Arctic Cooling Freezer 7 Pro HSF

Our favorite low cost quiet heatsink/fan maker has a new flagship with some unusual and innovative heatures. Multiple copper heatpipes and thin fins, of course, as in the original Freezer, and an improved “frameless” fan mounted with miniature soft rubber grommets. It’s quite the cooling performer. What about noise? Read on to find out about Arctic Cooling’s latest.

November 28, 2005 by Mike Chin and Devon
Cooke

Arctic Cooling fits the stereotype of a European engineering company very well.
For the most part, their products are smart implementations of elegantly simple
ideas, such as their Super
Silent series of heatsinks or their
VGA Silencer. Unlike most famous European brands, Arctic
Cooling doesn’t go after the high end market. Most of their products are modestly
priced, and, even if they’re not the top performers in their class, tend to be
very good value. In this respect they’re a bit like IKEA:
Cheap but usually nice and a bit quirky.

The Freezer
series served as their “high end” model, the low ~US$35
price notwithstanding. Even though it was Arctic Cooling’s best CPU cooler, it was no match for the latest and greatest from Scythe, Thermalright and Zalman, which may be why Arctic Cooling
released a new version called the Freezer Pro. It features
a larger fins, an extra heatpipe, and a different, soft-mounted fan. The soft-mounted fan is especially interesting because it shows how seriously Arctic
Cooling considers noise in their design process ? not just their
marketing materials.

The package is small and mostly monochromatic. Classy.

The Freezer Pro comes in two variants: The Freezer 7 Pro and the Freezer
64 Pro. The main difference between the two is the mounting system. The Freezer
7 Pro is compatible with Socket 775 processors (Intel), and the Freezer 64 Pro
with Socket 754/939/940 processors (AMD). According to Arctic
Cooling’s datasheet, the two models also feature slightly different fans.
This makes sense, since most Intel processors produce more heat ? and thus
require more airflow ? than AMD’s models. Our sample is a Freezer 7 Pro,
which comes with a 4-pin fan that is a little faster than the 3-pin fan that
ships with the Freezer 64 Pro. No samples of the Freezer 64 were available in early November.

A fitted plastic casing protects the Freezer 7 Pro in transit.
The only other things included in the box are an instruction sheet and a small,
branded sticker.

DETAILS

The Freezer 7 Pro is built around three U-shaped heatpipes (for 6 heatpipe paths) and
a stack of 42 aluminum fins, which is fairly typical of a tower-style heatsink.
The heatpipes receive heat from the CPU at the bottom of the U and transfer
it up to the ends where it is dissipated into the aluminum fins and then eventually
into the surrounding air. The individual heatpipes are located quite close together. This is bit surprising; distributing them farther apart would seem better for heat dissipation.

The fins are large and tightly spaced, so the total surface
area is high. With unlimited airflow, this would be a good thing, but
the close spacing of the fins could hurt cooling efficiency in low airflow conditions.

The 92mm fan is rated for 0.16A at 12V, providing 45 CFM. The latter number seems very high for the low power rating, but the noise and the perceived airflow aross our fingertips suggest it’s accurate. We tried to measure the airflow of the fan, both on the HS and also removed from the heatsink. The results from our anemometer testing were lower than the 45CFM claimed. but they were too inconsistent to be relied upon. This may be due to the absecne of a frame. Barring evidence to the contrary, we have no choice but to accept their claim for now.

The total mass of 520 grams is very modest for a high performance heatsink/fan, exceeding Intel and AMD recommendations by only 70 grams. Many high end heatsinks are far weightier, often approaching a whole kg, which does not bode well for a vertically mounted motherboard, especially with tall heatsinks that apply a lot of cantilever force on the mounting points and the board. Nothing like that to worry about here.

Just 520g despite the large size.

Tower heatsinks like the Freezer 7 Pro provide very little airflow across the motherboard
itself. This is a problem, as motherboard makers expect the CPU heatsink fan to provide cooling
for the voltage regulators (often called voltage regulator module or VRM) around the CPU socket. The illustration below (from
the retail box) shows how Arctic Cooling intends to solve this problem.

The three bottom fins direct airflow downwards towards the motherboard.

The gist of the solution is to deflect a little bit of airflow down towards
the motherboard by bending the bottom three fins and spacing them more widely.
The frameless design of the fan should also help draw air across the motherboard.
Because the bottom three fins are barely in the path of the fan, this solution
is unlikely to provide a lot of airflow, but it should be enough to prevent
pockets of dead air from forming under the heatsink.

