Would it have anything to do with static pressure? Perhaps the Noctua is "sucking" up plenty of air, but can't really push it through the fins of the sink, and a lot of the air is kinda venting out the side.

Maybe the way the air comes out of the fan is different. I think the noctua's output airflow is much higher on the outside edge of the fan rather than more spread out across the fan's outpu area (like a cone shaped ariflow, with the air following the sides of the cone, without much of anything in the middle). It's the feeling I get with the scientifically proven "hand-in-the-airflow" test
That would explain why it performs worse while still having a higher airflow than other fans : overall, the air the noctua is pushing is in contact with less surface from the heatsink, hence moving less heat out of the way. Maybe the Nexus' airflow is a bit more evenly spread out, so even if it moves less air, the air it moves is more "efficent" because it hits more surface from the heatsink.
I'm not sure that makes senses to anyone but me...

So if we could find a way to visualise the airflow comming out of a fan in 3D that would answer some questions I think.

[EDIT : typos and clarification, hopefully]

dev, dev, dev....

How have you missed it again? The age of the HS design is completely irrelevant, and even how good/bad it is hardly matters. The only thing that really matters is that it has neither too high nor too low impedance to airflow, and that the heatsink/heat combination is very responsive to changes in fan airflow. We have verified that the TRUE is both of these things, more so than other HS we could have chosen.

I think...the fact that you used an off the shelf CPU cooler, and that you mentioned that you were emulating an i7 in terms of heat load is causing many to incorrectly go into "CPU cooling test" mode, so to speak, as opposed to seeing the setup as a controlled test load for fans. =\

I'm just sayin'.

You're judging the fan performance based on the temp you get from the TRUE. All I'm saying is some folks might not get comments like: " a couple of degrees cooler but 6 dB noisier". Louder, higher quality fans have their merits and might work better on a radiator or a different heatsink. I think just like the heatsink testing platform you've put yourself in another minority.

Lets say I own a Thermolab Baram or even a newer revision of the TRUE and I own a Slipstream and based on your results I buy a Nexus. Surprisingly there is no improvement and there are situations where the Nexus is worse. Am I entitled to blame you for this?

I know this is a worst case scenario but I think its plausible.

Quite the reverse -- you are the one who's in the minority here. No one here cares about absolute cooling performance -- it's always relative to the noise.

If you disagree with everything we do here, just bug off w/ your irrelevant comments.

I think most here should really care about enough cooling performance at the lowest sound level possiable. So yes absolute cooling performance matters, if you need to cool 60w, there isn't much of a good reason to be able to cool 200w, even if it causes an increase in sound of only 2SPL. The ratio is important, but being able to cool 1kW@50SPL, isn't all that useful when you only have a 200W load, and it can be cooled by a fan that only has a 25SPL.

You can replace watts with degrees C if you like.

So the second fan would look worse than the first, but the first is much louder, and complete overkill for most modern CPUs. Thats my only complaint with the method. Absolute SPL for a giving cooling capacity is what really matters. Granted i'm not expecting to see a gap in performance like that in the real world.

I agree with the comments about needing some ducting between the fan and the anemometer. I also thing that a manometer is probably needed across the anemometer to account for any pressure drop caused and correct the flow.

Those two statements contradict each other. When you say "enough cooling performance at the lowest sound level" this is already not an absolute, there are several conditions other than lowest temperature you seek: "enough" cooling (not the lowest temp) -- "at Lowest sound" (condition other than temp). So your second statement does not follow.

Look, let's not get into splitting hairs even before a single review from the new system is posted. What we will find, I suspect, is that the noise-to-temp curves for most fans will lie very close to each other. At the 16 dBA/1m level, if a fan can get that low in level, I would be surprised if there is more than 2deg C variation amongst all the fans. At the 13 dBA/1m level, there might be slightly more variation.

