Madshrimps - 120mm Fan Roundup - 35 Models Compared

Control: management of fans, temp/rpm monitoring via soft/hardware

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Madshrimps - 120mm Fan Roundup - 35 Models Compared

Post by MikeC » Mon Feb 16, 2009 8:14 am


jmke
*Lifetime Patron*
Posts: 495
Joined: Thu Apr 15, 2004 3:53 am
Location: In Front of PC
Contact:

Post by jmke » Mon Feb 16, 2009 8:34 am

Thank you for sharingthis review with your readers MikeC, much appreciated, I know it doesn't match your test setup and methodology, but hope it will give some useful data, if not just to compare to findings of others

I saw that SPCR tested the SilenX IXP-74-11 too here: http://www.silentpcreview.com/article739-page3.html
findings are very similar, in my testing it did very well in the performance/noise test, landing near the top when undervolted, but at 12v it has very noticeable motor noise, undervolted still quite audible, even at 5v, although dBA meter tells a different story; good thing SPCR include the subjective noise comments too, because the numbers alone could paint a different picture.

don't know if you guys tested those Noiseblocker fans yet?

pcy
-- Vendor --
Posts: 137
Joined: Sat Aug 13, 2005 2:46 am
Location: Central London, England

Post by pcy » Mon Feb 16, 2009 9:03 am

Hi,


Given the huge number of fans they did test, the absence of Yate-Loon is disappointing. Yate-Loon are a major fan manufacturer, whereas many of the fans that were tested are re-badged and made by someone else. Maybe some were in fact Yate-Loon...


Just for the record, I should declare my interest: I use Yate-Loon fans in the PaQ cases, so I'd like to have seen them included.



I'm struggling to find the justification for the formula used to produce the MadShrimps performance index. It looks like it makes some sort of sense, but I'd prefer to have more explanation of that formula in order to better interpret the results.



Measuring noise at such a close distance and such high levels, rather than using (admittedly much more expensive) better equipment and measuring at 1m or 2m from the noise source may have biased the results. Also, open air sound levels are also not always representitive of real-world results, because the case and the room the computer is in may be much better at eliminating some (typically higher) freqencies than other (typically lower) ones.



I'm also concerned to see CPU temp with a Scythe Zipang being used as the sole determinent of cooling efficenecy for the fans, because the Zipang's fins are very closely packed, which means that it works best with a fan designed to counter substantial backpressure. By contrast, a cooler optimized for quiet running would have fewer, less closely packed, cooling fins, and would work best with a fan less able to counter backpressure, but able to deliver a higher CFM , at given rpm.

These tests, as run, are therfore biased, to some extent, towards fans designed to produce the highest possible levels of cooling with low noise as a secondary issue.

However, I'm not convinced that these standard axial fans are actually designed with such subtle orientations in mind, so the actual scale of any bias may be tiny.





Between them, these factors make it hard to draw any firm conclusions from these tests, IMO.





Peter

jmke
*Lifetime Patron*
Posts: 495
Joined: Thu Apr 15, 2004 3:53 am
Location: In Front of PC
Contact:

Post by jmke » Mon Feb 16, 2009 9:08 am

no single test setup will ever cover all bases; but it would take an enormous amount of time and resources to test pressure, CFM, noise in different situations (as case fan, as radiator fan, as CPU heatsink with small fins, large fin) the possibilities are limitless; you have to draw a line somewhere.

the most important piece that needs to be cooled inside a PC right now is the CPU, case fans you can be more relaxed with performance wise as even a tiny amount of the air suffices in most cases, whereas CPU cooler fans do have to perform or performance will suffer, especially in low noise environment.

I don't disagree with what you are saying, just explaining why I chose to do the setup as is.
Maybe some were in fact Yate-Loon...
Nexus 120mm fans are Yate-Loon I think;)
I'm struggling to find the justification for the formula used to produce the MadShrimps performance index
scatter chart approach delivers best results; but given the amount of samples tested that chart is really too unclear imho, pure numbers alone don't do justice either; the formula used is invented to give some kind of rating, but no definite or final one, just a helping hand to give you an idea of how they match up performance/noise wise.

fri2219
Posts: 222
Joined: Mon Feb 05, 2007 4:14 pm
Location: Forkbomb, New South Wales

Post by fri2219 » Mon Feb 16, 2009 9:32 am

They did explain their rating system in the header above the main bar chart- I don't think anyone on the web seems to have ever heard of Edward Tufte.