The most unusual part of the Freezer 7 Pro is the semi-frameless 92 mm fan. One of our complaints
about the original Freezer was how fragile its fan was with only two spindly supporting arms for a frame. In fact, one of our original Freezer samples was damaged in transit, which surely must have occurred much more than with typical HSF and perhaps spurred development of this new model. The fan here is still missing most of its
frame, and this means it is not as strong as a regular “box” fan. The only thing
preventing the fan from catching a stray wire are two thin plastic rings that
surround the blades. This provides a minimum of safety while keeping the
frame as open as possible. Unlike the fan on the original Freezer which used four tiny screws, this one can be easily taken off and put back on without any tools.

For some reason, the fan frame is ever-so-slightly oval shaped.
Two plastic tabs keep the fan clipped in grooves on two side of the fins.

Oddly, the minimal frame is not quite circular; it’s a little wider than it is tall,
giving it a slightly oval shape. Given how well the frame fits on the fins,
this was probably done to meet the geometry of the heatsink itself.

The fan frame is slight oval.

A frameless fan usually generates a little less turbulence
than a standard fan. However, we suspect it also generates less directional airflow, which may make it
more difficult to force air through tightly spaced fins. The Zalman 7000, 7700 and 9500 series heatsinks also use frameless fans, but they are significantly different from Arctic Cooling’s HS fans. In the Zalman models, the fan is nearly surrounded by the fins, it fits into a recess in the fins. This suggests that all the kinetic energy of the spinning blades is converted into some kind of airflow through the fins, which would improve cooling efficiency. The Arctic Cooling heatsinks, in contrast, position the frameless (or near-frameless) fan conventionally. It reduces turbulence, but the energy at the tips of the blades may not necessarily be converted into airflow towards the fins.

RUBBER GROMMETS

The fan is decoupled from the frame by four rubber grommets. This is
a selling point for Arctic Cooling that’s well described in the
marketing literature. Their purpose is to reduce the amount of fan vibration
that is transferred to the heatsink and the motherboard, thereby reducing noise.

How significant is this? It is worthy of note. Only serious silencers are likely to notice it much,
and the difference is likely to be subtle. But, in a properly built
system it could be the difference between a quiet system and a silent one.

Tiny rubber grommets attach the fan to the frame…

…so that the fan hangs from the frame instead of being firmly attached.

INSTALLATION

Arctic Cooling provides
a single bilingual sheet of instructions with the Freezer 7 Pro (German
and English). There are seven steps to cover everything from
removal of the old heatsink to the correct orientation of the heatsink to the
actual installation.

The mounting system is the same as used by Intel’s stock heatsink, so mounting
the Freezer 7 Pro is fairly simple. Four white plastic fasteners are positioned
at the four holes around the CPU socket on the motherboard. The black plastic head has to be in the most clockwise position. Then, each of the black plastic heads is pushed down to lock it in place.

It is important to make sure that each of the four pins is engaged tightly, and to the same depth. It is possible to have one or more of the locking shafts be only halfway engaged, yet still look and feel reasonably tight. If this happens, the base of the HS will not be sitting with even pressure on the CPU, and cooling will suffer.

The heatsink can be mounted in any orientation, so it should always be possible
to set it up correctly with the fan blowing towards the rear of the case. Arctic
Cooling recommends removing the fan before installation, then putting it back on after the HS is engaged on the motherboard.

The base is copper, and comes with thermal interface material pre-applied.
With the TIM removed, the base is flat and quite smooth, though not polished.

TESTING

The Freezer Pro series is not available for socket 478, which means that the
test bench that we have used for the past year could not be used. A socket 775 test
platform was put together using the parts outlined
below. The data from this review cannot be directly
compared to the data from past reviews. It also means the test has become a bit tougher
because a hotter processor is used. The Pentium 520 used in this test is about
the coolest Intel P4 on the market today, but it is still 15-20W hotter than our previous
P4-2.8 Northwood.

To establish some references on this test platform with heatsinks that we have already tested and reviewed, we took time to install and run stress tests on the Thermalright XP-120 and the Scythe Ninja, both top performers, as well as a standard Intel socket 775 HSF. This is the cooler that’s packaged with Intel’s retail processors.

Measurements of the total AC power consumed by the system during the load tests were also recorded, as a check on Arctic Cooling’s claim about improved cooling for the VRM. We know that improved cooling for the VRM results in slightly higher efficiency and lower power draw.