We will see about the higher numbers, but in truth, I personally am not interested in 20 dBA/1m or higher performance -- this is already too loud when you put the fan in a case and add the cavity resonance and vibration effects that always come into play. (point in fact -- there's a system in the chamber now that measures 14 dBA/1m which is low, but it is audible and I'm not certain I find it acceptable. In other words, sound quality is a huge component of our fan analysis, of all our sound analysis, and it will always continue to be so.) The data at the higher SPLs will be collected, but its not for our core audience; any time fans are tested, the OC'ers come around and it's in SPCR's best interest to feed them what they want.

Good stuff Mike. I guess I never really thought about it before, but delta C does seem like a much better metric for measuring fan performance on a heatsink. One simple measurements that takes into account static pressure and other variables that would be difficult and time consuming to measure separately. The flow rate results are kind of funky, though, especially for the Noctua. I think KayDat is on the right track, air must be "leaking" through the space between the fins and frame. One of the other Noctua fans designed for higher pressure that doesn't have that spacing would probably leak less and perform better.

Also, you mentioned in the article about finding different reference heatsinks for other fan sizes. There's probably a good reason you guys decided against this that I'm not thinking of, but could you jury-rig the Ultra 120 to accept other fan sizes? Maybe just tape on smaller fans or something like that? Would save you the trouble of having to swap out heatsinks on the test platform.

I can think of several reasons for these discrepancies that you found Mike. They all revolve around fan (and blade) geometry. We're starting to dive into some complex fluid dynamics here...so I will try to keep this brief.

Here's an example of how blade geometry could give rise to this otherwise inexplicable phenomenon. The pictures below are my attempt to represent possible velocity profile differences between the two models. The longer the green line, the faster the air is moving at that particular radius from the axis. (again this is all theoretical..just to make a point)



*pretend the noctua is facing the other way when looking at my lines..couldn't get an angled picture of the back...lazy*


Now, velocity profiles like this could possibly explain your measured difference in FPM. The ultra sensitive velocity meter could be affected by more air moving closer to it's axis (hence making it read a faster velocity due to the smaller moment arm).

But I digress, the main thing I'm getting at here, is possibly the velocity meter you have is more geometry dependent than originally thought, and that the velocities that it measures aren't necessarily accurate correlations to the actual volume of air passing through each test fan. The complexity of fluid dynamics is staggering...

No question your conjectures have merit, and yes, this is a most complex subject. The FPM measurements actually create much more confusion than clarity -- perhaps the best thing is to forget about ever measuring airflow again, sell off all the anemometers and eliminate all talk of FPM, CFM or APH.

I think it might be worthwhile to measure open-air FPM, because that would probably be pretty relevant for people who want to use them as case fans or in other situations involving low static pressure. But I agree that measuring FPM with the fan mounted on a heatsink doesn't really seem to provide any useful insight into the fan's performance.

I don't think this would be the case, since on page 2 it says the airflow was measured at the intake side of the fan. Any velocity profile created from blade geometry would appear at the exhaust side of the fan. Also, since the fan is strapped directly to the heatsink, there is virtually no room for a proper velocity profile to develop. My guess would be that the measured airflow going into the fan isn't the same as the airflow that's actually passing through the heatsink. Like mentioned before, probably air is being lost flowing out the sides.

Hi Mike,

Glad to see the new setup is easier to use, faster, and more repeatable. Can you talk a bit more about repeatability? How many runs are done, are the results averaged, etc.. Are you seeing +/- 1.0dB or 0.1dB and 0.5C or 5.0C repeatability at a given rpm, etc...

Might be useful to add this to the test procedure boilerplate as well.

Thanks!

This is a dumb question but have you tried adding smoke to the airflow to see the difference in the pattern how the air itself is ejected from each fan?

Nope, and we're not doing it!

The SPL variations with multiple runs are very small -- under 1 dBA.

The temperature might vary a bit more but it's harder to tell because this involves two temps to monitor -- it might be 1C or a bit higher, but it probably depends on how long the ambient/CPU temps are monitored. There's always a bit of bounce up/down in the thermometer, mostly under 1C, but if the CPU temp is up by .5C and the ambient temp is down by .5C at point of recording, then the variance is 1C.

