• Home
  • blog
  • ASUS GeForce GTX 670 DirectCU II

ASUS GeForce GTX 670 DirectCU II

The ASUS GTX 670 DirectCU II is one of the few high-end graphics cards that run really quiet and cool right out of the box.

December 16, 2012 by Lawrence Lee

ASUS GTX 670 DirectCU II (GTX670-DC2-2GD5)
PCI-E Graphics Card
Sample Provider Puget Systems
Street Price

Our first experience with the Nvidia’s latest Kepler core came with the ASUS
GTX 680 DirectCU II OC
, which was armed with a massive cooling solution
with dual fans. The card impressed us greatly with its mix of superb performance,
high energy efficiency, and reasonable noise output. But being a GTX 680, Nvidia’s
flagship single GPU card, it’s quite expensive, not only for the vast majority
of users, but even gamers willing to spend a good chunk of cash for a smooth
gaming experience.

GTX 600 Series Comparison (Single GPU)
GTX 680
GTX 670
GTX 660 Ti
GTX 660
Transistor Count
3.54 billion
3.54 billion
3.54 billion
2.54 billion
Die Size
294 sq. mm
294 sq. mm
294 sq. mm
221 sq. mm
SM count
Shader Units
Core Frequency
1006 MHz
915 MHz
915 MHz
960 MHz
Avg. Boost Frequency
1058 MHz
980 MHz
980 MHz
1032 MHz
Memory Frequency
6008 MHz
6008 MHz
6008 MHz
6000 MHz
Memory Bus
Street Price
(2GB version)

Like most types of PC hardware, video cards suffer from a case of diminishing
returns as the price increases. The GTX 670 promises to be an attractive alternative,
essentially a slightly cut-down version of the GTX 680 for about US$80
less. Lower power demand also reduces electricity consumption and makes the
GPU easier to cool quietly. This is quite evident when you look the respective
DirectCU II versions of the two cards from ASUS; the GTX 680’s triple slot cooler
dwarfs the dual slot version found on the GTX 670.

The ASUS GTX 670 DirectCU II.

While it’s more modestly sized, the GTX 670’s DirectCU II’s (and all DirectCU
II variants) has a similar look. The heatsink spans over the majority of the
card with heatpipes intersecting the main body at various points. Airflow is
provided by two fans surrounded by a plastic shroud that directs air toward
the GPU and board components rather than escaping out the sides.

The cooler appears virtually identical to the one found on the HD
7870 DirectCU II,
which interestingly enough, didn’t produce very good
environmental results in our GPU test system despite having a similar TDP to
the GTX 670. And yet, our GTX 670 sample came from the Puget
Serenity Pro
, which exceeded the demanding acoustic requirements for
our Silent PC Certification
by a healthy margin. We’re anxious to investigate this odd discrepancy.

ASUS GTX 670 DirectCU II (GTX670-DC2-2GD5): Specifications (from the product
web page
Graphics Engine
NVIDIA GeForce GTX 670
Bus Standard
PCI Express 3.0
Video Memory
Engine Clock
GPU Boost Clock : 980 MHz
GPU Base Clock : 915 MHz
CUDA Cores 1344
Memory Clock 6008 MHz ( 1502 MHz GDDR5 )
Memory Interface 256-bit
Resolution DVI Max Resolution : 2560×1600
Interface DVI Output : Yes x 1 (DVI-I), Yes x 1 (DVI-D)
HDMI Output : Yes x 1
Display Port : Yes x 1 (Regular DP)
Power Consumption
up to 225W2 additional 6 pin PCIe power required
1 x Power cable
ASUS Utilities & Driver
ASUS Features
DirectCU Series
Super Alloy Power
10.7 ” x 5.4 ” x 1.7 ” Inch
*To have the best cooling performance, ASUS GTX670-DC2-2GD5 extends the fansink to 2 slots. Please double check you Chassis and Motherboard dimension prior to purchase to make sure it fits in your system!
*Note that the actual boost clock will vary depending on actual system conditions. For more information, please visit http://www.geforce.com/