While I appreciate their efforts, a scatter chart with some standard deviation bars would have been nice. That said, their results pretty much mirror the consensus from spcr over the last couple of years.

pcy
-- Vendor --
Posts: 137
Joined: Sat Aug 13, 2005 2:46 am
Location: Central London, England

Post by pcy » Mon Feb 16, 2009 9:34 am

jmke wrote:no single test setup will ever cover all bases; but it would take an enormous amount of time and resources to test pressure, CFM, noise in different situations (as case fan, as radiator fan, as CPU heatsink with small fins, large fin) the possibilities are limitless; you have to draw a line somewhere.

I fully understand the problem, so we don't disagree. I've been struggling with the problem of how properly to measure the noise output of my PaQ case for some time now.


the most important piece that needs to be cooled inside a PC right now is the CPU, case fans you can be more relaxed with performance wise as even a tiny amount of the air suffices in most cases, whereas CPU cooler fans do have to perform or performance will suffer, especially in low noise environment.

Absolutely. I didn't make myself quite clear. My objection to the use of the CPU temperture is that it is only reported to 1 degree and tends to fluctuate.

My main concern, though, is the use of the Scythe Zipang rather than choosing a cooler intended for low noise opertion -i.e. with greater cooling fin separation.
I'm struggling to find the justification for the formula used to produce the MadShrimps performance index
scatter chart approach delivers best results; but given the amount of samples tested that chart is really too unclear imho, pure numbers alone don't do justice either; the formula used is invented to give some kind of rating, but no definite or final one, just a helping hand to give you an idea of how they match up performance/noise wise.
Understood. Even so, an explanation of the thinking behind the formula would help, plus, perhaps a sensitivity analysis - if you change the formula in certain ways how much (in relative terms, one fan to another) does it affect the results?

PayBas
Posts: 2
Joined: Mon Feb 02, 2009 2:19 pm
Location: 127.0.0.1

Instant classic

Post by PayBas » Mon Feb 16, 2009 10:14 am

I will certainly be using their monster chart for my next PC (gonna buy one in 1 or 2 weeks).

Simply awesome roundup, thanks.

jmke
*Lifetime Patron*
Posts: 495
Joined: Thu Apr 15, 2004 3:53 am
Location: In Front of PC
Contact:

Post by jmke » Mon Feb 16, 2009 11:01 am

Hey PCY, I wanted a balance of the dBA reading, CPU temp and subjective motor noise; while a scatter chart might balance the first two, adding the third in there is not straight forward, since dBA scaling is not linear and the subjective rating is, unless I starting applying some fancy math there which would make it more complicated then need be.

the formula once completed by our local code guru asked me how much difference in "score" I would like to see between the best and worst fan, I thought a 40-50% difference would seemed reasonable, manipulating the different factors it would be easy to scale it much higher or lower; but gotta stop/start somewhere.

you are correct about CPU temp fluctuating, the temperatures measured are after 1 hour load, max temperature reached and corrected with room temperature at that time; the importance of raw CPU temperature I don't see as the holy grail, as long as the CPU runs stable and does what it's supposed to do without overheating I'm fine with it; so a drop of 1°C in temps by swapping fans is not as important as a drop of 1dBA in noise; so I wanted to put more weight on the dBA reading for the score, or inversed: less importance to the CPU temp; in the total score the CPU temp measures for 66%; this number can of course be lowered, the raw data sheet is easily manipulated and can tip it more towards the silence side of things; there are soo many variables to contemplate... nevertheless we (credit to Massman who helped with the score math and tweaking) think this score balances performance and noise for the average user;

those seeking high performance no matter the cost have it easy, 4000rpm 120mm Delta fan and you're settled; but as soon as you remove that "no matter the cost" you come into a gray area of how much noise somebody can tolerate before it becomes annoying; MikeC hears the buzzing the power adapter of a device 2m away and searches out that noise in order to kill it and not be annoyed; others can live with a 2000rpm 120mm fan in their case and never bother to ask if it can run quieter, it just doesn't bother them to a point where they'll start searching for a solution. so that's a lot of area to cover between the "silent" and "performance" users; no matter if you are performance/silence orientated, dBA >>> CPU temp any day of the week; if your day-to-day system run only stable with a CPU temp which can only fluctuate a few degrees you have other issues:)

