VGA Coolers: Thermalright V1 Ultra, Zalman 700 & 900, AC Silencer 5 v.3

High-end VGA Cards are a major source of heat, so most silencers know to avoid them if they want quiet. But, what about people who need a high end video card? Gamers can’t seem to do without them… and increasingly, they want quiet, too. That’s where aftermarket VGA coolers come in. Like CPU coolers, VGA coolers can be bigger, cooler, and quieter. So here is our long-awaited, much-retooled review of the biggest, coolest, and quietest VGA heatsinks on the market in a 4-way battle for the SPCR crown: Zalman VF900CU & VF700CU, Thermalright V1 Ultra and Arctic Cooling NV5 Silencer.

June 5, 2006 by Devon
Cooke

Even though some of today’s graphics cards get hotter than most CPUs, the variety of aftermarket coolers for graphics cards is
surprisingly limited. Large tower heatsinks have made it easy to cool even the
hottest processors with very little airflow, but the same cannot be said for
graphics cards. Building a graphics cooler is a challenge because many of the
conditions that allow a CPU to be well cooled do not apply to graphics cards.

For starters, the market for hot high-end graphics cards is a fraction of the
size of the CPU market. Graphics cards have less standardized mounting systems
for heatsinks, they cannot support as much weight, and they must deal with more
stringent space restrictions than a CPU. Motherboards are designed to accommodate
the cooling requirements of hot processors, but graphics cards are limited by
the layout of the expansion cards plus whatever wiggle room the manufacturers
think they can get away with. In practice, this has meant that good graphics
coolers nearly always protrude above and below the graphics card, potentially
causing compatibility problems and blocking the use of one or more expansion
slots.

Despite all of these challenges, a number of companies sell aftermarket coolers
that purport to be cooler, quieter, more space efficient, or all three at once
compared to the stock heatsink/fan modules. Some manufacturers have even adopted
their own versions of these aftermarket coolers instead of using the reference
designs from ATI and nVidia.

Over the last few months, a small pile of these aftermarket coolers has been
growing in the SPCR lab as we pondered about the challenges of testing them
effectively. With the development of our test bed for graphics cards a couple
months ago, we finally had a tool we could use to
separate the gems from the pretenders.

Our pile contained four different coolers:

• Arctic Cooling NV Silencer 5, rev. 3
• Thermalright V1 Ultra
• Zalman VF-700CU
• Zalman VF-900CU

These coolers use a variety of different approaches
to improve on stock cooling. There are two heatpipe-based designs, one that
exhausts waste heat outside the case, and a simple one that simply boasts of
better cooling. All of them were tested under the same thermal conditions in
our VGA test bed, described in on page 6.

Arctic Cooling was one of the very first companies to offer an aftermarket
VGA cooler called, appropriately enough, the VGA
Silencer. The NV Silencer 5 is the successor of the VGA Silencer and, as
its name suggests, it is designed specifically for nVidia-based cards. In fact,
it’s even more specific than that — the “5” in its name is there
to distinguish it from the other five nVidia-based versions that Arctic Cooling
produces. Arctic
Cooling’s product page has the complete details about which coolers go with
which graphics cards.

SPCR has reviewed the NV Silencer
5 before, so the product is not new to us. It was included in this round-up
for two reasons:

1. It makes a good reference against which to judge the other coolers in the
   review
2. The original review was of the noisy first revision of the cooler; Arctic
   Cooling released revision 2 with a quieter fan shortly after the review was
   published. (Revision 3 was added to allow compatibility with the GeForce 7800
   series and did not affect noise).

Plastic bubble packaging shows the main heatsink on the front with the bracket
and a secondary RAM heatsink on the back.

The NV Silencer is more than just a big heatsink with a quiet fan. One of the
reasons why VGA cards are so difficult to cool is that in the mid-tower style cases still favored by diehard enthusiasts, the heatsink typically
hangs below the card where hot air gets trapped and recirculated several
times before being exhausted. The AC Silencer line is special in that the
hot air from the heatsink is largely blown out the back of the case so it cannot be
recirculated.

