Archive Home arrow Reviews: arrow Video Cards arrow NVIDIA GeForce GTX 670 Video Card Tests

NVIDIA GeForce GTX 670 Video Card Tests E-mail
Reviews - Featured Reviews: Video Cards
Written by Olin Coles   
Thursday, 10 May 2012
Table of Contents: Page Index
NVIDIA GeForce GTX 670 Video Card Tests
First Look: GeForce GTX 670
DX10: Crysis Warhead
DX11: 3DMark11
DX11: Aliens vs Predator
DX11: Batman Arkham City
DX11: Battlefield 3
DX11: Lost Planet 2
DX11: Metro 2033
DX11: Unigine Heaven 3.0
Temperature and Power Consumption
GeForce GTX 670 Conclusion

First Look: GeForce GTX 670

The NVIDIA GeForce GTX 670 is a 1.5" tall double-bay, 3.9" wide, 9.5" long graphics card that will fit into nearly all mid-tower computer case enclosures with room to spare. GeForce GTX 670 is shorter than NVIDIA's GeForce GTX 570, GeForce GTX 580, and also the AMD Radeon HD 6970, and Radeon HD 7970 (each 10.5" long).


A rear mounted 60mm (2.4") blower motor fan with a slight offset takes advantage of the chamfered depression to draw cool air into the angled fan shroud, allowing more air to reach the intake whenever two or more video cards are combined in close-proximity SLI configurations. NVIDIA's add-in card partners with engineering resources may incorporate their own cooling solution into the GTX 670, but most brands are likely to adopt the cool-running reference design.


Specified at 170W Thermal Design Power output, the GeForce GTX 670 requires less power than its predecessor and several other flagship products. Because TDP demands have been reduced, NVIDIA's GeForce GTX 670 has also reduced power supply requirements to a pair of six-pin PCI-E power connections - identical to the GeForce GTX 570. However, with GeForce GTX 670 the two power connections are relocated to the side of the video card so it fits better into small enclosures.


The GTX 670 offers two simultaneously functional dual-link DVI (DL-DVI) connections, a full-size HDMI 1.4a output, and a DisplayPort 1.2 connection. Add-in partners may elect to remove or possibly further extend any of these video interfaces, but most will likely retain the original engineering. Only one of these video cards is necessary to drive triple-display NVIDIA 3D-Vision Surround functionality, when using both DL-DVI and either the HDMI or DP connection for third output. All of these video interfaces consume exhaust-vent real estate, but this has very little impact on cooling because the 28nm Kepler GPU generates less heat than past GeForce processors, and also because NVIDIA intentionally positions the heatsink far enough from these vents to equalize exhaust pressure.


As with past-generation GeForce GTX series graphics cards, the GTX 670 is capable of two and three card SLI configurations. Because GeForce GTX 670 is PCI-Express 3.0 compliant device, the added bandwidth could potentially come into demand as future games and applications make use of these resources. Most games work well using moderate settings on a single GeForce GTX 670 graphics card, but multi-card SLI configurations are perfect for gamers wanting to experience high-performance video games played at their best quality settings with bells and whistles enabled.


The exposed printed circuit board on the backside of the video card reveals an interesting discovery: GeForce GTX 670 uses a much smaller PCB then the profile suggests. Past GeForce products generally use a shroud to cover the entire length of the circuit board, but with GTX 670 the PCB measures only 7.0" of this 9.5" card, with a 2.5" extension to support the cooling fan.


In the next section, we detail our test methodology and give specifications for all of the benchmark tools and hardware used for our performance tests...

VGA Testing Methodology

The Microsoft DirectX-11 graphics API is native to the Microsoft Windows 7 Operating System, and will be the primary O/S for our test platform. DX11 is also available as a Microsoft Update for the Windows Vista O/S, so our test results apply to both versions of the Operating System. The majority of benchmark tests used in this article are comparative to DX11 performance, however some high-demand DX10 tests have also been included.

In each benchmark test there is one 'cache run' that is conducted, followed by five recorded test runs. Results are collected at each setting with the highest and lowest results discarded. The remaining three results are averaged, and displayed in the performance charts on the following pages.