On the test bench…

Test Platform

• Intel
  520 processor (P4-2.8 Prescott, 1Mb cache, 800 MHz FSB in 775 casing).
  The Thermal Design Power is 84.0W. The Maximum Power,
  as calculated by CPUHeat
  & CPUMSR, is 100.76W.
• AOpen
  i915Ga-PLF motherboard – Intel 915G Chipset; built-in VGA. Reviewed in SPCR.
• OCZ DDRAM PC-4000, 512 MB
• Fujitsu
  MHT2080BH 80GB notebook drive (in Smart
  Drive)
• Seasonic Super Silencer 400 power supply
• Arctic Silver
  Ceramique Thermal Compound
• Two-level open platform with foam damping feet. Motherboard on
  top; most other components below. Eases heatsink changes and setup.

Measurement & Analysis Tools

• CPUBurn
  processor stress software
• SpeedFan
  version 4.27 software to show CPU temperature
• A custom-built variable DC power supply that allows us to dial in exactly what voltage
  is powering the fan
• B&K model 1613 sound level meter
• Seasonic Power Angel AC Power meter
• Electronic Anemometer, to measure airflow
• High resolution digital audio recording system

Noise measurements were made with the fan powered from the lab DC power supply with everything else turned off to ensure minimal ambient noise.

Airflow measurements for this heatsink were not made due to the difficulty
of measuring the stock fan accurately. Instead, RPM measurements were
made using the onboard monitoring chip and a Zalman
Fanmate to vary the input voltage.

Load testing was accomplished using CPUBurn to stress the processor, and the
graph function in SpeedFan was used to make sure that the load temperature was
stable for at least ten minutes.

The ambient conditions during testing were 17 dBA and 21°C.

TEST RESULTS

Fan @ 12V: Cooling performance was excellent. A rise over ambient of just 18°C on a CPU that’s putting out as much as 100W is pretty good. Whether you use the 100W or 84W rating, the °C/W numbers are very good.

The noise level was oddly variable. The variability came from
air turbulence, but it didn’t have the more-or-less constant whoosh that most
fans exhibit. Instead, it seemed to go up and down in surges, sounding smoother
one moment and “stormier” the next, sometimes a bit randomly, and other times in a more regular, cyclical rhythm. Depending on who is listening, this variable effect can be described as either obvious or subtle. (The effect can be heard somewhat in the audio recording at 12V on the next page.) From a meter distance in open air, it could be heard easily in a quiet room, although the overall volume may be more annoying than its variability. The base SPL of 33 dBA@1m is already well above our 30 dBA maximum SPL for a quiet product.

The hum
of the motor sounded lower in pitch when there was less turbulence
noise. The volume of the noise was reasonable for a fan at full speed, but the
variability of the noise emphasized its presence and made it harder to tune
out.

We don’t know why this fan sounded the way it did. The absence of a frame should have reduced the turbulence, if anything. There were no odd fluctuations of voltage seen in the variable power supply driving the fan, either. Call it a mystery for now. Perhaps AC will send us another sample or two for us to see if this effect is common to other samples.

Fan @ 10V: This is not one of our usual measurement points, but it marked something of a halfway point between the variable noise
at 12V, and the barely audible noise level at 9V. The cooling performance of 26°C
rise from ambient was still very good. The noise was
constant, much smoother than at 12V. Turbulence was less
of an issue, and there was a touch of low-frequency growl. The measured
noise level of 28 dBA@1m would be considered perfectly acceptable in many environments but
unlikely to be below ambient noise in a quiet room.

Fan @ 9V: It was impressively
quiet at this level. The noise was broadband, slightly biased
towards the lower frequencies. Only a trace of the growl
at 10V remained. The 32°C
rise from ambient was still acceptable; in <25°C room temperature and a well-built PC system the maximum CPU temp would probably not exceed the low 60s, Celsius. Intel processor typically don’t start throttling till ~70°C or higher.

Fan @ 8V: It was difficult to hear the fan even from as close as a couple feet. However, the CPU temperature of 67°C is not acceptable. It would be >70°C if the system was in a PC case.

Fan @ 7V: Virtual silence was achieved. Using a pair of high fidelity headphones at fairly high gain, we were unable to tell the
difference between a recording of the Freezer Pro at 7V with our sensitive Taylor Hohendahl Engineering (THE) KP- 6M Reference Microphone just 3″ away, and a recording of
the ambient room noise ? with no noise sources in the room. However, the CPU temperature soared quickly
to 76°C and the CPU throttled.

Given that our Intel 520 is about the coolest of the 775 processors, it’s not safe to run the Freezer 7 Pro at less than ~9V in any system. Between 9V and 12V, there is a fairly wide range of cooling and noise options, however.

COMPARISONS

Along with the Freezer 7 Pro, three other heatsinks were tested on our new
socket 775 test platform: A Scythe Ninja,
a Thermalright XP-120, and
a stock Intel heatsink. (The links take you back to our original reviews of these heatsinks, tested on our P4-2.8 Northwood socket 478 platform.) Both the Ninja and the XP-120 were tested with a Nexus
120mm fan, our current reference fan of that size.

