In spite of our best efforts, some SPCR readers found fault with a few of the airflow measurements in our recent fan roundups. We revisited the airflow metrology. After much experimentation and some test methods that were rejected, we settled on a new way of assessing fan airflow under some load (impedance) that seems more consistent and reliable.
May 2, 2007 by Devon
Cooke and Mike Chin
What is that in the photo below? Three guesses — no cheating!
Need
a hint? Remember the title of the article.
Is it modern sculpture? No? Then what…?
Neither the photo nor this article would exist if it weren’t for the
fact that, occasionally, we get things wrong. We make a recommendation, people
take us at our word and wham — our
forum is full of people complaining that our recommendation didn’t
make their system any quieter. These situations don’t arise often, but we take
them seriously when they do.
Most recently, we identified the Noctua’s
NF-S12 series as the fan with the best airflow-to-noise ratio. Naturally,
we were surprised when reports began to surface that, while the Noctua was indeed
very quiet, some users were noticing their system temperature going up, not down.
Since our recommendation assumes that higher airflow normally leads to lower temperatures, these reports cast our measurements
into question.
Could a fan that we measured as having higher airflow provide poorer
cooling in the real world?
Since our usual method of holding an
anemometer directly in front of the fan to find the peak airflow didn’t
seem to be working, we needed to get creative. And that brings us to the odd-looking
piece of modern art in the photo above. In proud SPCR tradition, we decided
to build what we couldn’t buy, and the strange device you see above was the
result.
PROBLEM FANS
To the best of our knowledge, our technique of holding an
anemometer directly in front of the fan to find the peak airflow works perfectly
well for the vast majority of fans. The ones that cause problems are like the
Noctua: They have some quirk that sets them apart from other fans. In the case
of the Noctua, that quirk is thin, propeller-like fins with a sharper pitch
than most fans, but we discovered several other quirks that seemed to affect
measurements. Among the most serious examples:
- Arctic Cooling’s frameless, reverse-direction Arctic Fan 12L and
its smaller cousins, including the
Arctic Fan 3 we reviewed last November. These fans spin in the opposite direction of the impeller of our anenometer. - The Noctua NF-S12 series.with their very thin fins.
- The 80mm Mechatronics
fan with its thin, stubby fins. - SilenX’ Ixtrema Pro series, with their wide, scooping fins and tiny
diameter hub.
All of these are problem cases because our standard measurements show an RPM-to-Airflow
ratio that is significantly different with other fans of similar size. Are these differences real or errors caused by our measurement technique? With the feedback on the Noctua fans, we weren’t sure.
Fan manufacturers use complex multichamber testing tools to measure airflow. The cost of these tools runs into many thousands of dollars, possibly into the tens of thousands. The excerpt below from Laboratatory Methods of Testing Fans for Aerodynamic Performance Rating, document ANSI / AMCA STANDARD 210, shows just how complex such tools can be.
Such tools are far beyond SPCR’s reach. We’d have to find our own way.
EXPERIMENT #1: STRAIGHTEN THE AIRFLOW
We began our quest for better measurements by starting with the most obvious
error: The Arctic Fan 12. We measured less than half the airflow of conventional
fans at similar RPM, and we were fairly certain we knew why: The direction of
rotation was wrong for our anemometer.
The rotation of a fan’s blades causes air to exit the fan in a swirling vortex,
not a straight line. Like a miniature tornado, the vortex spins in a particular
direction. Because 99% of fans turn in the same direction, this effect can usually
be ignored. However, a the Arctic Fan 12 produces a vortex that spins
in the opposite direction from “normal”, which has a serious affect on our anemometer
since it spins in the “normal” direction.
Our anemometer in use.
The solution was fairly obvious: Find a way to straighten the vortex airflow. The photo below shows our first attempt.
A bunch of tightly packed straws mounted on a foam base. The straws are set
about two inches in front of the fan so that the airflow is more evenly distributed
across the straws.