A combination of synthetic and video game benchmark tests have been used in this article to illustrate relative performance among graphics solutions. Our benchmark frame rate results are not intended to represent real-world graphics performance, as this experience would change based on supporting hardware and the perception of individuals playing the video game.GPUZ-NVIDIA-GeForce-GTX-670.gif

Intel X79 Express Test System

DirectX-10 Benchmark Applications

  • Crysis Warhead v1.1 with HOC Benchmark
    • Settings: Airfield Demo, Very High Quality, 4x AA, 16x AF

DirectX-11 Benchmark Applications

  • 3DMark11 Professional Edition by Futuremark
    • Settings: Performance Level Preset, 1280x720, 1x AA, Trilinear Filtering, Tessellation level 5)
  • Aliens vs Predator Benchmark 1.0
    • Settings: Very High Quality, 4x AA, 16x AF, SSAO, Tessellation, Advanced Shadows
  • Batman: Arkham City
    • Settings: 8x AA, 16x AF, MVSS+HBAO, High Tessellation, Extreme Detail, PhysX Disabled
  • BattleField 3
    • Settings: Ultra Graphics Quality, FOV 90, 180-second Fraps Scene
  • Lost Planet 2 Benchmark 1.0
    • Settings: Benchmark B, 4x AA, Blur Off, High Shadow Detail, High Texture, High Render, High DirectX 11 Features
  • Metro 2033 Benchmark
    • Settings: Very-High Quality, 4x AA, 16x AF, Tessellation, PhysX Disabled
  • Unigine Heaven Benchmark 3.0
    • Settings: DirectX 11, High Quality, Extreme Tessellation, 16x AF, 4x AA

PCI-Express Graphics Cards

Graphics Card GeForce GTX570 Radeon HD6970 GeForce GTX580 Radeon HD7970 GeForce GTX670 GeForce GTX680
GPU Cores 480 1536 512 2048 1344 1536
Core Clock (MHz) 732 880 772 925 915 1006
Shader Clock (MHz) 1464 N/A 1544 N/A Boost 980 Boost 1058
Memory Clock (MHz) 950 1375 1002 1375 1502 1502
Memory Amount 1280MB GDDR5 2048MB GDDR5 1536MB GDDR5 3072MB GDDR5 2048MB GDDR5 2048MB GDDR5
Memory Interface 320-bit 256-bit 384-bit 384-bit 256-bit 256-bit



# ImpressiveBob 2012-05-10 08:42
Really impressive performance. Probably the best bang for buck in current cards. Good review too Olin.
Report Comment
# RE: ImpressiveOlin Coles 2012-05-10 08:44
Thank you! Unfortunately I was absent for about three of the five days I had this video card, so I didn't get to test as much as I would have liked.
Report Comment
# RE: RE: ImpressivetheUglyman0war 2012-05-11 21:33
Hey Olin great Job! would love to see some overclocks and SLI benchmarks at higher setting like enthusiast crysis warhead and why not turn on physx on the Batman benchmark?
Also was wondering if Battlefield 3 on three monitors and 3dVision will be as memory hungry or if I should wait for the upcoming 4 gig cards?
Report Comment
# RE: NVIDIA GeForce GTX 670 Video Card Testsclaydough 2012-05-10 09:35
Holy Moly! Jeez Louise!

Go to Toughest bench first -> Metro2033 3 frame difference
Go to To will it play nerfed -> Crysis2 3 frame difference
Go to Gbuffer tech to see memory intensive app -> Battlefield3 3 frame difference!
Finish off with Physx example/open city that opened with shakey driver support: ArkhamCity... 3 Frame Differnce

?Check Price again...
According to Nvidia Math 3 frames represents the single GPU flagship performance that justifies a $100 price difference?

Perhaps a metro 2033/ Battlefied3 3Dvision + NV surround With dual tri ( quad? ) SLI is needed to show more advantage not yet exposed?
Otherwise I sure am glad there have not been any GTX 680 stock.

( I suppose a new wave of powerful games that will not be as nerfed by console economics when the next gen of consoles are released will validate the $100 difference as well... But I would suspect that Metro 2033 should have been enough of a burden for the gtx 680 to prove the advantage of it's muscle?? )

Was a GK 100 chip or GTX 780 rumor ever accounted for??
It would seem reasonable by looking at the results of this bench that perhaps some of the rumors have some truth? Perhaps the gtx 680 is a different flagship class then the 480 and 580???
( otherwise the gtx 570 was virtually equal to a 480... Compared to the KO "beatdown" the 670 gives the 580? ( boom! )

