GHz would be impossible with that voltage..

 

[amdftw]

This is fake...
Kingpin replaced the evga precession image or after 3D load he set the 2003Mhz clock.
Facts:
The power graph is not show that there was any 3dmark load.
I or everybody can set higher clocks in 2D which freeze in 3D.

So the simply way that kingpin is chose: He ran the 3dmark with lower clock, it would was 1800-1900Mhz, and after in 2D he set 2003Mhz.
This way he can tell he was the first who broke 2ghz but actually not...

 

[damric]

This is fake...
Kingpin replaced the evga precession image or after 3D load he set the 2003Mhz clock.
Facts:
The power graph is not show that there was any 3dmark load.
I or everybody can set higher clocks in 2D which freeze in 3D.

So the simply way that kingpin is chose: He ran the 3dmark with lower clock, it would was 1800-1900Mhz, and after in 2D he set 2003Mhz.
This way he can tell he was the first who broke 2ghz but actually not...

This is the conclusion that I came since there is not a GPU-Z validation of 2GHZ.

 

[hardcore_gamer]

That is extremely impressive.

I'll breach the 3Ghz barrier. Gimme some time to install Photoshop.

 

[Andrew Moore]

Did Galaxy not hit this already at Computex with their white PCB card?

 

[Hayder_Master]

GPU-Z shows it's on 1.845ghz, i think the test on this clocks and the 2ghz just a screen shot from freezing windows.

 

[ShockG]

I'd normally not reply to this but the ignorance about extreme overclocking by the posts above is staggering. None of you have run 3DM11 at anything close to those clock speeds or in any extreme environment and for you to call KNGPIN a cheat is evidence of this. Tee sub scores are in line with the clock speeds and indeed that score is perfectly acceptable and legitimate.

If you didn't know we change clock speeds during different tests on the fly during the benchmark. This applied to 3DMark Vantage GT2 as well where you had to clock lower than in GT1. You can use keyboard short cuts for this. GPU-Z validation isn't the end all be all, it's useful but it isn't always reliable and in the desktop it indeed may be 1846 that does not mean it was that during the benchmark. He's done 1950MHz before btw

 

[sanadanosa]

check this out.
http://hwbot.org/submission/2293742_kingpin_3dmark11___performance_geforce_gtx_680_16472_marks

 

[20mmrain]

That is extremely impressive. Has the 7970 breached the 2GHz barrier yet?

No it hasn't although 2nd place is held by a HD 7970..... only 200 points lower then this card. Guess what it was done @ 1700mhz. It shows... that the 7970's Stream Processors are more powerful then the GTX 680's CUDA core. It wouldn't take much to overcome those 200 points.
So while the 2ghz clock is impressive.... I don't believe this score will take too much... to be beaten by the HD 7970.

 

[Quake]

No it hasn't although 2nd place is held by a HD 7970..... only 200 points lower then this card. Guess what it was done @ 1700mhz. It shows... that the 7970's Stream Processors are more powerful then the GTX 680's CUDA core. It wouldn't take much to overcome those 200 points.
So while the 2ghz clock is impressive.... I don't believe this score will take too much... to be beaten by the HD 7970.

No, it doesn't show that 7970 is more powerful.. The difference is that 680 runs with default level of tessellation and its score is valid on ORB while 7970 runs with disabled/tweaked tessellation which you can adjust in ccc and its score isn't valid on ORB

 

[ShockG]

No you're mistaken.
HD7970 results are done with Tessellation turned off, that's why the scores are so high. With Tessellation turned on, the scores on the HD7970s are significantly lower.
We all know this who take part in Extreme OC

 

[okidna]

No you're mistaken.
HD7970 results are done with Tessellation turned off, that's why the scores are so high. With Tessellation turned on, the scores on the HD7970s are significantly lower.
We all know this who take part in Extreme OC

So, that's "Non-standard AMD Tessellation Setting 1" means tessellation OFF?

Thank you. You answered my curiosity.