I'm typing too much here :)

total score as mentioned in the article is an invention to mainly help me sift through the dba/cpu/motor noise results and make sense of it; I might do a flash version of the data so the end user can manipulate the importance of certain scores to give them more weight and thus change the ranking. my skills are not high enough to do this, had to ask help already for the HSF chart generator (link in sig) and it's far from ideal still.

anyway, appreciate your feedback and input. thank you for your time.

Worker control
Posts: 126
Joined: Wed Nov 16, 2005 12:16 pm

MicroCenter / Scythe

Post by Worker control » Mon Feb 16, 2009 7:29 pm

I stopped by a MicroCenter today and was pleasantly surprised (even shocked!) to see that they had a range of Scythe fans for sale. Perhaps they've been reading MadShrimps and SPCR? :)

In any case, nice to see that my ears aren't lying when they tell me my 850RPM SlipStream is really quiet.

Otto69
Posts: 43
Joined: Tue Sep 18, 2007 12:38 pm

Post by Otto69 » Mon Feb 16, 2009 8:29 pm

I'll stick with my Noctuas thank you very much. They are the quietest at low speed for the most air moved I've found. Too bad they aren't PWM.

jmke
*Lifetime Patron*
Posts: 495
Joined: Thu Apr 15, 2004 3:53 am
Location: In Front of PC
Contact:

Post by jmke » Tue Feb 17, 2009 12:36 am

Otto69 wrote: They are the quietest at low speed for the most air moved I've found. .
as a casefan :)
Last edited by jmke on Tue Feb 17, 2009 3:17 am, edited 1 time in total.

pcy
-- Vendor --
Posts: 137
Joined: Sat Aug 13, 2005 2:46 am
Location: Central London, England

Post by pcy » Tue Feb 17, 2009 2:52 am

Hi,

jmke wrote:
Otto69 wrote: They are the quietest at low speed for the most air moved I've found. .
as a casefan :)[/i]
By "as a casefan" do you mean:

1. A CPU cooler fan is working against a much higher backpressure than a case fan so the performance rankings might change?


2. Or are you simply referrring to the available electical connections?




Peter

jmke
*Lifetime Patron*
Posts: 495
Joined: Thu Apr 15, 2004 3:53 am
Location: In Front of PC
Contact:

Post by jmke » Tue Feb 17, 2009 3:17 am

the Noctua NF-S12 series performs wonderful in open air, or as a case fan with not a lot of restriction; on a heatsink it's a different story, the NF-P12 does better here.
Noctua NF-P12 1300rpm
Noctua NF-S12-1200 1200rpm

temps: 5/7/12v
Noctua NF-P12 56 48 45.5
Noctua NF-S12-1200 55 51 46.5

dBA: 5/7/12v
Noctua NF-P12 37.1 42.6 49.1
Noctua NF-S12-1200 38.1 42.9 49.6
although NF-P12 is rated higher at 1300rpm, is less noisy on the dBA meter as well as subjective motor noise; CPU temp is noticeably lower at 7v, lower at 12v, slightly higher at 5v, but larger noise difference.
1. A CPU cooler fan is working against a much higher backpressure than a case fan so the performance rankings might change?
yes exactly; the SPCR tests are closer to results based on fans used as case fans since they are tested in less/no restriction setup.

pcy
-- Vendor --
Posts: 137
Joined: Sat Aug 13, 2005 2:46 am
Location: Central London, England

Post by pcy » Tue Feb 17, 2009 3:59 am

Hi,


Intereting.


Those temps were with the Scythe Zipang? so the balance might shift when using a cooler with more open fins?


You're up early, or do yo ulive in Europe?