Don't know why the article starts with a concentration comparing against the 570 it replaces? The obvious story is: Why is this card practically a gtx 680 gk104 with the same memory?? Just turned down slightly? Better power requirement handling under load? While at lower temperture under load? ( with a 3 frame difference at the max settings given I would love to see if it can overclock as well? )

If it can... Then for $300 dollars cheaper! I can easily see the gtx670 tri sli the big story of the year ( if it scales well? )

Just seems so weird at the price point to actually dive in just yet. But if nothing develops that explains or ruins the ridiculous value... after x-mas I will definitely tri Sli the 670.
Report Comment
# RE: RE: NVIDIA GeForce GTX 670 Video Card TestsDavid Ramsey 2012-05-10 09:49
Since you can't buy any new Kepler-based video card yet, it's largely academic.
Report Comment
# RE: RE: RE: NVIDIA GeForce GTX 670 Video Card Testsclaydough 2012-05-10 10:01
More like a blessing :)
Report Comment
# RE: RE: NVIDIA GeForce GTX 670 Video Card TestsSempifi99 2012-05-11 10:57
Accounting for rumors is pointless. For all we know the GK 100 will be for workstations only, or perhaps it was scrapped for "other" reasons.
Report Comment
# RE: RE: RE: NVIDIA GeForce GTX 670 Video Card TeststheUglyman0war 2012-05-11 21:09
Considering I pulled the brakes on Keplers launch cuz the air stank of fish...
Knowing that I might have saved $300 on my tri sli replacement makes me so happy I have spent the past round of benchmarks squealing in glee like a little girl riding a pretty pink bony bareback!(some of which showing even less of a divide like Tom's showing "less than" < the 3 seconds in Metro 2033 even at the extrem rez of 2650 x 1600!!)
Sure is hard for me to consider considering the possible implications "pointless"? Tho I suppose current developments might just be dumb luck.
But let us assume u r exactly right? In which case, what happens to gtx 680 production? How can they possibly sell any more of them at $100 more now? What kind of idiot would u have to be? (I suppose some might just blindly trudge on despite the obvious value disparity?)
When the single gpu flagship pride has been a point of prestige as much as anything else! And in light of the recent Jon Pedi Research numbers showing the projected multi-Billion ( $23 Billion ) Dollar PC gaming market explosion driven by performance gaming systems. Where according to their research:
The Enthusiast Market of systems costing over $3400 dollars a build represent as much of the final dollar share in the end as the performance systems costing over $1000 a build! And each of which alone, are more than double the markets dollar share compared to the ghetto gamer (gaming builds costing about $800)who invariably QQ's the loudest in support of their needs when they are contributing the least to the detriment of the gaming experience suffered by the peformance and above market who actually created this explosion in the first place!) In other words, The low and Mid Range users at the end of this console cycle are poised to become the new least common denominator that brings down the level of gaming for the rest of us! (Now that soon the blame can not be leveled squarely with the consoles)
In light of the new PC gaming revolution the last thing Nvidia would want to do is to lose credibility on their flagship model when the enthusiast Market numbers show how important they actually are.
( Unless they r going to play smoke and mirrors and just Promote their dual GPU flagship and the GTX 670 and Ignore the GTX680?
Report Comment
# RE: RE: RE: RE: NVIDIA GeForce GTX 670 Video Card TeststheUglyman0war 2012-05-11 21:27
That strategy of simply ignoring the gtx 680 actually works for me come to think of it?
Just as long as there is a price drop now on the gtx 690 so original gtx 680 hopefuls do not do something really dumb out of anger and haste like "go with AMD".
That would be a tough choice between them.
I am kind of excited now to see what results from the upcoming 660 benchmarks?
At this point maybe we will see another deal. Where the card is to powerful compared to the 670!

It could happen. I haven't gone cheap in a long time. it would be awesome actually? Considering Maxwelll seems to be creeping up incredibly fast?
Report Comment
# RE: NVIDIA GeForce GTX 670 Video Card Testsclaydough 2012-05-10 09:54
Oh no! wait Batman was actually a 2 frame difference not 3...