Unfortunately, our hopes for this contraption were dashed when we started measuring
airflow. The measured airflow of most fans that we tried in the device dropped by almost half (and even more at low rotation speeds) compared to our directly-in-front measurements. The one exception was the Arctic Fan 12L,
whose measured airflow barely dropped at all and thus brought its result in line with the rest of the fans we tested. This suggested that we had fixed the problem we
set out to solve — the airflow reaching the anemometer blades was no longer swirling, so the spin direction of the fan did not affect the measured airflow. But the device introduced far too much impedance, especially at low speeds.
The fan mounts in the square cut-away.
Looking straight down the airflow path.
EXPERIMENT #2: A FAN AIRFLOW BOX
Our next attempt took a different approach. Rather than trying to straighten
the airflow directly, we used an airtight box (an adapted acrylic computer
case) with a baffle to eliminate the direct flow of the vortex to the anenometer. The only intake is the fan itself; the only exhaust, the anemometer
at the other end. Every seam, screw hole and vent other than the fan intake and the aneometer vane exhaust was sealed with clear packing tape. A baffle made of neon yellow postercard was affixed inside the box to ensure that there is no direct path between the fan and the anemometer.
Some judicious use of packing tape made this acrylic case completely
airtight.
The baffle ensures that the swirling fan exhaust vortex
does not reach the anenometer while introducing very little impedance. On top of that,
we no longer need to move the anemometer around hunting for the peak reading
during measurement because sealed box guarantees that any air blown into the box
flows out through the anemometer.
Fans are mounted using custom-cut grommets made of closed-cell foam.
Any air blown into the case is forced through this anemometer.
PRELIMINARY RESULTS
To verify that the new airflow test box
could produce useful results, we tested a handful of fans that have already been measured the previous way. The results are summarized in the table below. You’ll notice the airflow is reported in LFM (Linear Feet per Minute) instead of the usual CFM (Cubic Feet per Minute) unit we’ve used in the past. The reason for this will come later.
| Airflow Measurements Compared: LFM of 120mm Fans | |||||
| Fan | Method | 12V | 9V | 7V | 5V |
| Arctic Fan 12L | old | 350 | 270 | 200 | 120 |
| new | 390 | 300 | 230 | 140 | |
| Antec TriCool | old | 770 | 610 | 470 | 320 |
| new | 810 | 650 | 530 | 370 | |
| Nexus 120 | old | 480 | 360 | 270 | 160 |
| new | 470 | 370 | 300 | 210 | |
| Noctua NS-S12-1200 | old | 650 | 530 | 430 | 310 |
| new | 590 | 480 | 390 | 280 | |
| Scythe S-Flex SFF21E | old | 550 | 390 | 280 | 140 |
| new | 600 | 430 | 310 | 170 | |
| SilenX IXP-74-11 | old | 580 | 470 | 420 | 330 |
| new | 510 | 420 | 360 | 270 | |
The exceptional fans, the ones that produced suspect measurements
before, now produced results more in line
with our expectations. The results for the Noctua and SilenX fans — both
of which previously measured unaccountably high — dropped down closer to
what we would expect. And the results for the reverse-direction Arctic Fan
12L rose slightly, though it still seemed to blow less air at a given speed than
the more conventional fans we tested. The Antec Tricool and Scythe fans also measured higher. The Nexus measurements were closest to each other, but the new method showed greater airflow at low speed.
The graphs of the data may be easier to understand. The two sets of measurements for each fan is in similar colors to make them easier to see and to compare. Note how most of them run parallel to each other.
All in all we were quite pleased with the result. This airflow measurement method
seemed to produce more reliable results, and it also simplified testing. We could now dial in a specific voltage
with any fan, and quickly obtain an airflow measurement consistent within 10~20
LFM. Quick, consistent results were very difficult to obtain with our old manual technique.
We also tested a few 92mm and 80mm fans to see how they were affected.