Peter

jmke
*Lifetime Patron*
Posts: 495
Joined: Thu Apr 15, 2004 3:53 am
Location: In Front of PC
Contact:

Post by jmke » Tue Feb 17, 2009 4:04 am

Hey PCY,

on a Scythe Ninja or similar product NF-S12 will do better.


btw, site URL should be a dead-giveaway, madshrimps.BE :D
1pm here atm.

pcy
-- Vendor --
Posts: 137
Joined: Sat Aug 13, 2005 2:46 am
Location: Central London, England

Post by pcy » Tue Feb 17, 2009 4:16 am

Indeed,


I could have sworn I typed in madshrinps.com



So if I sent you some Yate Loon 120 or 140mm fans would you want to include them on your test.


I'm currently testing a Xigmatek S1384 and S1283 for use in my 5U format PaQ cases, both with 140mm fans rather than 120.


The theory is that the 1283 will work better on Q9xxx CPUs but the 1284 will work better on the i7s because the larger die size means all four piped have good thermal contact and work effectively.



Peter

bluepanda
Posts: 7
Joined: Mon Sep 25, 2006 5:31 am
Location: Australia

Post by bluepanda » Tue Feb 17, 2009 4:42 am

As much as i would like a SPCR comparison to make my decision on according to this my next fans will probably be Scythe SY1225SL12L's. They work sweet with no undervolting required and I can get them at my favourite online store. :-) Handy because i need 10.

They are also cheaper and much the same performance wise as nexus fans which i have and love. Awesome.

jmke
*Lifetime Patron*
Posts: 495
Joined: Thu Apr 15, 2004 3:53 am
Location: In Front of PC
Contact:

Post by jmke » Tue Feb 17, 2009 5:16 am

So if I sent you some Yate Loon 120 or 140mm fans would you want to include them on your test.
my fan test setup has been disassembled for CPU cooler testing which will last more than a month; will have to look into another test setup/method in the future, depends on how much money I can invest in the site this year.

ame
Posts: 488
Joined: Tue Jul 10, 2007 10:35 pm
Location: Israel

Post by ame » Tue Feb 17, 2009 6:22 am

Thanks for the link.
Most of the fans on the list are unavailable or too pricey here.

I build systems for others but rarely OC and honestly don't care if temps are 29/45 or 35/56 (idle/load) so performance is less of a factor for me.
Noise is however, so if a fan runs 800 RPM inside a case I just assume its low noise enough for my clients.
Case fans I typically replace with AC PWMs 500-700 RPM motherboard controlled, or Scythe S-flex 800s depending on budget (each s-flex is $18 :evil: ).

jmke, the 160+ CPU cooler article I just stumbled upon from your sig is IMO the best CPU cooler comparison tool I have encountered so far.
Just used it to choose OCZ Vendetta over AC freezer 7 pro, A choice I had already made but now validated.

Olle P
Posts: 711
Joined: Tue Nov 04, 2008 6:03 am
Location: Sweden

Post by Olle P » Tue Feb 17, 2009 7:23 am

pcy wrote:Measuring noise at such a close distance and such high levels, ... may have biased the results. Also, open air sound levels are also not always representitive of real-world results, because the case and the room the computer is in may be much better at eliminating some (typically higher) freqencies than other (typically lower) ones.
Given their available test equipment I think they did the right choices, with one detail omitted.

- The relatively insensitive audiometer made measuring close up a must.
- Measuring close up will also reduce influence of background noise and "filtering".
- The detail they forgot was to deduct background noise from the presented noise value. (Relevant for readings <45dBA.)

- To get the equivalent noise value for 0.5m subtract 20dB, for 1.0m subtract 26dB and for 2.0m 32dB.

- The formula used for "efficiency" can be discussed, or better still dropped altogether.

Cheers
Olle

jmke
*Lifetime Patron*
Posts: 495
Joined: Thu Apr 15, 2004 3:53 am
Location: In Front of PC
Contact:

Post by jmke » Tue Feb 17, 2009 7:36 am

or better still dropped altogether.
yup, definitely not needed, it was more a tool for me to sift through the results; the scatter chart verifies most of those findings though, I made a bigger version of the original here
jmke, the 160+ CPU cooler article I just stumbled upon from your sig is IMO the best CPU cooler comparison tool I have encountered so far.
should be more dynamic, with some nice flash or ajax, but not my strongest, already happy I could get it to generate a PNG chart:)