But then again Metro 2033 was actually 4 seconds not 3!
So maybe the 680 did "start" to show it's muscle by pulling away at the burden of Metro 2033?
( So perhaps a stress test representing next gen 2.0 upcoming market with the added burden of extreme resolution and 3dvision? )

That Metro 2033's burden still only resulted in a 1 frame difference does not seem to inspire much confidence though? ( heck 1 frame isnt enough concern for most users to give any credence to an PCI-E X8 speed difference )

Report Comment
# RE: NVIDIA GeForce GTX 670 Video Card Testss2korpio 2012-05-10 14:19
The red team is gonna have a hard time swallowing this one. I hope AMD brings out their best on the mid and budget range.
Report Comment
# RE: NVIDIA GeForce GTX 670 Video Card Testsweilrich 2012-05-10 14:32
I had hoped that that GTX 660Ti would be out (around May 10, 2012) as suggested in related news, but it appears that nVidia wants to wait until Q3 to release it. The price difference ($249 for the 660 versus $399 for the 670) makes the GTX a more economical choice for me - I still can't justify spending $400 for the screaming GTX 670, even with these really good reviews. I suppose I should be glad that the Kepler cards aren't available yet, and limp along with my gimpy old Radeon HD 5770 and 4650, but as I watch the GPU temperature on the 5770 hover around 90C, I start getting really nervous - I've already moved all my games over to the older 4650 and only use the 5770 for word processing and the like. Excellent review, Olin - it really makes me want one of these new Kepler cards.
Report Comment
# RE: RE: NVIDIA GeForce GTX 670 Video Card TestsAthlonite 2012-05-10 17:24
if your getting 90c on an HD5770 there's something wrong either it's choked up the waazoo with dust or your case sucks for air throughput

take it out clean it replace the thermal paste with some AS5 and come back and tell us how well the temps dropped
Report Comment
# RE: NVIDIA GeForce GTX 670 Video Card Testsweilrich 2012-05-10 18:29
I've already done that - took the card out, disassembled it, blew compressed air through the fan, cleaned every surface, applied fresh heat sink compound, cleaned the case, checked all the case fans, made sure there was a blank slot between it and the other card, check the temperature in the case (never exceeded 30C), made sure the GPU fan turned when the card was on (it does). I contacted ATI support about it, nothing - they only sent me an auto-reply. I tried to set the fan speed manually using CCC, but it doesn't do anything - the fan never changes speed. I'm using the latest driver (12.4). The fan may be defective, but it is still turning, and there is no easy way to replace it. What does work, is to leave the case open and place a floor fan up against it to blow cold air on it. I wear a sweater and turn up the A/C, but my wife hates having the house so cold so I can only use that tactic when I know she is going to be out for a few hours.
Report Comment
# RE: RE: NVIDIA GeForce GTX 670 Video Card TestsAthlonite 2012-05-11 03:32
hmmm sounds like then the only option open to you is an third party HSF for it then as 90c is ridiculously high temp mine never goes over 57c when folding 54c when gaming if I leave the fan set to auto.. Arctic have some nice and reasonably priced HSF's for graphics cards you may want to try one of those
Report Comment
# $550???Sempifi99 2012-05-11 10:54
I have not seen a 7970 at $550 in a long while. On newegg, stock ones are around $450-470. Still the 470 provides better value, but by a much smaller margin.
Report Comment
# So add up the numbers and compareKiernan 2012-09-02 18:29
670 is only 50% of whatever 570 is.. And The 680 is just a little bit more than the 670..
Report Comment
# BTWKiernan 2012-09-02 18:31
You can get a 570 at TigerDirect for about 240 after rebate..
I'm looking to use it for CUDA rendering..
Report Comment
# Am I wrong?Kiernan 2012-09-02 18:33
It would be better to get two 570's than one 680..
Report Comment
# RE: Am I wrong?David Ramsey 2012-09-02 20:21
Unless, of course, you want to take advantage of some of the new features in the Fermi cards, like their motion-based antialiasing, adaptie vSync, and such...
Report Comment
# oh yaKiernan 2012-09-02 18:36
You gain greater longevity with redundancy.. Hence it is better to have two kids than one.. If you have one, greater likelihood it will die, unless with two one kills the other..But lonely people are prone to suicide..
Report Comment
# RE: oh yaKiernan 2012-09-02 18:39
I mean greater likelihood one of the two will be around longer than just the one.. So get two cards not one.. One fails you at least can go 2 thirds the speed until the one that failed is fixed..
Report Comment
# RE: RE: oh yaKiernan 2012-09-02 18:41
Unless of course you don't have two PCIe ports on your motherboard, in which case go with one..
Report Comment

Comments have been disabled by the administrator.

Search Benchmark Reviews

Like Benchmark Reviews on FacebookFollow Benchmark Reviews on Twitter