A double width cooler that exhausts hot air out the back.

This idea is no longer unique — a number vidcard manufacturers use OEM variants of the Arctic Cooling — cooler. The basic idea is excellent especially
if the rest of the system also runs hot. The down side of the design is its
size and weight. Like many VGA coolers, it is too thick to fit into a single
slot, meaning that whatever slot is below the graphics card is unusable. At
428g, it is also the heavier than any of the other coolers in the roundup by
more than 100g.

Hot exhaust air gets blown out the back of the case instead of allowing it
to recirculate.

The GPU and the RAM chips on the front of the card are all cooled.

Installation

Because the NV Silencer 5 was designed for a limited range of cards, installing it is as simple
as peeling off the protective paper on the heat pads, threading the bolts through
the holes on the card, and screwing on the backplate. The entire process took
less than five minutes. (On the other hand, removing the stock cooler took a
good half hour.)

The fan is connected to two different headers that match the ones found on
the most common nVidia cards. However, Arctic Cooling notes that “Asus
is unfortunately not following ATIs/NVIDIAs solutions and therefore our plug
doesn’t fit.” They recommend removing the plugs and connecting the
wires directly in this case. Unlike the other coolers in this review, a motherboard-compatible
header is not provided.

A cross-shaped backplate holds the cooler securely in place.
The rubber mat prevents short-circuits and damage to the delicate electronics
underneath.

The last step is optional but highly recommended: A separate aluminum plate
provides additional cooling for any RAM chips on the back side of the card.
It also draws some heat through the PCB, so it is useful even on cards without
rear-mounted RAM chips.

Thermalright is well known for making high-end CPU heatsinks, so it is no surprise
that they have an offering for VGA cards. The V1 Ultra is the successor of the
original V1 cooler, which was rendered obsolete when ATI and nVidia released
their latest generation of GPUs. Aside from fixing compatibility, the V1 Ultra
also adds a third heatpipe.

Thermalright has dealt with the space constraints around the VGA card by relocating
the most bulky parts. A medium-sized passive heatsink provides a small amount
of direct cooling, but the bulk of the heat is transferred to a smaller heatsink
that is suspended 6mm above the back of the card. A low profile 80 mm fan dissipates
heat away from the heatsink. For cards that require extra cooling, an second
fan can be attached to the passive heatsink.

All these heatsinks and fans make the cooler quite thick. Like the NV Silencer,
the heatsink protrudes into the slot below the card, preventing its use. The
fan and heatsink on the top also add take up space equivalent to another two
slots, rendering any slots above the card useless as well. The V1 Ultra
is unlikely to work well in a dual VGA configuration, since even if the fan
can be forced to fit, the fan on the lower card will likely end up blowing all
of the heat from the upper card down onto the lower card.

Thermalright’s signature brown box and a wide gamut of bits and pieces.

Individual components add up: The V1 Ultra is exceptionally thick.

Passive base in the foreground, main heatsink in the background.

A low profile 80mm fan is included, but a standard fan can be used if there’s
room.

A trio of heatpipes move heat away from the GPU.

Installation

Installing the V1 Ultra was much more complex than the NV Silencer. The cooler
was designed to work with a wide range of cards, so instead of fixed position bolts, it uses
a bracket with adjustable arms that can adapt to various configurations of mounting
holes. The system is somewhat similar to those used in some northbridge heatsinks
(Zalman’s come to mind).

There were two main steps and numerous intermediate steps in the installation.
Step one: The bracket was screwed onto the card in preparation for step two:
The heatsink was screwed onto the bracket. The difficult part was the first
step, which was imprecise and required a little experimentation. And, despite
an instruction booklet that is well illustrated and looks complete, there were
a number of small details that needed to be figured out on their own.