ASUS has a few different variants of the GTX 670; our sample is the GTX670-DC2-2GD5, the most basic model they offer, sharing the same specifications as the reference version. The card’s GPU is clocked at 915 MHz but can upped to 980 MHz under favorable thermal conditions thanks to Nvidia’s GPU Boost features which adjusts the clock speed in a similar fashion to Intel and AMD’s respective dynamic CPU overclocking features. It also ships with 2GB of ultra-fast 1502 MHz GDDR5 memory though 4GB versions are available as well.


The ASUS GTX 670 DirectCU II is 23.2 cm long measured from the rear panel to the far edge of the circuit board and the cooler adds an additional 3.5 cm to its length. The heatpipes also hang over the edge of the PCB, increasing its effective width by 2.0 cm. The heatsink accounts for 500 grams of the card’s 710 gram total weight.

The heatpipes protruding out the side are noticeably thick, making the
heatsink seem quite formidable. The fans aren’t particularly large for
a dual fan GPU cooler, though, equivalent in diameter to a standard 80
mm case fan and very thin in profile.

The card has a standard set of video outputs: DisplayPort, HDMI, DVI-I and DVI-D.

The trace side of the PCB is covered with a thick metal backplate secured at multiple mounts. Ventilation holes have been punched out over much of the surface.

The backplate extends over the edge of the PCB forming a grill. The card’s two 6-pin PCI-E power connectors are positioned on the side so they don’t add any length to the card. However one of the heatpipes nearby makes it difficult to get a grip on the release underneath the plugs.

The heatsink is secured only by four spring-loaded screws on the back. It has three 8 mm thick direct-touch heatpipes made of copper which intersect aluminum fins at three different spots. The fins are approximately 0.42 mm thick and spaced a tight 1.35 mm apart.

The PCB layout is remarkably tidy with nothing notable interfering with the area around the GPU core. There is no additional cooling aside from the backplate and a small black heatsink near the rear of the card, presumably for the voltage regulation circuitry.


Our test procedure is an in-system test, designed to:

1. Determine whether the cooler is adequate for use in a low-noise system.
By adequately cooled, we mean cooled well enough that no misbehavior
related to thermal overload is exhibited. Thermal misbehavior in a graphics
card can show up in a variety of ways, including:

  • Sudden system shutdown, reboot without warning, or loss of display signal
  • Jaggies and other visual artifacts on the screen.
  • Motion slowing and/or screen freezing.

Any of these misbehaviors are annoying at best and dangerous at worst
dangerous to the health and lifespan of the graphics card, and sometimes to
the system OS.

2. Estimate the card’s power consumption. This is a good indicator of how efficient
the card is, and it affects how hot the GPU runs. The lower the better.

3. Determine how well the card decodes high definition video.

Test Platform

GPUs Compared:

GPU-Z screenshot: ASUS GTX 670 DirectCU II.

Measurement and Analysis Tools

3D Performance Benchmarks (for low-end/budget graphics processors only)

Estimating DC Power

The following power efficiency figures were obtained for the
Kingwin LZP-1000
used in our test system:

Kingwin LZP-1000 Test Results
DC Output (W)
AC Input (W)

This data is enough to give us a very good estimate of DC demand in our
test system. We extrapolate the DC power output from the measured AC power
input based on this data. We won’t go through the math; it’s easy enough
to figure out for yourself if you really want to.

Ambient Noise Level

Our test system’s CPU fan is a low speed Scythe that is set to full speed at all times. The two Antec TrueQuiet 120 case fans are connected to the motherboard and are controlled using SpeedFan. Three standard speed settings have been established for testing.

GPU Test System:
Anechoic chamber measurements
System Fan Speed
System SPL@1m
High (loud)
1130 RPM
26 dBA
Med (quiet)
820 RPM
18 dBA
Low (silent)
580 RPM
12~13 dBA
Note: mic is positioned at a distance of one meter from the center of the case’s left side panel at a 45 degree angle.