Here, we found bigger differences between the old measurements and the new ones. The measured airflow of virtually every fan we tested dropped substantially.
| Airflow Measurements Compared: LFM of 80mm & 92mm Fans | |||||
| Fan | Method | 12V | 9V | 7V | 5V |
| Nexus 92 | old | 490 | 390 | 300 | 190 |
| new | 410 | 320 | 250 | 150 | |
| Fander FX92-W | old | 810 | 640 | 480 | 340 |
| new | 580 | 460 | 360 | 250 | |
| Nexus 80 | old | 550 | 370 | 250 | 120 |
| new | 390 | 280 | 180 | 100 | |
| Scythe SA0825FDB12SL | old | 630 | 440 | 300 | 140 |
| new | 420 | 310 | 200 | 100 | |
| Arctic Fan 3 | old | 320 | 260 | 170 | 90 |
| new | 410 | 320 | 260 | 160 | |
The graph below illustrates it more dramatically. This time, the pair of curves for each fan is the same color. Note how much wider the gap is between the two lines for most of the fans, compared to the 120mm fans above. The greater difference at 12V is more marked as well.
EXPERIMENT #2 PROBLEMS
Why is the difference between the old results and the new ones greater with the smaller fans? We’re not positive about the answer. Fan aerodynamics
is a complex field, and what we can offer here are observations and hypotheses.
Our old technique measured the airspeed directly
in front of the fan, without worrying about capturing all of the air
flowing through the fan. The airspeed was assumed to be more or less constant
across the whole area of the fan, and that speed (in linear feet per minute)
was multiplied by the area of the fan (in square feet), to produce a final
flow volume (in cubic feet per minute). Our readings were affected by extra turbulence caused by placing the anenometer so close to the fan blades themselves; with most fans, the effect was to raise the CFM readings.
With the new technique, because
the test box is sealed, every bit of air pushed into the box
by the fan will eventually pass through the anemometer and contribute to the LFM measurement.
The anenometer impeller measures 68mm (2.67″) in diameter, which is a lot smaller than 120mm. It is also smaller than 92mm or 80mm.
| FANS versus ANEMOMETER IMPELLER | ||||
| Radius | Anenometer | 120mm | 92mm | 80mm |
| Area* | 36cm² | 113cm² | 66cm² | 50cm² |
| *This calculation includes the central hub area. | ||||
The airspeed
through the anemometer is not the same as the airspeed right at the
fan. Instead, it is governed by a complex formula involving the difference in size
between the fan and the anemometer, the speed of the fan’s rotation, and the impedance of the test box, which includes the anenometer vane.
We are no longer measuring unimpeded airflow; our test box brings
pressure into the equation as well.
What effect does this have on our measurements?
The instruction manual of most handheld anemometers contain this information:
To determine the volume of air flowing through a duct, take the area of the duct in square units (like square feet) and multiply the value by the measured linear velocity (ft/min).
In the airflow measurement technique we’ve been using all along, the area of the test fan’s impeller opening (diameter of the blades minus the diameter of the center hub) is considered to be the duct. In the airflow test box, this can no longer be valid. It is the anenometer impeller which becomes the duct. This means the fan diameter no longer has any bearing on the relationship between the measured LFM and calculated CFM. It also means that the relationship between measured LFM and calculated CFM is now linear regardless of fan size; the area of the anenometer impeller is used for every fan.
Now compare the LFM results between the various size fans:
| LFM measurements in Airflow Test Box | |||||
| Fan | rated CFM | 12V | 9V | 7V | 5V |
| Arctic 3 (80mm) | 28 | 410 | 320 | 260 | 160 |
| Scythe 80 | 27 | 420 | 310 | 200 | 100 |
| Nexus 92 | 27 | 410 | 320 | 250 | 150 |
| Fander FX92-W | 35 | 580 | 460 | 360 | 250 |
| Scythe S-Flex SFF21E | 49 | 600 | 430 | 310 | 170 |
| Arctic Fan 12L | 37 | 390 | 300 | 230 | 140 |
| Noctua NS-S12-1200 | 48 | 590 | 480 | 390 | 280 |
| Antec Tricool 120 | 79 | 810 | 650 | 530 | 370 |
Let’s accept the manufacturers’ CFM claims for the time being. The discrepancies between the claimed CFM and our measured LFM are dramatic in some cases, and consistent:
1) The Antec Tricool 120 has the highest rated airflow of 79 CFM. But its measured LFM is only double that of the Scythe 80mm or Nexus 92, both rated at less than a third of the Tricool’s CFM. Its rated CFM is more than double that of the Fander 92, but the measured LFM is only about 35% higher.