Problems arose from the very beginning, starting with the cushion pad that
is intended to protect the delicate corners of the GPU while maneuvering the
heatsink into place. Unfortunately, the cushion was too small to fit around
the twin chips on our card; had we left it in place the heatsink would not have
made good contact. In addition, the metal frame around the chip did not allow
the cushion to lie flat. Oddly, the cushion didn’t seem to be very well matched
with the mounting bracket either, as two of the edges would not have provided
any cushioning at all. Eventually, we gave up and installed the cooler without
the cushion.

This soft silicone pad is designed to protect the GPU chip, but it was too
small to fit our card.

The next challenge was to install the bracket. Unlike the NV Silencer, which
used all four of the available mounting holes, the V1 Ultra used only two flimsy
machine screws to hold the bracket in place. In fact, the screws weren’t even
attached to the bracket itself; instead, they screwed into metal arms that were
then attached to the bracket. The illustration below shows the correct installation.
Tension on the bracket is controlled by the tightness of the machine screws,
which bend the arms as they are tightened.

The arms themselves are worthy of note. Thermalright includes two sets of arms,
a long set and a short set. Contrary to the advice in the instruction booklet,
we found it best to use the longer arms to avoid warping the bracket. Although we were able to complete the installation with the
short arms, they did not allow the bracket to lie flat under tension.

The fully installed bracket felt quite flimsy, and seemed unlikely to hold
the cooler in place if the card is dropped or jostled in transit. The individual
screws and the threads on arms also seemed less secure than they should have
been.

The bracket, fully installed.

The next step was to maneuver the heatsink into place. Thankfully, this was
much easier than installing the bracket, although it too required several intermediate
steps. First of all, the RAM chips had to be cooled. Thermalright includes six
individual heatsinks for the RAM chips which were easily applied by peeling
off the protective backing and sticking them in place. However, our card, like
most others, shipped with eight RAM chips, not six.

This was no accident; the remaining two chips are cooled by the main heatsink
via an extra-thick heatpad. Two 1mm pads and a 2mm pad are included so that
the height of the pad can be adjusted from 1~4mm. For our card, the single 2mm
pad was adequate. The pad needed to be applied before we slid the heatsink onto
the card.

The heatpipes extend well above the top edge of the card.

The main heatsink would have slipped easily over the end of the card were it
not for the aluminum backplate from the NV Silencer. This backplate was left
in place for all of the coolers, since the NV Silencer was the only model to
ship with a rear-mounted heatsink. As it was, it took a little force to move
it into position but everything worked out in the end. The heatsink was secured
to the bracket with four hefty machine screws that pulled the bracket up so
it was snug against the heatsink, adding some much needed tension. The final
result seemed tight and secure, whatever our worries about the security of the
bracket.

The back side. The aluminum plate is left over from the NV Silencer, and
does not come with the V1 Ultra.

The VF700-CU is basically one of Zalman’s signature flower heatsinks shrunk
down to fit in the space of two expansion slots. Our sample is an all copper
version with a blue LED fan, but there are also versions that use a conventional
fan or a combination of aluminum and copper. There’s even a bright red version
sponsored by world renowned (apparently — I’ve never heard of him) gamer,
Fatal1ty.

Like the V1 Ultra, the VF700 is meant to be a one size fits all design that
supports the majority of the cards on the market. Zalman maintains an extensive
list of compatible
cards on their web site. Or, if you want to attack the problem in the opposite
direction, they also have a list of incompatible
cards, mostly of older vintages.

The VF700-CU is a simpler design than both the V1 Ultra and NV Silencer because
it doesn’t try to move the heat around with heatpipes or a plastic duct. It’s
closer to a conventional stock heatsink than either of the other two heatsinks
we’ve seen so far. Because of this, it is smaller and lighter than both of them.
Even so, it is still thick enough that it occupies the PCI slot below the card.

Zalman’s flower design comes to a VGA cooler.

