Some new test equipment

[Cathar]

33cfm @ 1450 RPM seems weak for a 120mm fan.

That's while sucking air through a radiator. BIG difference between free-air flow and flow through a radiator.

How much does the fan speed & CFM effect the chip (cpu / gpu) temperatures? Are we talking about 1-5C differences b/t 30cfm and 50cfm or more like 8C+ differences?

I'll be measuring this later. It affects the water temperatures though. Your CPU, GPU, etc will go up and down directly proportional to the water temperatures.

[Cathar]

How much does the fan speed & CFM effect the chip (cpu / gpu) temperatures? Are we talking about 1-5C differences b/t 30cfm and 50cfm or more like 8C+ differences?

Okay, we'll use the test data on this graph.

The Nexus at 12v is a ~1000cfm fan, of which being like the curved bladed Yates is a very good fan for the RPM. At 1400rpm (a Yate L with curved blades) would be around mid-way between the dotted orange and blue lines (by extrapolating from my own earlier radiator testing).

So that's a ~0.08C/W at 1400rpm, and ~0.094C/W at 1000rpm (@1.5gpm). So we'd be looking at around a 1.4C increase in water temperature per 100W of heat load. In a CPU only system? ~1.4-2.0C warmer full-load temps.

That's for a 120.1. For a 120.2, divide the differences by 2. For a 120.3, divide by 3.

All approximations only - so take with a large grain of salt.

Overall, if you're after silence, it really pays to have a 120.3, and use slow spinning fans.

We can also see from the Nexus curves that dropping fans below 1000rpm (eg. Nexus at 7v which would be around 600rpm) has a BIG performance hit. The difference between 1000rpm and 600rpm for a CPU only system with a 120.1 would be around 7.5C. It really pays to keep fan RPMs up around 900-1000rpm mark at the very least.

[jinsean]

Great stuff man. Thanks for elucidating the subject at hand. I can't wait for your results =D

That's still pretty good imo: a nexus fan @ ~600rpm dissipating 60w of heat. I wonder how much heat would be dissipated w/o any fans for comparison...

[Cathar]

That's still pretty good imo: a nexus fan @ ~600rpm dissipating 60w of heat.

Remember, that's with a 10C air-water temperature delta. i.e. Your water temps will increase by 1.0C for about every 6W of heat load.

I wonder how much heat would be dissipated w/o any fans for comparison...

If anyone really wants to attempt this, I recommend setting up a thermal chimney. Basically construct a small tower box out of MDF that's around 4' high. Mount your radiator flat near the bottom (about 1' high off the floor) with a middle cut-out section such that air is forced to flow through the radiator. Cut some crennelations into the bottom of your tower where it sits on the floor so air can get in. The idea being that air comes in the bottom from the cool floor, and is forced through the radiator in the middle, and up and out the top. We want there to be some chimney above the radiator because as the radiator warms up, it'll heat up the air above the radiator which will rise and suck in cool air from the bottom.

That would be the best way to achieve effective fully passive radiator cooling, and I reckon that it might just give a 600rpm fan a pretty good run for its money.

The more adventurous could cut crennelations into the top of the chimney, sit a flat board on top (so air can still rise and get out) and then sit your PC on top of your little thermal chimney box.

[jinsean]

@ cathar

I was looking at that graph again and I realized that w/ low CFM/RPM, the flow doesn't effect the heat dissipation as much as high CFM/RPM. I guess it makes sense b/c the bottle neck would be on the amount of heat the fan could blow away from the rad.

This is all great news for me b/c my loop will be a low flow, restriction, RPM setup - ultra quiet =D

No need for a high powered pump when I'll be using low powered fans basically

[Cathar]

Have finished testing the:

CoolerMaster AAF-B12-E1
Tricod Science SPDL1225S
Scythe S-Flex SFF21F

For those of you who were hankering after the Coolermaster, truly, don't bother. On radiator airflow vs rpm performance lags behind the other two, and it excited the radiator shroud almost across the entire rpm range, causing a ringing noise. As a case fan only, and properly isolated from everything else, it'd do a fine job of moving air, but for a radiator application, forget it.