When testing video cards and coolers with active cooling, the low setting will be used. For passive cards and heatsinks, all three settings will be tested to determine the effect of system airflow on cooling performance.

Video Test Suite


1080p | 24fps | ~22 mbps

H.264/MKV: A custom 1080p H.264 encoded clip inside an Matroska container.


1080p | 24fps | ~2.3 mbps

Flash 1080p: The Dark Knight Rises Official Trailer #3, a YouTube HD trailer in 1080p.


Testing Procedures

Our first test involves monitoring the system power consumption as well as CPU and GPU temperatures during
different states, idle, under load with Prime95 to stress the processor, and Prime95 plus FurMark to stress both the CPU and GPU simultaneously. This last state is an extremely stressful, worst case scenario test which generates
more heat and higher power consumption than can be produced by a modern video
game. If the card can survive this torture in our low airflow system, it should be
able to function normally in the vast majority of PCs. Noise levels are measured and recorded as well; if we deem the card’s fan control to be overly aggressive, we can adjust them at our discretion using various software tools.

Our second test procedure is to run the system through a video test suite featuring
high definition clips played with PowerDVD and Mozilla Firefox (for Flash video). During playback, a CPU usage graph is created
by the Windows Task Manger for analysis to determine the average CPU usage.
High CPU usage is indicative of poor video decoding ability. If the video (and/or
audio) skips or freezes, we conclude the GPU (in conjunction with the processor)
is inadequate to decompress the clip properly.

Lastly, for low-end and budget graphics cards, we also run a few gaming benchmarks
to get a general idea of the GPU’s 3D performance. We don’t feel this is necessary
for high-end models as there are many websites that do this in painstaking detail.


Baseline Power with Integrated Graphics:

Power Consumption Measurements:
GPU Test System (Intel HD 2000 IGP)
CPU Load
CPU + GPU Load
Sys. Power (AC)
Sys. Power (DC)
System fan speeds: 580 RPM
Ambient noise level: 10~11 dBA
System noise level: 12~13 dBA
Ambient temperature: 22°C

System with Discrete Graphics:

System Measurements: GPU Test System
(ASUS GTX 670 DirectCU II)
CPU Load
CPU + GPU Load
GPU Fan Speed
1140 RPM
1890 RPM
SPL @1m
13~14 dBA
16 dBA
Sys. Power (AC)
Sys. Power (DC)
System fan speeds: 580 RPM
Ambient noise level: 10~11 dBA
System noise level (on int. graphics): 12~13 dBA
Ambient temperature: 22°C

With our test system fans on our low speed 12~13 dBA@1m setting, the GTX 670 DirectCU II ran very cool and quiet when idle. The fans were spinning at 1140 RPM but had a minimal effect on the overall noise level. Power consumption was only 54W AC which is fairly reasonable for a modern high-end GPU. Putting the system on CPU load was only slightly detrimental to GPU cooling. The addition of a GPU load pushed the GPU fans to 1890 RPM for a total noise level of just 16 dBA@1m. The fan control was quite aggressive as the GPU core was only registering a comfortable 73°C (we consider anything under 90°C to be acceptable).

Thanks to Nvidia’s GPU Boost feature, the GPU core clock stabilized at 967 MHz on load, a significant increase from its 915 MHz base frequency. On a lark, we manually increased the GPU fan speed to 100% (3540 RPM) to give it more thermal headroom. The system produced a noise level of 29 dBA@1m and the GPU frequency topped out at only 980 MHz.

The acoustics of the GTX 670 DirectCU II were simply excellent. At idle, the extra noise it produced was imperceptible compared to our system running solely on integrated graphics. On load, the machine was 2~3 dB louder but the overall sound didn’t really change aside from some extra output at ~1000 Hz. Subjectively, it was very smooth and unobtrusive.