Zalman has always made a point of catering to low-noise enthusiasts, and the
VF700 is no exception. Included in the package is a Molex to 3-pin adapter that
supports two 12V and two 5V connections. This is not quite as good as the variable
fan controller that Zalman ships with some of its heatsinks, but the gesture
is appreciated even so. The fan uses a standard 3-pin motherboard plug, so finding
ways to control the fan speed shouldn’t be a problem.

Extras include a simple 12V/5V fan header and 8 tiny RAM heatsinks.

A classic Zalman flower.

One thing that sets the VF700 apart from other Zalman coolers is its asymmetrical
design. The reason for this seems to be compatibility: A simple fully circular design with would be unlikely to fit any of the cards on the market.
Our card in particular only allowed about an inch of space between the bottom
of the GPU and the PCI Express slot. With the cooler centered over the chip,
there is simply no room for the fins to extend downwards very far. Zalman’s solution places the base of heatsink off-center so that the fins extend in a larger arc above the GPU and a smaller arc below. The squashed egg shape works to ensure physical fit with most graphics cards.

This side has shorter fins that aren’t likely to bottom out when the card
is installed.

The base is polished to a shine.

Installation

Installing the VF700 was almost as simple as the NV Silencer. The mounting
system is similar to some of Zalman’s past heatsinks: A metal harness that screws
onto two sturdy mounting posts.

The first step is to install the eight RAM heatsinks. These are smaller than
the heatsinks included with the V1 Ultra, but the installation procedure is
identical: Peel off the backing and stick them on the RAM chips. The rear heatsink
from the NV Silencer was again left on as a precautionary measure.

The most complicated part of the installation was putting together these mounting
posts. Each is built from a pair of metal sleeves that screw together, sandwiching
the graphics card between them. Rubber washers need to be placed on either side
of the card to prevent short-circuits. A well illustrated instruction manual
and a
very helpful flash animation illustrate the procedure well.

In progress. On the left is a fully installed mounting post.
On the right are the four pieces that make it up.

A brace is then screwed onto the back side of the card, presumably to give
the mounting posts some stability. This adds a little bit of height to the overall
installation and could probably be safely removed if it interferes with any
components above the card (a tall northbridge heatsink, for example).

The back brace is simple and secure.

Once the mounting posts are firmly secured, it is but the work of a minute
to screw the heatsink in place. The screws need to be tightened gradually, alternating
from one to the other, to avoid crushing the corners of the GPU.

Mounting posts are provided.

That done, all that’s left is to plug in the fan, preferably into
one of the 5V headers. Enjoy!

Installed and glowing.

Moving up the price range, the Zalman VF900 does what the VF700 could not:
It closes the flower to make a fully circular cooler. Despite being both smaller
and lighter than any other heatsink in this review, including the VF700, Zalman
positions it as their high-end VGA cooler and has priced it accordingly.
The difference? A pair of heatpipes that snake around the outside edge of the
flower.

Like the VF700, the VF900 is relatively simple and intended for low noise operation.
A Zalman Fanmate2 is included in the package, so the fan can be set to quite
a wide range of speeds. That aside, it is fairly similar to its cheaper cousin:
The fan is located directly over the GPU, with no attempt to move the heat around.
It is up to the system airflow to move the heat out of the system once it has
come off the heatsink.

A full flower… how romantic (your girlfriend won’t think so).

Even though the heatsink is made of copper, it is by far the lightest cooler
of the bunch, coming in at just 185g — almost 100g less than the VF700
and less than half of the NV Silencer. It is also the smallest cooler, although
it still takes up an extra PCI slot.

By this time, we were starting to scratch our heads and wonder if Zalman had
reversed the roles of their two coolers by accident. Surely this tiny thing
couldn’t be Zalman’s performance cooler… could it? Of course, the true
test is how it stands up under testing, not how it looks.

A Zalman Fanmate2 and the same RAM heatsinks as the VF700.