2) The Noctua 120 and Scythe 120 are rated at 48 and 49 CFM, while the Fander 92 is rated at 35 CFM. Yet the measured LFM of the Fander is the same as the bigger fans.
3) The Arctic Fan 12L, rated at 37 CFM, has slightly lower measured LFM than the Arctic and Sycthe 80mm fans and the Nexus 92, all three of which are rated at 27 or 28 CFM.
The consistent part of the above data is that it’s almost always the larger higher CFM fans that don’t seem to have high enough measured LFM. This suggests that there is higher airflow restriction at the anenometer for the bigger fans at higher speeds. In other words, the pressure in the box rises for the bigger fans at higher speeds, resulting in a compression of the airflow measurements.
If we take the point of view that to minimize pressure and airflow restriction, the exhaust vent (the area of the anemometer impeller) should be at least as big as the intake vent (fan impeller area), then our test box with this anenometer should be limited to fans no bigger than about 70mm diameter.
In conclusion, we can probably say the airflow box of Experiment #2 provides…
- a fairly low resistance condition for 80mm and 92mm fans spinning not too fast; we don’t know how fast is too fast.
- a higher impedance to almost any larger diameter fan even when it’s spinning quite slowly.
This means data between smaller and larger fans, and perhaps between faster and slower fans, cannot be compared fairly due to the increasing pressure that prevails with higher airflow.
In the end, we had to call Experiment #2 an interesting failure.
A HOT WIRE ANEMOMETER
There was no question that the airflow box solved at least one of our problems: The swirling vortex of fan-generated airflow and the effect of its spin direction on measured LFM. Was there some way that we could measure airflow using the box without running into the higher pressure problem with bigger fans?
One solution would be an anenometer with a much larger impeller. Perhaps there was one with 120mm diameter blades?
An extensive search through test equipment manufacturers’ catalogs ensued. In short, the anwser was no. There does not appear to be any anenometer on the market with a 120mm diameter impeller.
However, the search did turn up a different type of anemometer we had not considered before: Hot Wire Anemometers. This type of tool was ignored in our initial market search for anemometers back in 2003; they were simply too costly to justify at the time. Today, with SPCR well established, the investment in an important tool is more easily justified.
“The hot-wire anemometer, principally used in gas flow measurement, consists of an electrically heated, fine platinum wire which is immersed into the flow. As the fluid velocity increases, the rate of heat flow from the heated wire to the flow stream increases. Thus, a cooling effect on the wire electrode occurs, causing its electrical resistance to change. In a constant-current anemometer, the fluid velocity is determined from a measurement of the resulting change in wire resistance. In a constant-resistance anemometer, fluid velocity is determined from the current needed to maintain a constant wire temperature and, thus, the resistance constant.” (Cited from The Engineer’s Edge web site.)
The most salient aspect of a hot wire anenometer is that its sensor is very small, and mounted at the end of a thin wand. This means that it can be placed in the airflow without creating any significant resistance.
Close market research turned up the Extech Model 407123 as the least expensive, full featured hot wire thermo-anemometer on the market. An order was duly placed, and a week later, the item arrived at the lab.
Extech Model 407123 hot wire thermo-anemometer
The cable from the telescoping antenna-like wand plugs into the top of the handheld digital readout meter. (The telescoping wand is useful for HVAC personnel when assessing airflow in large building ducts.) The black colored head of the wand is where the sensors are located. A close up photo is shown below.
Sensors at the end of the wand.
So how were we going to use this new test instrument? Well, we wanted to stay with the fan test box, for sure. It’s a low impedance setup, but has some impedance, and it’s more representative of actual use conditions than free air. We also wanted to stay with fixed position measurements, as opposed to hand held muddling.