Noise & Cooling Comparison

Comparison: GPU Test System (Load)
Est. Power Draw (DC)
GPU Temp
SPL @1m
DirectCU II
16 dBA
HIS HD 5870 Turbo
GELID Icy Vision @5V
17~18 dBA
AMD HD 6870 +
GELID Icy Vision @5V
17~18 dBA
DirectCU II OC
27~28 dBA
ASUS HD 7870
DirectCU II
30~31 dBA
Gainward GTX 560
Ti Phantom
37 dBA
System fan speeds: 580 RPM
Ambient noise level: 10~11 dBA
System noise level (on int. graphics): 12~13 dBA
Ambient temperature: 22°C

The load results were very impressive compared to previously tested high-end
graphics cards. The DirectCU II versions of the GTX 680 and HD 7870 were inferior
by a substantial amount both in terms of temperature and SPL. The noise difference
is especially staggering as the decibel scale is logarithmic. Equipped with
a sizable third party cooler, the GELID Icy Vision, the older HD 5870 and 6870
were also easily beaten out. These two cards aren’t that fast by today’s standards
but they still have hefty power draws, far exceeding that of the GTX 670.

The stark contrast between our results for the GTX 670 and HD 7870 DirectCU
II is particularly noteworthy considering the similarity of the two heatsinks
— they are virtually indistinguishable from one another. One wouldn’t expect
that a 27W reduction in power draw could end up making such a dramatic difference.
After examining the two coolers closely we realized there was another possible
explanation — the proportion of the GPU core making contact with the direct-touch
heatpipes. The Radeon core has a smaller die than the GTX 670 (212 vs. 294 square
millimeters) and it is positioned diagonally so only small portions of it are
touched by the outer heatpipes, making the center heatpipe responsible for transferring
most of the heat produced. The GeForce die’s larger footprint and “straight”
orientation seems to take greater advantage of this heatpipe layout.


The power consumption of an add-on video card can be estimated by comparing the total system power draw with and without the card installed. Our results were derived thus:

1. Power consumption of the graphics card at idle when Prime95 is run on a system, the video card is not stressed at all and stays idle. This is true whether the video card is integrated or an add-on PCIe 16X device. Hence, when the power consumption of the base system under Prime95 is subtracted from the power consumption of the same test with the graphics card installed, we obtain the increase in idle power of the add-on card over the integrated graphics chip.

2. Power consumption of the graphics card under load the power draw of the system is measured with the add-on video card, with Prime95 and FurMark running simultaneously. Then the power of the baseline system (with integrated graphics) running just Prime95 is subtracted. The difference is the load power of the add-on card. Any load on the CPU from FurMark should not skew the results, since the CPU was running at full load in both systems.

Both results are scaled by the efficiency of the power supply (tested here) to obtain a final estimate of the DC power consumption.

Note: The actual power of the add-on card cannot be derived using this method
because the integrated graphics may draw some power even when not in use. However,
the relative difference between the cards should be accurate.

According to our calculations, the GTX 670 DirectCU II consumed about 20W when idle and 162W on full synthetic load. Considering it’s only one step down from Nvidia’s flagship single GPU video card, the GTX 680, it’s a spectacular result.

In our video playback tests, the GTX 670’s lower idle draw carried over, allowing it to edge out GTX 680 by a few watts. In general, Nvidia’s performance models are substantially more energy efficient than AMD’s in this regard. We believe this is due to higher core and memory speeds being used when hardware acceleration is activated on the various members of the HD 7000, 6000, and 5000 series. During playback, the GTX 670’s core/memory clocks stayed the same as when idle, just 324/162 MHz.

All of the AMD/NVIDIA cards from the last three generations had very similar CPU usage during video playback, 1~2% for our 1080p H.264/MKV test clip, and 8~10% for our YouTube HD sample the GTX 670 was no exception.