Look closely: There’s actually two heatpipes here, one for each half.

Although it looks similar, the VF900 actually shares more in common with the
CNPS9500 CPU heatsink than
the VF700. The difference is the way that the flower of fins is constructed.
The VF700 consists of a large number of fins that are clamped together to make
the base and then spread apart at the edges to make the flower, but the VF900
does things differently. It relies on the ring of heatpipes to give it its structure
and is hollow in the middle. The base plays no role other than to serve as a
transfer point between the GPU and the heatpipes.

The heatpipes are more visible on this side.

Smooth contact surface for the GPU.

Installation

The VF900 is about as easy to install as the NV Silencer, but it manages to
retain compatibility with a wide range of cards while doing so. The incompatibility
list for the VF900 is short and sweet, with the GeForce 7800GS being the only
current-generation card that will not work. Zalman deserves extra kudos for
making the installation process 100% tool free.

Amazingly, Zalman has managed to make the VF900 compatible with no less than
five mounting hole patterns without resorting to a bracket like the one used
by the V1 Ultra. Installation is as simple as screwing the mounting posts into
the correct holes on the heatsink, threading them through the card, and tightening
the nuts on the other side. Just make sure that washers are used to insulate
the card from the metal hardware.

The mounting posts are attached to the heatsink itself…

…the card is placed on top, and hand-tightened nuts hold the heatsink in
place.

Voilà.

TEST METHODOLOGY

This sturdy modified LX-6A19 (D8000) case from Cool Cases housed our test
system.

Our test procedure is an in-system test, designed to 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 or reboot without warning.
• 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.

The test system was built around the coolest Intel Prescott-based processor
we had on hand. It’s a reasonable example of a mid-powered system that is fairly
easy to keep quiet. Almost all AMD-based processors should run cooler than our
Pentium 520, and almost all Intel-based desktop processors run hotter.

Test Platform

• Intel
  520 processor (P4-2.8 Prescott, 1Mb cache, 800 MHz FSB in 775 casing).
  The combined power draw for the processor and the VRMs on the test motherboard
  was measured at 85.3 watts. Intel’s official TDP is 84 watts.
• AOpen
  i945Ga-PHS motherboard – Intel i945Ga Chipset; built-in VGA.
• CoolerMaster Hyper 48
  heatsink, cooled by a Nexus 92mm fan undervolted to 7V.
• Corsair
  DDR2 RAM, 1024 MB
• Seagate Momentus 5400.2
  120 GB, 2-platter drive, suspended in a NoVibes
  III just inside the front intake vent.
• Antec Neo HE 430
  ATX12V 2.01 compliant power supply, with a custom-built fresh air duct to
  ensure that the internal fan did not ramp up during testing.
• Modified case from Cool
  Cases, outlined in detail below.
• Nexus 120mm
  fan controlled by a variable voltage fan controller.

Measurement and Analysis Tools

• CPUBurn
  processor stress software
• RTHDRIBL
  graphics demo to stress the GPU
• ATI Tool
  v0.24 as an alternate tool for stressing the GPU.
• SpeedFan
  version 4.28 to show CPU and system temperature
• NVTempLogger
  v3.37 to show GPU temperature
• Seasonic
  Power Angel AC power meter, used to measure the power consumption
  of the system
• A custom-built variable DC power supply to power the system fan

System airflow is quite good, allowing the CPU and system fans to run at close
to inaudible speeds without compromising system cooling. The intake is about
the size of a 120mm fan. The only restriction is an air filter. A much more
restrictive cover for the filter was removed because it impeded the airflow
too much.

The one and only intake…

…and the same view, with the bezel removed.

There is only one point of exhaust: The 120mm case fan, which can be run at
a number of different speeds. The 80mm fan in the Neo HE power supply was taken
out of the picture by using a custom-built duct to ensure that the fan never
ramped up. The amount of airflow through the system can be controlled by adjusting
the speed of the case fan, thereby giving us a way of controlling how difficult
the thermal environment inside the case is.