EXPERIMENT #3: THE NEW AIRFLOW TEST SYSTEM
After a few days of experimentation, we came up with a setup and methodology that appears consistent, repeatable and reliable. Here it is in a photographic nutshell:
Fan airflow test box set up with hot wire anemometer.
The exhaust vent is the cutout for the power supply on the back panel of this acrylic ATX computer case. This is where the airflow is measured. The exhaust vent is about 10% larger than the impeller area of the typical 120mm fan. This means the exhaust vent should not constrict the airflow in any significant way, except perhaps in extremely high airflow models — but we wouldn’t be interested in such fans anyway because they’d be far too noisy from the air turbulence alone.
Experimentation with sensor wand placement indicated that despite the ~1.5′ distance from the fan and the baffle to minimize direct airflow between fan and exhaust vent, the airflow rates are not the same across the exhaust opening. To be specific, there is always a peak of up to 10% on the left side of the exhaust opening, compared to the center. Further to the right, the flow tended to dip about 5%.
There are all kinds of possible reasons for this positional variance in airflow — the airflow path is not perfectly symetrical, and neither is the vent opening, and the swirling vortex effect of the fan exhaust may still be apparent at the exhaust. The reasons are not really that important. What’s more important is whether we can get rid of the variance, and if not, how we can deal with the variance.
In the end, we decided to accept the positional variance, and to deal with it by taking the high and low readings from the three positions, then use the average of all six readings. This averaged LFM figure would be multiplied by the area of the exhaust opening to give us the CFM value. We set up a jig with grooves to secure the sensor into the same positions every time; this is the white closed-cell foam block that’s holding the sensor wand in the photos above and below.
Sensor wand locked in the right position.
EXPERIMENT #3 PRELIMINARY RESULTS
Results for selected fans retested in the airflow box with the hot wire anemometer are listed in the table below. The 80mm fans are on top, followed by the 92mm and 120mm fans.
| CFM results from Airflow Test Box w/ Hot Wire Anenometer | |||||
| Fan | Rated CFM@RPM | Measured | |||
| 12V | 9V | 7V | 5V | ||
| Nexus 80 | 20@1500 | 18 | 14 | 10 | 6 |
| Scythe 80 | 19@1500 | 19 | 14 | 10 | 6 |
| Arctic 3 | 28@1900 | 23 | 18 | 15 | 10 |
| Nexus 92 | 27@1500 | 20 | 17 | 14 | 9 |
| Fander FX92-W | 35@2000 | 34 | 26 | 19 | 13 |
| Arctic Fan 12L | 37@1000 | 28 | 20 | 17 | 11 |
| Nexus 120 | 37@1000 | 33 | 26 | 20 | 15 |
| Noctua NS-S12-1200 | 48@1200 | 38 | 30 | 23 | 15 |
| Scythe S-Flex SFF21E | 49@1200 | 40 | 27 | 19 | 11 |
| Antec Tricool 120 | 79@2000 | 58 | 43 | 33 | 21 |
A few generalizations can be made from the above data:
1) The airflow readings are significantly lower than with the direct-at-fan method we’ve been using. This suggests some impedance in the box, which is expected. We no longer have any expectation that our measured CFM will match the free-air CFM specified by the manufacturer. Yet, some of our measurements are identical to the manufacturer-specified CFM; we believe this is mostly happenstance.
2) There is a general correlation between RPM and measured airflow for fans of the same size. Good examples are the Nexus and Scythe 80mm fans, and the Noctua and Scythe 120mm fans; each pair, which spins the the same speed, have the same or very close measured CFM. Another example is between the 1000 RPM 120mm fans and the 2000 RPM Antec Tricool 120: The latter has roughly double the CFM of the former, which is expected from the fan laws (assuming the same fan blade geometry).
3) The bigger fans with higher specified airflow do measure proportionately higher than smaller fans with lower specified airflow. In other words, there is no serious compression of measured airflow for larger, high speed fans, unlike the results with Experiment #2. For example, the Antec Tricool 120, rated at about double the CFM of the Fander 92, does give us roughly double the airflow measurement.