The GTX 670 DirectCU II ships with ASUS’ GPU Tweak utility which offers a wide variety of fan, clock, and voltage adjustments as well as monitoring functionality. It bares more than a passive resemblance to MSI’s popular Afterburner application, albeit skinned with ASUS’ attractive red and black Republic of Gamers motif.

The user defined fan speed control could use some work however, offering stepped speeds rather than the gradual linear approach used by most comparable utilities. If you prefer the latter, applications like MSI Afterburner and Sapphire TriXX work well with the card despite being developed by different manufacturers.

MP3 Sound Recording

This recording was made with a high
resolution, lab quality, digital recording system
inside SPCR’s
own 11 dBA ambient anechoic chamber
, then converted to LAME 128kbps
encoded MP3s. It’s intended to give you an idea of how our test system sounds
in actual use one meter is a reasonable typical distance between a computer
or computer component and your ear. The recording contains stretches of ambient
noise that you can use to judge the relative loudness of the subject. Be aware
that very quiet subjects may not be audible if we couldn’t hear it from
one meter, chances are we couldn’t record it either!

The recording starts with 5~10 seconds of room ambiance, followed by 5~10 seconds
of the GPU test system with its case fans at various speeds. For the most realistic results, set the volume
so that the starting ambient level is just barely audible, then don’t change
the volume setting again.


According to credible gaming-oriented review sites like PC Perspective, HardwareCanucks and AnandTech, the GeForce GTX 670 is a high-end graphics card that can play most titles smoothly with high detail levels at the native resolutions of large, premium monitors (2560×1600, 2560×1440). It’s also one of the few single GPU cards that can tackle multi-monitor gaming to some degree. The general consensus is that it’s roughly 10% slower than the GTX 680 and trades blows with the Radeon HD 7970.

The GTX 670 is also fairly energy efficient, at least compared to other performance models, using about 20W when idle and 162W on load according to our estimates. The load figure is particularly impressive, approximately 40W less than the GTX 680. For most system configurations, a modest 500W power supply is all that’s required to accommodate the GTX 670 with room to spare. It’s also quite thrifty when accelerating HD video, as are Nvidia cards in general. Rendering H.264 and Flash video, it uses more than 20W less than the most recent AMD card we tested, the Radeon HD 7870.

The DirectCU II cooler is the cherry on top, making it the coolest and quietest
high-end graphics card we’ve ever tested that wasn’t using third party cooling.
In our low airflow video card test system, on full synthetic load, the GPU stabilized
at a comfortable 73°C with a total system noise level of just 16 dBA@1m.
Also keep in mind, the fan control can be customized to be less aggressive,
making it even quieter — there’s plenty of thermal headroom to do so.

Altogether, the ASUS GTX 670 DirectCU II is the complete package. With high
performance, modest power consumption, and a highly efficient cooling solution,
there’s very little room for improvement. The only thing that makes us hesitate
to recommend running out and purchasing one immediately is the price. Even after
excluding users who can’t fathom spending a hefty amount on a graphics, the
question of value is still in play. At US$400, it’s US$50 more
than an entry level GTX 670 with the same reference clock speeds. It’s also
notable that GTX 680’s can be found for not much more. Whether or not it’s worth
the cost really depends on whether value a superior heatsink with superbly low
noise. Puget Systems obviously thinks their customers are among those who do.
Good choice, Puget!

Our thanks to Puget
for the GTX 670 DirectCU II video card review sample.

ASUS GTX 670 DirectCU II wins the SPCR’s Editor’s Choice Award

* * *

Articles of Related Interest
ASUS Radeon HD 7870 DirectCU II

Prolimatech MK-26 Multi-VGA Cooler

ZOTAC GeForce GT 640 ZONE Edition Fanless GPU

SPCR’s 2012 Graphics Card/Cooler Test System

ASUS GeForce GTX 680 DirectCU II OC

Sapphire HD 7750 Ultimate Edition

* * *

Discuss this article in the
SPCR forums.

Leave a Comment

Your email address will not be published. Required fields are marked *