A fresh air duct isolates the power supply from the rest of the system.

Only one possible point of exhaust: The orange case fan.
The fan in the power supply draws its air from a duct that does not interact
with the rest of the system airflow.

The airflow in our test rig is typical of an ATX case. Air flows in through
the intake near the bottom of the front panel, and is pulled up to the top rear
corner. Most of this air will bypass the expansion cards altogether, but a small
amount will be pulled across the rear of the card as it is pulled towards the
CPU heatsink and the case fan. All of the air will exit the case via the exhaust
fan.

The air will flow from the lower right to the upper left, drawing a small
amount of air across the VGA card.

The coolers were installed on an AOpen
Aeolus PCX6800GT-DVD256 — a medium-powered video card that features
a thermal sensor build into the GPU core. The accuracy of the sensor is unknown,
but it is good enough to make relative judgments between the various coolers.

This GeForce 6800GT has stood up to a year’s worth of abuse around the SPCR
lab.

TEST RESULTS

Ambient conditions during testing were 19 dBA and 22~23°C. Tests were run
with the system fan at 12V, 9V, and then 7V. An additional test was run with
both the system fan and the VGA fan at a full 12V as an estimate of the best
possible cooling, regardless of noise.

Thermal testing consisted of running CPUBurn and the artifact scanner
built into ATI Tool simultaneously to generate as much heat as possible.
An initial test was run with the system fan running at 12 volts, and then the
fan was progressively slowed down to make the thermal environment more difficult.
As a general rule, the fan on the VGA cooler was kept at a quiet 5 volts, with
the single exception of the NV Silencer, which was powered by the fan header
on the test card. Based on the additional noise in the system, we would estimate
that the fan received approximately 10V from the header. The noise level did
not change appreciably during testing, so we can assume that the fan did not
change significantly in speed.

Once the temperature on the card had stabilized, the stress software was left
running for at least another 20 minutes while we watched the screen carefully
for visual artifacts that might indicate overheating. The last test, with the
system fan running at 7 volts, was left running for more than an hour. Even
during our most strenuous test, none of the coolers ever allowed our test card
to get hot enough that there were visual artifacts on the screen.

TEST 1: System Fan @ 12V, VGA Fan @ 12V

At full blast, all of the coolers were more than capable of cooling our test
card. The Thermalright V1 Ultra and the VF900CU were more or less tied for the
best cooler, with the Thermalright delivering a slightly better core temperature
and the Zalman giving a better “ambient” temperature. The location
of the ambient sensor is unknown, so the result probably has more to do with
the location of the airflow than any absolute difference in cooling ability.
As a side note, the top mounted fan of the V1 Ultra provided a little bit of
extra cooling for the CPU.

On the other hand, the NV Silencer, which provides no airflow
that isn’t around the GPU, produced a much higher ambient temperature than any
other heatsink. It seems that the secondary electronics (RAM, VRMs, etc.) are
better cooled with the unducted coolers.

All of the coolers were too loud to consider in a quiet system, although the
NV Silencer might be on the borderline. In all cases, the primary source of
noise was the VGA fan, not the rest of the system.

TEST 2: System Fan @ 12V, VGA Fan @ 5V

What a different view from the quiet end of the spectrum! This time, the V1
Ultra was the worst performer and the VF900… hadn’t changed. Sure, the ambient
temperature bumped up a few degrees, but the actual core temperature didn’t
change. Shocked, we filed the result away for closer examination later.

The VF700 also did quite well with the fan turned down. GPU temperature crept
up by just two degrees, putting it at roughly the same level as the V1 Ultra
— at least where core temperature is concerned.

This time around, the mysterious “ambient” sensor was worth paying
attention to. While the two Zalman coolers kept the increase in “ambient”
to a respectable 3~4°C, the V1 Ultra skyrocketed by 15°C! Wherever the
sensor is, it is quite clear that the Zalman coolers do a better job of cooling
it with low airflow than the V1 Ultra.