Worker control
Posts: 126
Joined: Wed Nov 16, 2005 12:16 pm

Post by Worker control » Tue Feb 17, 2009 11:23 am

bluepanda wrote:As much as i would like a SPCR comparison to make my decision on according to this my next fans will probably be Scythe SY1225SL12L's. They work sweet with no undervolting required and I can get them at my favourite online store. :-) Handy because i need 10.
SPCR review here:
http://www.silentpcreview.com/article832-page4.html

Aris
Posts: 2299
Joined: Mon Dec 15, 2003 10:29 am
Location: Bellevue, Nebraska
Contact:

Post by Aris » Tue Feb 17, 2009 1:36 pm

its nice to see em all listed, but its really doesnt do us much good without individual subjective noise evaluations.

I'll likely stick with my Sflex 1200. I may try one of the 800rpm slipstreams though. I've heard people on these forums recommending them as well.

Plus *everyones fav online store* carries the full line of Scythe fans. That alone should give every scythe fan reviewed a bonus 1000 points.

guises
Posts: 55
Joined: Thu Oct 26, 2006 11:48 am

Post by guises » Tue Feb 17, 2009 1:40 pm

Hey, I'm a little confused as to how the motor noise rating was applied to some of your results. The formula says that the subjective noise rating was simply subtracted, and that holds for some of the fans on the graph, but for others it looks like two times the noise rating was subtracted.

For example, the Scythe SA1225HP12LVR @ 7v gives me 137.12 unadjusted with your data, just like you've got, but the subjective motor noise is 3 and you're giving the weighted value as 131.11.

jmke
*Lifetime Patron*
Posts: 495
Joined: Thu Apr 15, 2004 3:53 am
Location: In Front of PC
Contact:

Post by jmke » Tue Feb 17, 2009 1:54 pm

good catch, yes I multiplied the subjective motor noise scare by 2, will adjust formula asap to reflect the one I used in Excel.

its nice to see em all listed, but its really doesnt do us much good without individual subjective noise evaluations.
pages 2,3,4?

guises
Posts: 55
Joined: Thu Oct 26, 2006 11:48 am

Post by guises » Tue Feb 17, 2009 2:24 pm

jmke wrote:good catch, yes I multiplied the subjective motor noise scare by 2, will adjust formula asap to reflect the one I used in Excel.
But you didn't for all of them. Some of them, like the NoiseBlocker M12-P @ 5v, don't have the motor noise multiplied. Did you have multiple columns or something, and forget to change it for all of them?

I've done that before.

jmke
*Lifetime Patron*
Posts: 495
Joined: Thu Apr 15, 2004 3:53 am
Location: In Front of PC
Contact:

Post by jmke » Tue Feb 17, 2009 2:29 pm

I'll look into this.

edit OK I recall now, was a long friday, I penalized those fans with 3/4 rating much harder as they were definitely not silent; fans with motor noise rating 1=dead silent even with your ear next to them, 2=silent from >30cm. 3=irritating in a quiet room, no matter where you are , 4=to avoid.

so I deducted motor noise *1 for 1/2 and *2 for 3/4 to really mark them down for having a noticeable and irritating noise, even though performance/noise ratio based solely on the dBA meter and CPU temps said they are better.

will update the article to reflect this, thank you for double checking that list and spotting this oversight :oops:

guises
Posts: 55
Joined: Thu Oct 26, 2006 11:48 am

Post by guises » Tue Feb 17, 2009 2:50 pm

Sure. Pedantry comes naturally to me.

jmke
*Lifetime Patron*
Posts: 495
Joined: Thu Apr 15, 2004 3:53 am
Location: In Front of PC
Contact:

Post by jmke » Tue Feb 17, 2009 2:55 pm

had to look up that word :)

bluepanda
Posts: 7
Joined: Mon Sep 25, 2006 5:31 am
Location: Australia

Post by bluepanda » Wed Feb 18, 2009 12:35 am

Worker control wrote:SPCR review here:
http://www.silentpcreview.com/article832-page4.html
Cheers dude. Im almost ashamed i missed that that review. :-P That seals it, next decent pay check im going to replace every fan in my two servers and the server cupboards exhaust fans with slip streams.

Post Reply