Folks, in the quest for lower server power consumption in light of a small but consistent storj workload (that turns out is excellent at preventig CPU package from entering deep C states) I attemptd to replace lound supermicro fans a quieter model, provided taht the server is idle most of the time, and even drive temperatures have extra 20C to go.
So I
considered competition,
Static Pressure, mm H2O
CFM
Noise at max, dBA
RPM
ARCTIC P8 PWM PST
1.9
23.4
10
200-300
NF-A8
2.4
32.7
17.7
520-2200
Noctua NF-R8
1.4
31.2
17
325-1800
be quiet! Pure Wings 2 80mm PWM
1.9
26.3
19.2
500-1900
Noiseblocker NB-Multiframe M8-P (80mm PWM):
?
31.78
17.6
200-3000
Supermicro FAN-0126L4
?
72.5
53.5
300-7000
Supermicro FAN-0074L4
?
68.3
45
300-5000
Supermicro FAN-0104L4
?
32.9
24
300-2800
and picked recetn revison Noctua A8 PWM, that claim half of throughput of supermicro fans, at max RPM of 2200.
After transplanting the fans into the supermicro cages,
it turned out that Supermicro 7000RPM fan is not only 4 dBA quieter on 3300 RPM, than the queen of silence Noctua at 2200, but also is moving significanly more air. Noctua fans at full blast could not move enough air to keep temperatures within reasonable values – drive termpertaures increased by 14-15C. WTF?
So what I did? I returned Supermicro fans, and just slowed them further down to 2800 (Fixed speed for testing. I probably shall continue tuning my PI controller… I’ll post parameters if I ever get around to it.) At that speed they are barely audible, definitely quieter than Noctua, and temperatures barely moved up (2-3 degrees)
Why woudl anyone buy Noctua at this point? I suspected it was a marketing gimming, but now I know. Hope save someone a few bucks.
So, of the general gist that noctua fans are overly expensive compared to similarly performing fans? Sure, I agree.
However, The supermicro (probably a rebadged delta or something?) has the advantage of being significantly thicker, 38mm vs 25mm. this often gives a big advantage in static presssure, which gives a big advantage is constrained spaces like a 2u rackmount case.
Yep, absolutely. The point is that general notion “noisy case? get Noctua” is deeply flawed. There is no magic.
[quote=“EasyRhino, post:2, topic:30215”]
fans are overly expensive compared to similarly performing
[/quote]
I’d say even compared to much better performing fans.
Why does Noctua not sell thick fans then? This should have solved the issue… I guess then the noise comes back, because again, there is no magic, just marketing and pollution (the amount of plastic shit that came in the boxes besides the fan itself is astounding. It all goes to the landfill for absolutely no reason. What a waste.)
Can’t comment on performance (Noctua did make my old HP Microserver quite a bit less noisy), but next time consider using Noctua Extension Cable NA EC1, changing the pinout is trivial there without destroying cables.
Probably because typical PC cases use the thinner fans. The thicker fans are for servers, but most people who have servers use original fans instead of trying to replace them. Sure, someone like you tried that, but you are probably in a very tiny minority.
By the way, I have multiple servers and they all have original fans, or at least fans specified by the manufacturer (I have one 1U server where I replaced the original Supermicro fans with newer Supermicro fans that have the ability to control speed with PWM).
Mine PWM controls fans, but even most “power efficient” config idles at too-high, too-cold RPM. There seemingly is no way to lower that setpoint, other than turning off fan control from IPMI and re-implemeting fan curves in software. Which I’m hesitant to to for multiple reasons. Mostly, becasue I would not trust my tuning to do the rigt thing under all conditions.
I’ll proably just attach acoustic foam on the walls of the closet and call it a day.
I don’t want rack in any space important enough as to be air conditioned
I already throw money at power company to heat up California 200 joules every second. I don’t want to throw more money at power company to have my AC combat that.
All I want the server to be queiet albeit hot, but not too hot. I don’t think it’s a lot to ask…
First mistake: you ignored your own advice: don’t buy consumer crap, only refurbished server parts worth the money.
Second: Noctua dosen’t have any magic, only pretty collors and smart marketing.
They always try to impress the consumer with nice desings, ignoring the specs. Slimmer fans, nice collored fans, pretty fans…
Just scroll and don’t look at specs… nice fans…
I use them too, btw, and with all those rubber feet and spacers, they are still noisy.
You could look at Corsair too, the ones without bearings, spinning on magnets.
Rubber helps avoiding transfer of vibtations in lighter (in terms of weight) systems, like 99% of consumer PCs. Servers weight quite a lot more, so rubber becomes less necessary. Same with other construction parameters of noise-reducing consumer brands, basically built to cover for deficiences of consumer-grade cases.
Try mounting these server-grade fans in a consumer case and you’ll learn what does “loud” mean
Yeah, each has his own noise limit. I probably can tolerate more noise than you can, so it’s not a problem for me, unless the server is really loud or has an annoying fan control (one server I have runs fans at low speed or full speed, nothing in between and can switch the speeds in a few seconds, that is extremely annoying).
There was a point when I did not have AC (incidentally I also prefer colder temperatures than most people I know), my servers would heat the room to about 40C, fans on the servers would run at 100%, I would have other fans to cool me down, the noise was such that it was difficult to talk. Now it’s really quiet in comparison with the AC keeping the room at 17-19C, though probably you would say it’s too loud.
I have noticed that Supermicro has different part numbers for same size, but different speed fans. Ones I bought for my 1U router were the lower speed, I could hae bought higher speed ones if I wanted, but with the 35W TDP CPU the slower speed fans are good enough.
One of them moves the most air without any noise restriction, one moves a lot of air within a reasonable noise level, and one moves a lot of air at a reasonable noise and price level.
I could not hear from your (very thorough!) review how are arctic fans better than stock?
It seems from your video that there is no difference on full blast, slight tone difference at default profile (probably also at different rpm, and a bit noisier, but more “pleasant”) and on custom profile any fan can be silenced.
I’m also not sure why did you glue the vents in the middle shut, and why was that even an option!
The idea from your other video to wrap inside with some sort of open cell foam to absorb sound is awesome. I completely forgot this was an option, althogh my older server I dampened significantly with dynamat (but that case was cheap chinesse sheet steel turd). Not sure where to get that thin foam though.
From what I understand, the hole in the middle is there to cool the fan motor when it is spinning at full speed. The problem with that at home (as opposed to a datacenter) is that it would soon fill up with dust and cause problems, as the distances in there are small and stuff is moving fast.
Agree, on full blast stock and the Arctics are very similar - but I don’t run them anywhere near full blast. On the custom fan profile, the stocks are more noisy than the Arctics and the tone which which they operate is more unpleasant to me. I rarely care about full-speed performance of fans, since I almost always create custom fan profiles for them. I think this is a stuck habit, back from my watercooling and hardcore-OC days
The vent in the middle of the fan allows extra air to cool the motors directly. Lower Motor hub temperature should significantly aid longevity of the fan (According to Arctics own makrketing material at least). Cooling the motor is great, but it allows dust directly into the fanbearings which is very counter productive to engine longevity.
Thank you on the foam! If I were to try and redo it, i’d just get dynamat, or any other heavy foam. The stuff I used was not great.