The results for the NV Silencer didn’t change — hardly a surprise because
we had no way of controlling the fan speed on the cooler, so the test that we
ran was identical. The results are included for comparison only.

The all of the VGA fans at 5V, the primary source of noise was no longer the
VGA fans but the system fan. Turning down the case fan was the logical next
step towards silencing the test system.

TEST 3: System Fan @ 9V, VGA Fan @ 5V

The NV Silencer and the V1 Ultra were more or less unaffected by the change
in system airflow. With either cooler, the temperatures on the video card remained
constant, although the CPU temperature crept further towards throttling.

The VF900 was only slightly more affected; it allowed the GPU core to creep
up by a couple of degrees. Only the VF700 was affected seriously by the reduction
in system airflow, once again dropping a little behind the V1 Ultra.

By this time, the NV Silencer was looking pretty good. Only the VF900 managed
to beat it in in terms of performance. But, (and it’s a big but) it was also
still the main source of noise in the system at this point, since its fan was
still running at close to full tilt. From a noise-for-performance perspective,
it was still less than satisfactory.

TEST 4:System Fan @ 7V, VGA Fan @ 5V

Reducing the system fan to 7V was the last step in our test. Reducing it any
more would have been pointless from an acoustic point of view, since the system
noise did not really change below this level. The residual noise level of 23
dBA@1m was a combination of all of the noise sources in the system, with none
distinguishing themselves clearly as the primary noise source. Further noise
reduction at this point would have needed to take a more radical form that just
reducing fan speeds.

The final test revealed little new information except to confirm what we had
already learned: The VF900 was clearly the best low airflow cooler in our test. The maximum GPU temperature of 67°C was at least 10°C better
than both the V1 Ultra and the VF700. It would undoubtedly have had a similar
advantage over the NV Silencer if we had been able to reduce the latter’s fan speed.

TEST 5: Experiments with the VF900CU

Much of the VF900’s advantage seemed to come from the fact that the speed of
its fan seemed to have very little effect on its cooling efficiency. When the
coolers were compared with the fans at full speed, the VF900 was good, but the
V1 Ultra was slightly better. However, when the fans were turned down, the VF900
was clearly better because its effectiveness simply didn’t change. In contrast,
the V1 Ultra ran much hotter under low airflow. Why? Re-testing proved that
the initial results were not an anomaly, but how was it possible that the speed
of the fan simply had no effect?

To answer this question, we decided to modify the airflow in our test rig.
A 92mm fan vent on the side panel — blocked off during ordinary testing
— proved to be the modification we were looking for. The vent was located
more or less directly above the expansion slots on the motherboard, slightly
below the top slot the was occupied by our test card. Unblocking the vent provided
a new source of fresh air — and an alternate system airflow which we could
use to investigate the VF900 further.

Two tests were run with this side vent unblocked: One with the VGA fan at 12V
and one at 5V. The system fan remained at 12V throughout the test so that we
could compare the results with our earlier tests.

With the side vent unblocked, the performance was more in line with our expectations.
There was a 6°C spread between the fan at 12V and the fan at 5V — not
large, but enough to let us know that the VF900 can be affected by fan speed.

What is more interesting is that, while the source of fresh air proved beneficial
when the fan was running full tilt, the temperature was actually higher than
any other test when the fan speed was reduced to 5V. In addition, the CPU temperature
was significantly higher whenever the side vent was unblocked. For a low noise
system, the original configuration was clearly a little better.

FAN NOISE RECORDINGS

By and large, the measured noise levels for the stock fans were all quite similar
at a given voltage level. Not surprisingly, the two Zalman fans were very close,
with the V1 Ultra being slightly louder and the NV Silencer being slightly quieter
across the whole range of voltages. However, these differences were quite minor.
All were just a little too loud at full tilt, and all were close to the ambient
noise level at 5V.