One fact not mentioned in the table is that there was virtually no difference in the RPM measured in free air compared to in the test box. The maximum variance we saw was a drop of about 40 RPM for a medium speed 120mm fan. This indicates that although the box does pose some impedance to airflow, it is very low, and the impedance does not load the fan enough to force any significant drop in fan speed. Low noise fans seem to be most susceptible to RPM drops. This may be a coincidence, but it’s probably of interest to most SPCR readers.
The anomalous fans which were partly responsible for this revisit of our airflow measurement technique no longer give us anomalous results. The specific fans in the above table are the Arctic Cooling 12L, which is virtually frameless and also spins in the direction opposite to most fans, and the Noctua, which has very thin blades and gave high airflow readings in our original tests.
The Arctic Cooling 12L gave us very low airflow readings previously. Now the measured airflow is much more reasonable, although still lower than specified, and lower than the same speed but conventional design Nexus 120. This may be a real weakness of the Arctic Cooling open frame design; we suspect that some of the kinetic energy of the fan blades is dispersed and lost on the intake side before it has a chance to be directed into the output stream.
Our original measured airflow on the Noctua 120, on the other hand, was unusually high. At 63 CFM, it was 15 CFM higher than specified. The new testing brings this number down to earth; the measured 38 CFM is more in line with other fans of similar RPM and size. We must retract the comments we made about the Noctua’s airflow in our original review. It does not push significantly more air for a given RPM than other fans. In this regard it is ordinary. But most importantly, it remains a very quiet fan, and comparison with both the Nexus and Scythe SFF21E 120mm fans, the Noctua still seems to have an edge in airflow at lower speed.
RPM VS. AIRFLOW VS. COOLING
This long study of axial fan airflow led us inexorably to the question of the relationship between fan speed, airflow and cooling. We’ve stated in the past that all other things being equal, higher airflow usually means better cooling. This statement is a simplification. The reality is that airflow is one of many factors of cooling in electronics.
Surprising though it may seem, increased airflow does not automatically or always mean improved cooling. One of the closely related factors is the rate at which heat is brought to the surface where airflow can carry it away. Looking specifically at CPU heatsinks, given a certain amount of heat generated by the CPU, a heatsink conducts heat at a certain rate to the fins where it can be dissipated into the air. If the flow of air across the fins remains below the rate at which heat is conducted from the CPU to the fins, then increasing airflow improves cooling and lowers CPU temperature. But once the rate of airflow matches the rate of heat conduction to the fins, further increases of airflow will not improve cooling.
Airflow Vs. RPM
The first matter is to look at the close relationship between RPM (fan speed) and cooling with a given fan size. All of the fans in the next table are 120mm fans. They were used with a Scythe Ninja (original version) heatsink on our current socket 775 heatsink test platform. The data in the table speak clearly enough.
| Thermal Test Results: 120mm Fans mounted on Scythe Ninja | ||||||
| Fan Speed | Nexus 120 | Antec TriCool | Scythe S-Flex SFF21E | Noctua NS-S12-1200 | SilenX IXP-74-11 | Arctic Fan 12L |
| 500 RPM | 49°C | 50°C | 49°C | 48°C | 50°C | 48°C |
| 800 RPM | 46°C | 45°C | 46°C | 44°C | 45°C | 43°C |
| 1,100 RPM | 41°C | 42°C | 41°C | 42°C | 41°C | 42°C |
| 25 CFM* | 45°C | 46°C | 46°C | 45°C | 45°C | 45°C |
| Test Platform Details: | ||||||
At the same RPM, the various fans achieved CPU cooling within a 3°C range. At 1100 rpm, the range dropped to just 1°C; the CPU temperature was 41°C or 42°C with every fan.
The last line shows the results with every fan set to the RPM at which it delivered 25 CFM. As expected, the temperature variance was a mere 1°C again. This helps to verify that our testing is giving us accurate airflow readings.