As for the winner of the other two, there is a clear victor, but I'm going to continue testing. I don't want people jumping the gun and running off trying to hunt down some fan before the test is done.

[dinos22]

As for the winner of the other two, there is a clear victor, but I'm going to continue testing. I don't want people jumping the gun and running off trying to hunt down some fan before the test is done.

me guesses tricod

[Vapor]

me guesses tricod

I wouldn't doubt it....the direct-from-Yate YL's and Nexus fans are some of the strongest in my 2nd roundup so far, on a radiator as well (although my radiator testing is a bit more primitive than Cathar's). My understanding of Tricods is that they're essentially Nexus quality, but not at only 1000RPM....which definitely sounds great.

Haven't tested the S-Flex F on the radiator yet, but its CFM/dBA curve was also very strong in open air.

[Cathar]

Tricods are "Yate-like". The frame body is similar, but not identical. The Tricods have a Yate-Loon straight-blade-like design, but it is subtly different too. The blade design performs a fair deal better than the Yate-Loon straight-blade design. The Yate-Loon curved blades (at least at this stage) appears to have an rpm-per-flow advantage, but the Tricod Science motor is different too. The Tricod Science easily has the quietest motor per rpm with the least ancillary vibrations to excite an enclosure. The drawback to them is that they generate a fair deal of blade noise in comparison to some other fans at similar rpm, eg. the curved-bladed Yates, but the Yates have more motor noise.

What would be really interesting would be to swap the blades from a curved Yate to the Tricod Science frame, and test that. I have a spare fan of each type, so I just may do that for purposes of experimentation. Matching the quietest motor to the most efficient blade design.

Okay, here's a couple of graphs of what I have so far. Please don't confuse these with Vapor's graphs. The axis' are switched, so we want to be seeing fans further to the bottom-right to indicate a "good fan". Also, you'll note the fluctuating lines in the 5cm graphs. These fluctuations are caused by varying degrees of excitation of the radiator/shroud/enclosure. Also, for the ambient tests, the noise floor is 31.2dBA, so there is invariably some amount of noise-floor compression occurring once we get below ~33dBA measured noise levels. Just be wary of that.

Plot of air-flow vs ambient noise levels:



Plot of air-flow vs noise level @ 5cm from fan hub:

[kevinbo03]

Go S-Flex, Go!

[dinos22]

wow big difference there cathar

what are you using to plot those graphs??

[Burn]

Great work Cathar- So the Scythe wins in both categories? I might be considering a couple new fans for my GT Stealth 360

[Cathar]

what are you using to plot those graphs??

Gnuplot

So the Scythe wins in both categories?

For as good as the Scythe is, it still excited the shroud+enclosure a fair amount, and generated singing and/or droning noises. Careful rpm selection would find various happy points where the noises were minimised, so unless the Scythe was matched with a non-PWM voltage controller, I would be wary of just slapping it on and hoping for low noise without any annoying excitations. I reckon a PWM controller would just make the droning worse.

The Tricod Science was exemplary in its minimal excitation. While overall it's maybe 1-2dBA noisier, it generated less annoying noises across the range, and the noise it made was more of a low rushing roar than any real singing or droning.

Between the two, despite what the dBA rating says, the Tricod Science played far more nicely. The Scythe is better, but you'd need to be really picking about the exact rpm's you run at it to avoid annoying excitations.

Please don't go rushing out to buy anything until I get more fans testing. I'm able to test about 1 fan/day, and there's still a lot more to get through. Plenty of time for a better choice to rear its head yet.

[Burn]

Could you please explain excitation a little further? I think if we all had a clear-cut definition it would make interpreting results a bit easier.

[dinos22]

Gnuplot



For as good as the Scythe is, it still excited the shroud+enclosure a fair amount, and generated singing and/or droning noises. Careful rpm selection would find various happy points where the noises were minimised, so unless the Scythe was matched with a non-PWM voltage controller, I would be wary of just slapping it on and hoping for low noise without any annoying excitations. I reckon a PWM controller would just make the droning worse.