It was the quality of noise that distinguished the fans from each other, not
their volume. In this department, the two Zalman fans were clearly superior,
although the VF700 with its clear LED fan was just a tad louder. However, they
both sounded quite good when undervolted. MikeC described the noise aptly as
“a bit like a whispery rubbing of paper”. This is a pleasant change
for Zalman, whose past fans have left a lot to be desired with their low voltage
noise. This noise character persisted even when the fan voltage was increased;
neither fan developed much whine until above 9V.

Compared to the Zalmans, neither the NV Silencer nor the V1 Ultra sounded very
nice, although the Revision 3 NV Silencer was considerably improved over the
original version. Between the two, the NV Silencer was nicer, but only because
it was quieter. Both had a pure tone that could be heard even at 5V. The whine
on the V1 Ultra was especially obnoxious because it was so easy to pick out
the specific pitch of the noise. The NV Silencer had a different problem: It
produced a distinct “chugging” that was noticeable because of its
rhythmic quality. Only when the fan was at 12V did the chugging dissipate into
a noisy clatter.

CONCLUSIONS

At the end of our roundup, Zalman has the advantage with their
simple designs, low noise, and excellent performance. The VF900CU in particular
had a very strong showing. It was quieter than any other heatsink with the possible
exception of the its brother, the VF700CU. And, when the fans were undervolted,
it also performed better than any other heatsink by a wide margin. At US$40,
it was one of the pricier heatsinks that we tested, but it’s still significantly
cheaper than many high performance CPU heatsinks.

It’s also cheaper than the Thermalright V1 Ultra which, although
it did better than any other heatsink with the fan at full speed, did very poorly
when undervolted. Its unusual back-mounted fan also raised questions about how
well the secondary components on the card were cooled, as the “ambient”
temperature reported by our test card was often much higher than the other heatsinks.
The large size and complex installation are also disadvantages when compared
against Zalman’s offerings.

The Arctic Cooling NV Silencer 5 was thrown into the mix as more
of a reference than a serious contender — we’ve reviewed it before. It
is costs the least of our contenders, and, although its new fan sounds much
better than the old one, it is still comparatively lousy. The fact that a number
of existing cards already ship with some version of the NV Silencer installed
means that buying it separately is probably not the best way of getting one.
The NV Silencer is beginning to look a little dated; we look forward to examining
their new Accelero line of VGA coolers sometime in the future.

Last but not least, the VF700CU LED is a good pick for a low or
midrange card that doesn’t need the extra oomph of the VF900CU. It can be found
for as little as US$25 online, and boasts a similar noise character to its bigger
brother. It also shares another important characteristic: It seemed to be affected
very little by the speed of the fan, which means that it can be undervolted
very easily.

As it stands, both of the Zalman heatsinks are viable replacements
for a noisy stock cooler. While the market for quiet CPU heatsinks has become
quite crowded of late, where VGA heatsinks are concerned, Zalman stands where
it did a few years ago with its CNPS7000 series of coolers: There are other
competitors on the market, but none is quite as quiet or quite as cool as Zalman’s.

Many thanks to Arctic
Cooling for the sample of the NV Silencer 5 (rev. 3)
to Zalman for the
samples of the VF900CU and the VF700CU LED,
to Thermalright
for the sample of the V1 Ultra,
to AOpen for the
samples of the VGA card and the motherboard in the testbed,
and to Newegg for
the Intel 520 processor.

*

SPCR Articles of Related Interest:
Fanless PCIe Graphics Cards
from Asus and Aopen
Gigabyte GV-N66256DP Fanless AGP video card
Chaintech AA6800GT
+ Arctic Cooling NV Silencer 5
Arctic Cooling ATI Silencer
2 VGA Cooler
Arctic Cooling VGA Silencer

Zalman ZM-80 VGA Heatpipe
Cooler

* * *

Discuss this article in the SPCR Forums.