The other heatsinks and fans.

For the record:

• The clips on our Thermalright XP-120 sample may have become worn over the course of many mounts and dismounts on many motherboards. It is very fiddly to mount on this 775 socket motherboard. It’s difficult to tell whether cooling was adversely affected.

• Both the Ninja and the XP-120 were also tested with an AcoustiFan AF120C Ultra Quiet Fan. At 12V, this fan is rated for 50CFM, and, more importantly, 34 dBA@1m. This happens to be about the same noise level as the Freezer 7 Pro at 12V. The SPL was confirms with our sound level meter in the lab. The cooling improved only by 1°C on the Ninja, and just 2°C on the XP-120. With the Ninja’s widely spaced fins, the ~40 CFM of the Nexus 120 fan was already wringing close to maximum performance. With the more tightly spaced fins of the XP-120, perhaps greater performance improvement could be had with higher airflow. The AcoustiFan actually only provides just over 10 CFM gain in airflow over the Nexus 120.

The XP-120…

…and the Ninja were both mounted with the 478 conversion adaptors that
came with the Ninja.

For your reading convenience, here’s the test data on the Freezer 7 Pro from the previous page:

Comparison 1: AT HIGH SPEED / NOISE

The Freezer 7 Pro did a great job of cooling at 12V. It blew past the stock cooler, and handily beat the XP-120 + Acoustifan 120 combo by a big 5°C

margin. Only the Ninja + Acoustifan 120 combo could keep up. For overclockers and performance enthusiasts, this is very good news. This kind of cooling performance at such a modest price is a pretty amazing bargain. Both the Ninja and XP-120 sell for US$40~50 without a fan. Any 120mm fan will set you back another $10 at least. But for anyone interested in low noise, none of the above heatsink configurations are worthwhile. They are all just too loud. As for the question of better cooling for the VRM, this was not reflected in the AC power draw measurements.

Comparison 2: AT LOW SPEED / NOISE

Now, if you examine all the above performance tables, you will find that even at 12V, the Nexus fan is just 23 dBA@1m. The closest matching acoustic points for the Freezer 7 Pro are 28 dBA@1m at 10V or 20 dBA@1m at 9V. We could go back and reinstall that heatsink yet once again and run yet another test at 9.5V, but what say we fudge a little here and just split the difference in the results between 10V and 9V? We think this would be fairly representative. (If you disagree with this minor compromise, you can always rant at us in the forums.)

Why is there such a quick drop-off in performance compared to the Ninja and the XP-120?

The biggest significant difference, we believe, is the airflow impedance in the heatsink fins themselves. The Freezer 7 Pro’s fins are very closely spaced, approximately 2mm, and the distance the air has to travel through is 58mm. It’s a pretty constricted space. There really isn’t that much airflow you can feel with your fingers in the center of the far side of the heatsink, even with the fan at full speed. Much of the airflow appears to come through along the perimeter of the fins, not the center. With lower fan speeds, the impedance is high enough to make a dramatic drop in cooling effectiveness.

In contrast, the Ninja heatsink fins are about 4mm apart, and even though the distance between the gaps is a long 110mm, there are deep cutaways in the corners of each fin, and a hole in the center of each fin. With the Ninja, the impedance is so low that you can feel the same airflow all around the exhaust side of the fins at any fan speed.

The XP-120 has fins as tightly spaced as the Freezer 7 Pro, but the distance the air has to travel through is no more than ~20mm for about 75% of the total area of the fins. In the very center, the distance is closer to 50cm. Again, like the Ninja, very little impedance to airflow can be felt.

All this means that the Freezer 7 Pro is really optimized for best cooling at high airflow and pressure. Its airflow-to-cooling ratio falls steeply as fan speed is reduced. Although the total cooling surface area is large, only at high fan speed (and pressure) can the airflow reach it to blow the heat way. Both the Ninja and XP-120 maintain a steadier airflow-to-cooling ratio throughout, as the airflow changes more linearly with fan speed changes.

Regarding the cooling for the VRM, again, there were no significant differences in the AC power draw measurements.

Comparison 3: AT MINIMUM SPEED / NOISE

At this near-silent level, the performance difference among the three heatsinks are the same as in the previous comparison. There is still a 10°C spread between the Ninja and the Freezer, whose performance is still acceptable for this CPU. However, this is about the limit of the Freezer 7 Pro. You cannot run it any quieter without running into overheating problems. In contrast, the Ninja and XP-120 still have room to move; they can go down to 18 dBA@1m and still have good enough cooling performance with this CPU. The Ninja, especially, has a lot more headroom; it kept temperature rise to just 23°C

at 18 dBA.