The Tricod Science was exemplary in its minimal excitation. While overall it's maybe 1-2dBA noisier, it generated less annoying noises across the range, and the noise it made was more of a low rushing roar than any real singing or droning.

Between the two, despite what the dBA rating says, the Tricod Science played far more nicely. The Scythe is better, but you'd need to be really picking about the exact rpm's you run at it to avoid annoying excitations.

Please don't go rushing out to buy anything until I get more fans testing. I'm able to test about 1 fan/day, and there's still a lot more to get through. Plenty of time for a better choice to rear its head yet.

i've tried a few fans myself now and i must say that is a sole reason why i like these tricods.........they are bloody quiet and no funny noise business even at 12V

i bought a couple of 47CFM yate loons....hopefully they are the same

[Cathar]

Could you please explain excitation a little further? I think if we all had a clear-cut definition it would make interpreting results a bit easier.

Every panel or object that the fan is attached to, whether directly or indirectly, will vibrate due to the fan spinning. Every object has a set of natural resonating frequencies, which if you hit that frequency, or get very near to it, the object will excite and start making a racket. Metal panels will buzz loudly, or "sing". Wooden objects (desks) will start to drone loudly.

The old Tacoma Narrows Bridge is an excellent and dramatic example of excitation in action. Once the wind flowing across the bridge starts to vibrate the wires and road at the natural frequency of the bridge, the whole thing will start to resonate and excite wildly.

The thing is, is that these resonating sounds with the fan on the shroud/radiator, and mounted in the test box, may not be extremely loud (although they can be), but they are of a very "pure" tone, emitting a real song note, or drone, or whatever, that is very clear and distinguishable from the general "white noise" of rushing air, and you can hear these noises quite distinctly.

"White Noise" (i.e. rushing wind/static) is generally tolerated far more easily by the human ear. There's no one single note or frequency for the brain to focus on, so we generally just shut it out almost automatically as background noise. Introduce a solitary resonating note though, and the brain will register it very clearly, and it can become annoying quite quickly, because the brain can't shut it out as easily.

Hope that explains it better.

The problem with a dBA rating, is that it's just a sound-wave pressure level. It doesn't really tell us how annoying/focused the sound is. Stick a person in a room with a constant 40dBA white-noise, and they won't really notice it after a while. Put the same person in a room with a constant 30dBA resonating pure note, and they'll quietly go insane.

[Burn]

Nice, thanks for the clarification

[Cathar]

Just added Delta TFB1212GHE results to the graphs. Do a <shift>+<reload> to reload the graphs.

[Cathar]

Quite interesting, the Delta TFB results, at least looking at them from an RPM vs flow perspective (not shown). The TFB, rpm vs flow, absolutely canes all other fans. At 1000rpm it's moving as much air as the Scythe does at 1150rpm. At 1400rpm it moves about as much air through the radiator as the Scythe does at 1600rpm.

The motor noise on the TFB though is astounding. A rasping whining roar that's about as enjoyable as having a hair-dryer on full blast next to your ear. I still can't help but think that if someone successfully managed to marry the TFB's blade design onto a 1300rpm 12cm 25mm thick fan that we'd have a very quiet fan that could be wound down to 800rpm and kick everything else in the rear. As it stands, the TFB's motor and bearings, wired and built as they are to operate an incredible 4600rpm (!!), redefines the definition of an annoyingly loud fan.

[Jedda]

I wasn't looking forward to polishing the al on the coolermasters anyway.
lol
I'd imagine any metal frame fan will tend to yield high excitation on rad. I'd noticed the CM has it's rubber mounts included.

Re the Scythe
That's the 21F, it'll be interesting to see how the 21E lines up?

[Cathar]

Is there a simple way to measure the frequency generated? Both noise and vibration?

Are you asking me for a sound recording? I've been looking into it. No solution as yet.

Next up on the testbed is the curved bladed Yate-Loon D12SL-12. Will be interesting to see how it goes in comparison to the Scythe and Tricod.