 

[creepingdeath]

Who the fuck cares what kingpin does...Kingpin uses hardware mods out the wazzoo, cherry picked cards, in a special testing environment...anyone who is stupid enough to believe this has any relevance to REAL WORLD overclocking has issues.

I bet EVGA sells cards from this nonsense, I feel sorry for the sucker that falls for it.

 

[ShockG]

Who the fuck cares what kingpin does...Kingpin uses hardware mods out the wazzoo, cherry picked cards, in a special testing environment...anyone who is stupid enough to believe this has any relevance to REAL WORLD overclocking has issues.

I bet EVGA sells cards from this nonsense, I feel sorry for the sucker that falls for it.

We care. The rest of us who mod our cards using the "untouchables" VRM mod, the rest of us who buy Lightning cards, and freeze them specifically for this purpose of achieving high scores. This isn't meant to impress you of "real world overclocking". If you can't appreciate it then it's not for you, but please don't knock it. To some of us, go cold or go home is the motto, this "real world" air cooling overclocking you're talking about doesn't appeal to us at all.

The amount of effort required to do and achieve this is lost to you and that's ok but don't speak ill of an overclocking milestone. Big ups to EVGA and Vince for this achievement

 

[beck24]

No it hasn't although 2nd place is held by a HD 7970..... only 200 points lower then this card. Guess what it was done @ 1700mhz. It shows... that the 7970's Stream Processors are more powerful then the GTX 680's CUDA core. It wouldn't take much to overcome those 200 points.
So while the 2ghz clock is impressive.... I don't believe this score will take too much... to be beaten by the HD 7970.

100% WRONG! Turn on tess and watch 7970 collapse.

 

[TheGuruStud]

100% WRONG! Turn on tess and watch 7970 collapse.

You mean like this? http://techreport.com/articles.x/21404/1

Nvidia and their tricks. They're a sad competitor.

 

[MxPhenom 216]

This is fake...
Kingpin replaced the evga precession image or after 3D load he set the 2003Mhz clock.
Facts:
The power graph is not show that there was any 3dmark load.
I or everybody can set higher clocks in 2D which freeze in 3D.

So the simply way that kingpin is chose: He ran the 3dmark with lower clock, it would was 1800-1900Mhz, and after in 2D he set 2003Mhz.
This way he can tell he was the first who broke 2ghz but actually not...

Everything that has to do with Nvidia and Intel is just fake too you huh? :shadedshu

100% WRONG! Turn on tess and watch 7970 collapse.

No. 7970 and the 680 are pretty equal when it comes to Geometry/Tesselation horse power now. AMD went to there new architecture that really boosted their power in Tesselation, while Nvidia went a different route and made the cores a bit weaker and less complex but added a shit ton of them compared to the prior gens.

 

[nleksan]

I am just pleased to see it done with a 4GB card... Anyone else?

 

[cadaveca]

I am just pleased to see it done with a 4GB card... Anyone else?

I'm still confused about the whole GPU-Z not showing 2.002 GHz. Apparantly K1ngP1n stil ahs a score or two in his back pocket as well, so this is al lrather unexciting, really. Let's see the back-up scores!


"Saving" scores waiting for someone to beat your own, so you cna then one-up them...man...This truly isn't the record...The real record is the score he's got saved and will submit later!:shadedshu

 

[LAN_deRf_HA]

I'm still confused about the whole GPU-Z not showing 2.002 GHz. Apparantly K1ngP1n stil ahs a score or two in his back pocket as well, so this is al lrather unexciting, really. Let's see the back-up scores!

These cards always boost higher than what GPUz shows on the main window.

 

[cadaveca]

These cards always boost higher than what GPUz shows on the main window.

BUt then what you are saying is that it's a boost clock...and might not have been running for the full test, or what, or...


It's not this result that is confusing, it's nVidia's boost that is..well..confusing!

 

[evulmunk33]

so was it 2g throughout the whole run?
or did it only turbo to 2g occasionally throughout the bench?

others have hit 1950 as well, so then this "2g" run wouldnt really be anything special...