In general, multiple test runs have resulted in no more than about 1C and 1dBA variance. It's quite repeatable; ie, if a test is run 4 times, all the measurements come in within 1 dBA and 1C. I'd say this is excellent consistency.

Given the above, it's probably not safe to say one thing is better than another if there is only 1C or 1 dBA difference between them... but if this difference is across all the test points (ie, fan speeds, SPL points, whatever), then we can safely say one "edges" the other.

On the other hand, there are fans which measure within 1 dBA of each other, yet we have a clear, unquestionable preference for one. This is where, as I said earlier, subjective listening becomes all-important, and SPL by itself cannot be the final arbiter of fan acoustics.

I like the new test concept. The only suggest I would make is to record the initial ambient temperature, humidity and barometric pressure. In theory, higher humidity and/or pressure should increase the cooling ability of the air. (You're measuring the velocity of the air, but not its mass.)

If it turns out that heat rise changes due to atmospheric conditions then a correction factor could probably be used. If it turns out that the temp rise doesn't vary by more than +/- 5%, for a given setup, regardless of the atmospheric conditions, then you can just ignore it. (+/- 10% is probably good enough, but the Scholastics on the forum would have a stroke.)

The data collected looks repeatable, useable, and accurate so the testing is valid. I like the the fact that during the summary you include the hypothesis of what a less restricted heatsink would do and therefore inferred is what a greater restriction would do. I like the RPM being the common factor with the voltage being stated along with the noise. In the community, RPM can be used to measure very easily with enough accuracy to surmise what approximate noise levels we should experience given the same environment factors.

The FPM is the variable that puzzles me as with so many that posted. The blade design of the Scythe and Noctua seems to create higher FPM with lower RPM which does not help the cooling delta. The FPM looks linear with RPM on each fan though not compared to each other. As a couple folks posted about 3D modeling of the exhausted air path would answer that mystery and perhaps an enclosed duct around heatsink. The heatsink does have a high quantity of obstructions in the air path such as the heatpipes which is not a great design as far as an air foil goes.

Thanks for the additional information. I agree with subjective vs objective results regarding SPL and 1dBA differences. 1C repeatability on the other hand is pretty darn good.

Very good initiative!!!

Mike did you miss this:

?

Can you measure the fans at 5V, 7V and 12V also because many of us are using the fans at 5V or 7V without a fan controller?

Any chance we can get a sneak peek at the list of fans you plan on testing.

As far as I know this is only useable for laminar flow anyway. There is not much laminar flow downstream of a fan.

I may be able to help you with Flot, please see this two examples:
http://people.iola.dk/olau/flot/example ... cting.html
http://people.iola.dk/olau/flot/example ... eries.html

I think the combination of the two examples would give a nice and usable interactive chart.


When do you need to have the chart ready?

I've never used Flot, so I invite people who already tried it to express their opinions about it and, if possible, give a hand.

This type of chart would work fine.

As an ex-audio component review, I can testify to the fact that subjective listening is more meaningful when describing noise as opposed to relying solely upon SPL. Some noise characteristics can cause a relatively quiet source to be more disruptive than a benign sound source that's twice as loud.

MikeC: Are there any plans to control PWM fans properly with PWM instead of reduced voltage? Dropping the voltage on a PWM fan is not running the way it was designed.

I posted the same question on the recent CM Hyper 212+ review, but never received a reply.

Yes, as mentioned in the article.

It took a bit of finagling and a lot of time, but we finally got a very useful test board from a fan control IC chip maker whose products are embedded in motherboards. The test sample board is a tool that is used by prospective engineering buyers to experiment with circuit implementations to tweak them for their requirements. See photo below.


There has not been enough to time play with this device, but a major problem is that tiny little blower fan, which cools a very hot resistor and transistor on the PCB. It's noisy as all hell, and we need to replace this with a passive device before the board can be useful for us.

The board does not need to go into a PC, is powered by 12VDC, and is software controlled with a PC through USB.

Wow, that looks nice. Could you fit a northbridge/chipset cooler in there?