[Vapor]

Hahahah, I didn't read that your axes were switched and got all freaked out that my beloved S-Flex F was doing so poorly! They were the first fan I had solidly beat the Sharkoon and was also impressed by the fact they're not sleeve bearings. I haven't really tested my E or D fans yet, just half-tested (open air only) my F and they're already among my favorite for open air.

Would love to see the hybrid Yate/Tricod.

I've also gotten a few problems with excitation on a few of my round2 fans that makes the graph get all screwy in some case and in others my dBA meter doesn't 'notice' it at all. Almost every fan of mine shows some form of it for sure....the S-Flex F was about mid-pack, something I reckon a silicone gasket or screws could nearly-completely cure, while the two best I've had so far were the Nexus (absolutely zero) and the direct-from-Yate SM curved blade.

I have also not come up with a good, consistent, low-cost way of recording sound....been searching for awhile now If I find anything, you'll definitely be one of the first to know Cathar

[ranker]

Hahahah, I didn't read that your axes were switched and got all freaked out that my beloved S-Flex F was doing so poorly! They were the first fan I had solidly beat the Sharkoon and was also impressed by the fact they're not sleeve bearings. I haven't really tested my E or D fans yet, just half-tested (open air only) my F and they're already among my favorite for open air.

Would love to see the hybrid Yate/Tricod.

I've also gotten a few problems with excitation on a few of my round2 fans that makes the graph get all screwy in some case and in others my dBA meter doesn't 'notice' it at all. Almost every fan of mine shows some form of it for sure....the S-Flex F was about mid-pack, something I reckon a silicone gasket or screws could nearly-completely cure, while the two best I've had so far were the Nexus (absolutely zero) and the direct-from-Yate SM curved blade.

I have also not come up with a good, consistent, low-cost way of recording sound....been searching for awhile now If I find anything, you'll definitely be one of the first to know Cathar

I wish I had the same experience as you guys with my SFLEXF fans. Full blast, they were great at moving air w/ acceptable noise levels like my Sharkoons. But any sort of undervolting with my Sunbeam Rheobus at the time would resort in all sorts of clicking/mechanical noises.

[Cathar]

I noticed that the Scythe's have exactly the same frame as the other Adda based fans (e.g. Globalwin, or Acoustfan, or ArcticCooling). I wonder how many similarities are really there.

As for rubber mounts, I ummed and erred over this, and just decided to mount everything without grommets or dampening rubbers. I've been taking notes as to which fans cause the most excitation, and I'll likely go back and revisit those fans with isolating silicone pads.

Rather than fit fans with any provided vibration dampers, I figured that this wasn't a level playing field. I want to see what the base fan is like without damping material.

You are right in that the Scythe F has very low to moderate excitation. The CoolerMaster in comparison had moderate to high excitation almost through the entire range, with only a couple of low points, and it is this which primarily is seen in its poor noise performance.

The bulk of the Scythe's noise/flow performance purely derives from its efficient blade design. It pushes more air through the radiator per RPM than either the Tricod or the Coolermaster, and lower revs means less noise.

What I'm also measuring is power draw too, with the goal of not only finding a quiet fan, but if two fans are very close together, then I'll separate them later and pick a winner decided through their power draw.

[Cathar]

But any sort of undervolting with my Sunbeam Rheobus at the time would resort in all sorts of clicking/mechanical noises.

Sounds like something that a PWM controller would do. I'm using a lab bench PSU. Don't know exactly what electrical mechanism it's using to allow me to vary the voltage, but I know it's not PWM. Alternately you can get weird noises from a rheostat if the power signal has a lot of ripple in it.

Just by wiring some caps (1000uF, 1uF and 0.1uF) across the 12v-ground lines you can cancel out almost all power ripple and should quieten down any fans experiencing strange noises. Should even work okay for fans driven from PWM controllers too, but don't hold me to that.

It wouldn't surprise me at all if most rheobus's you buy don't make any attempt to properly control ripple.