Regarding the cooling for the VRM, again, there were no significant differences in the AC power draw measurements. You will probably have noticed that for all the heatsinks, there appears to be a direct relationship between CPU temperature and system power draw: The hotter the CPU gets, the more power the system draws. The range between the lowest and highest power draw was 132W to 141W. Does this mean higher CPU temperature always results in poorer VRM or system efficiency? Or is it that the reduced airflow that causes higher CPU temperature also happens to provide less cooling for the VRM? If the VRM can be cooled independent of CPU temperature, will power draw still go up when the CPU gets hotter or will it stay stable? These are all questions for another article.

Finally, it’s too bad that we didn’t have a suitable Zalman 7000AlCu or 7000Cu model to compare against the Freezer 7 Pro on this platform. We have a couple of them in the lab, but they are older models without proper hardware for socket 775. We’re pretty sure the Freezer 7 Pro would stomp all over the 7000s at the full 12V speed, but they are close in price, and at the lower, quieter fan speeds we’re more interested in, they might be close competitors. Perhaps later in an update, or when we look at the Freezer 64 Pro.

NOISE RECORDINGS

Arctic Cooling Freezer 7 Pro:

MP3: Arctic
Cooling Freezer 7 Pro – 7V / <17 dBA@1m

MP3:
Arctic Cooling Freezer 7 Pro – 9V / 20 dBA@1m

MP3:
Arctic Cooling Freezer 7 Pro – 10V / 28 dBA@1m

MP3:
Arctic Cooling Freezer 7 Pro – 12V / 33~37 dBA@1m

Recordings of Comparable HSF:

MP3:
Arctic Cooling Freezer 4 (original version) – 7V – 20 dBA@1m

MP3: Arctic
Cooling Super Silent 4 Ultra TC, 22 dBA@1m

MP3:
Nexus 120mm fan – 12V – 41 CFM – 22.5 dBA@1m

MP3:
Nexus 120mm fan – 8.8V – 35 CFM- 19 dBA@1m

FINAL CONCLUSIONS

Straight out of the box, the Freezer 7 Pro has the kind of cooling performance that will make it a darling with overclockers.
It can take on the highest performance heatsinks head on without hesitation. At its suggested price of US$34, we know of nothing that can compete for cooling performance. Rave reviews and market success with overclockers and gamers is virtually guaranteed. The same can probably be said of the forthcoming Freezer 64 Pro, unless the fan is substantially changed.

Given that the product comes without any fan speed reduction device, it is clear that Arctic Cooling expects performance-oriented buyers. It’s not that AC doesn’t make fan speed controllers; many of their HSF come with multi-speed switches or thermal sensors to control fan speed. Perhaps their attitude is that if you’ve chosen a socket 775 platform, you already know you need all the cooling help you can get, and a little noise probably won’t bother you as long as that hot CPU is kept cool.

Compared to the original Freezer, the new model is larger, quieter, and
its fan is more securely mounted. The fan is also uniquely soft-mounted to reduce vibration-induced noise.
These are all improvements over the original model.

But, the Freezer 7 Pro is not for everybody. At normal (12V) speed, the fan is too loud for a quiet computer. Dropping the fan speed requires another device like the Zalman Fanmate, not a big expense, but something to consider. At lower speeds, the fan is much better behaved, but the cooling performance also falls off fairly quickly. The usable fan voltage range is just 12~9V, which seems quite narrow. In that range, the fan drops from 34~37 dBA@1m down to just 20 dBA, while its cooling performance also drops from only 18°C rise over ambient to 32°C. Because of its high impedance fins, the Freezer 7 Pro does not downscale gracefully like the best of the low airflow optimized heatsinks. Anyone expecting the same stellar cooling performance at lower fan speeds and noise will be a bit disappointed.

This is not to say the Freezer 7 Pro cannot be used successfully to make a quiet Prescott system. It can be done with good case / system configuration to ensure a cooler ambient for the reduced thermal headroom at lower fan speed.

Overall, the Freezer 7 Pro is an impressive HSF with just a couple of faults, at least for us: The high level and odd surging turbulence noise at full speed (which might be specific to our sample?) and the rapid drop in performance at lower fan speed. Arctic Cooling could make a very attractive quiet variant by simply removing a few fins to reduce the airflow impedance and adding some kind of speed control for the fan.

Much thanks to Arctic
Cooling for the Freezer 7 Pro sample.

* * *

Comment on this article in our Forums.