What all of this data tells us is that airflow is very closely related to RPM and fan diameter. This is a basic fan law. Given the same RPM, most fans of the same size provide about the same airflow. There are no miracles, despite extensive technical marketing talk about blade design and geometry, which seem more or less irrelevant.
Airflow Vs. Cooling
The above table shows a steady drop in CPU temperature as the fans were sped up, from ~50°C at 500 RPM to ~41°C at 1100 RPM. It might be tempting to conclude that this supports the idea that higher airflow results in better cooling. But does this hold true even as we keep increasing airflow? To answer this question, a high speed 120mm fan was brought into play. We feel that the result is important enough that it deserves a highlighted text box.
| JMC model 1225-12HB At 2700 RPM, this fan spins some 2.5 times faster than the Nexus 120 at 12V (and most of the other 120mm fans in the table). It also generates more than double the CFM, according to our measurements. It makes enough noise to give most SPCR enthusiasts a headache in a couple of minutes. What was the CPU temperature achieved by this high airflow fan? 41°C. Exactly the same temperature reached by most of the 120mm fans at 1100 RPM. What this tells us is that for this combination of CPU, load, and heatsink, the airflow of an 1100 RPM 120mm fan matches the rate of thermal conduction from CPU to heatsink fins. This is why increasing the airflow does not decrease the CPU temperature. The increased fan speed does not change the ability of the cooling system to conduct the heat any faster to the fins. Only if the heat of the CPU was increased or the fin area of the heatsink expanded would the increased airflow result in a change. |
So what practical implication does this have on the quiet-seeking PC builder or modder? What we’ve been saying all along for years:
- Choose the quietest fan with the best noise signature and adjust its speed to give a balance of cooling and noise you can live with.
- Don’t worry about CFM. Just go for the quietest fan and the best low-airflow performance heatsink, and minimize airflow impedances in your case.
It’s funny, but true: We went through this entire rigamarole of establishing an accurate system of measuring axial fan CFM only to tell you it doesn’t really matter, and not to worry about it. It may seem a waste of time, effort and money to some, but this type of process has always been an integral part of empirical, scientific exploration at SPCR.
CONCLUSIONS
The airflow measurement system we described for Experiment #3 with the test box and the hot wire anemometer has become our new fan airflow test methodology. The resulting CFM data makes sense for lower and higher airflow fans of all the standard sizes (80mm to 120mm diameter). It is also repeatable and consistent with all the fans we’ve retested. The fact that it represents flow into some kind of impedance rather than free air makes the results more relevant for practical purposes.
Our CFM data is not directly comparable to manufacturers’ specifications, which are purported in free air, without any load. They may never have been directly comparable anyway.
Keep in mind the comment made recently by Russ Kinder in the SPCR forums:
“General rule (about fan specs):
“The people who actually make the fans tend to have reliable specs, for legal and liability reasons. If an engineer spec’s a fan for a piece of equipment, and that equipment then dies a fiery death due to bogus fan specs from the fan manufacturer, you can guarantee that the engineer’s company lawyers will be very interested.
“But the people who resell the fans are free to lie, cheat, and steal to their hearts’ content, because there are no consumer protection labeling laws regarding fans or PC components.”
We will be updating the airflow information for all the fans in all the roundups thus far. Yes, we’re busy collecting this data now. The airflow information will be the only data in the reviews that will be changed. In many cases, the airflow data will not change by much. There will be some changes in the text and analysis for the unusual fans whose odd characteristics prompted these airflow measurement experiments in the first place. Neither noise nor RPM data will be changed. Most significantly, our assessments of the noise characteristics of each fan will stay unchanged.
* * *
SPCR Articles of Related Interest:
SPCR’s Fan Roundup #3: 92mm Fans
SPCR’s Fan Roundup #2: 120mm Fans
SPCR’s 80mm Fan Roundup #1
SPCR’s Fan Testing Methodology
Anatomy of the Silent Fan
Simple Fan Controllers from Zalman
Get 5V, 7V, or 12V for your Fans
* * *
Discuss this article in the SPCR Forums.
