
Synology NAS Lüftertausch: Was das Dezibelmessgerät nicht erfasst hat
I did a Synology NAS fan replacement on my DS1621xs+ because my family kept complaining about the noise in the living room. I could hear the rack from my bedroom, five metres away, through a closed door. The swap took 27 minutes of physical work and about 23 minutes of node downtime. Afterwards the room went silent, and when I compared the sound level meter readings before and after, the number had not dropped at all. It had gone up by 0.8 dB. This article is the full measured record of that Synology NAS fan replacement, including the part where the data contradicted the result, and the part where I admit I bought the wrong fan.
SynoPower Club Point: everyone shopping for quieter NAS fans compares one number, and that number is dB(A). I measured 48.2 dB(A) before and 49.0 dB(A) after. By that number the swap was a failure. By every other measure it was the best thing I have done to that machine all year. What actually changed was that the two fans stopped fighting each other.
What is a Synology NAS fan replacement, and when is it worth doing?
A Synology NAS fan replacement means removing the rear case fans your NAS shipped with and fitting aftermarket ones. On a six-bay unit like the DS1621xs+ there are two 92 mm fans on the back panel, held on with four screws each and plugged into a four-pin PWM header on the chassis. Mechanically it is a twenty minute job with one screwdriver.
The reason people do it is almost always noise. The reason it works is usually not the one they expect. My stock fans were Y.S.TECH FD129225 units rated DC12V 0.32A, and after five years they were caked in dust and no longer spinning at the same speed as each other. That second detail turned out to be the whole story.
It is worth doing if the noise bothers someone in the room. It is not worth doing for temperature, and as you will see below, it can easily make temperature worse.
Was the Synology NAS fan replacement about volume, or about pulsation?
Before starting the Synology NAS fan replacement I logged both fan speeds and the CPU temperature every five seconds, straight off the hardware monitor, and left it running all day. That gave me 420 samples of the old fans and more than 2,500 of the new ones. The number that mattered was not the speed itself. It was the gap between the two fans.
| Measurement | Old fans | New fans |
|---|---|---|
| Average gap between fan 1 and fan 2 | 40.3 RPM | 16.3 RPM |
| Resulting beat frequency | 0.67 Hz | 0.27 Hz |
| Samples | 420 | 2,531 |
Two fans running at slightly different speeds produce sound waves that drift in and out of phase. The loudness rises and falls at the difference frequency. Forty RPM of mismatch works out to roughly 0.67 Hz, which is a swell about every 1.5 seconds. That is squarely inside the band of amplitude modulation that human hearing is most sensitive to, and low-frequency modulation passes through walls and doors far better than steady broadband noise does. Steady noise gets filed away as background. A sound that pulses does not.
How do you empty a NAS before you open it?
My DS1621xs+ is one of two nodes in a Virtual Machine Manager cluster, and it was hosting the VMs behind several live websites, a mail server and a reverse proxy. I could not simply power it off. VMM Pro protection plans let me snapshot each VM, replicate it to the second node, shut it down, switch over, and boot it on the other machine.
| VM moved | Method | Downtime |
|---|---|---|
| Web and shop VM | VMM interface, by hand | 303 seconds |
| Mail and reverse proxy VM | Scripted through the API | 426 seconds |
| Web and shop VM, moved back | Scripted through the API | 192 seconds |
Three to seven minutes per VM, and the node was free. Worth being honest about what visitors experienced: pages served from the Cloudflare edge cache never went down at all, because the edge kept serving them. Anyone trying to log in, add to a cart or check out during those windows was stopped for the full few minutes. Cached browsing and live transactions are not the same thing, and only one of them was protected.
One thing I did not plan for: emptying the VMs does not empty the host. That NAS was also running Surveillance Station with eight cameras attached, and those eight cameras stopped recording for the entire 23 minutes the machine was off. Check what packages are running on the host, not just what VMs are on it.
Synology NAS fan replacement in 4 steps
Move every workload off the unit
Migrate virtual machines to another cluster node with VMM protection plans, and check which packages still run on the host itself. Surveillance Station, Hyper Backup jobs and Download Station all stop when the unit powers down, and an empty VM list does not mean an idle machine.
Shut down and open the rear panel
Use a proper DSM shutdown, never the power switch. Unplug everything, lay the unit on a soft surface, and remove the screws holding the rear panel. The panel carries both fans and stays tethered by the fan cables, so lift it gently rather than pulling it away.
Swap the fans and keep the orientation
Unplug each four-pin header, unscrew the fan from the panel, and fit the new one with the airflow arrow pointing out of the case. Note the thickness of the fan you removed before you buy a replacement, because a thinner fan will fit the screw holes but move much less air.
Reassemble and confirm both fans spin
Refit the panel, reconnect power, and boot. Open the DSM information centre and confirm the thermal status reads normal, then check that both fan speeds are reported and that they are close to each other. A large and persistent gap between the two is what causes the pulsing noise in the first place.
What did the sound level meter actually record?
I took a sound level reading before and after the Synology NAS fan replacement with a phone app, 20 cm from the front of the chassis, before and after. These are not calibrated laboratory figures and I am not going to pretend otherwise. But they were taken the same way on the same day with the same phone, so the comparison between them is fair even if the absolute values are not.
| Reading | Old fans | New fans |
|---|---|---|
| Average dB(A) | 48.2 | 49.0 |
| Sample length | 13 minutes | 1 min 23 sec |
| Room temperature | 27.8 degrees Celsius, 56 percent humidity | |
The new fans measured 0.8 dB louder. Within the error of a phone app that means identical. And yet the subjective result went from two people complaining in the living room and the noise being audible five metres away through a door, to the rack being completely inaudible from the next room. If I had chosen fans by the spec sheet number alone, I would have concluded this swap did nothing.
Why did the CPU run 8 degrees hotter?
This is the part of the Synology NAS fan replacement that went wrong. I bought a Noctua NF-A9x14, which is the low profile 14 mm version. The fans I took out were 25 mm thick. The screw holes are identical, it fitted perfectly, and I did not think about it until I looked at the logs.
| Load average | Old fans | New fans | Difference |
|---|---|---|---|
| 0 to 1.5 | 1801 RPM / 63.7 C | 1506 RPM / 72.1 C | plus 8.4 C |
| 1.5 to 2.5 | 1903 RPM / 64.3 C | 1529 RPM / 72.6 C | plus 8.3 C |
| 2.5 to 3.5 | 2085 RPM / 65.8 C | 1549 RPM / 73.0 C | plus 7.2 C |
Both sets of readings were taken in the same DSM fan mode, binned by system load so I was comparing like with like. The penalty is remarkably consistent at around eight degrees across the whole load range. Drive temperatures stayed fine, between 34 and 40 degrees for the hard disks and 39 to 43 for the NVMe cache, and DSM still reported thermal status as normal. Nothing is in danger. But eight degrees of headroom is eight degrees I gave away for nothing.
Worth being precise about the mechanism: the new fan is not running at its limit. It is rated to 2200 RPM and it was only turning at around 1550. DSM drives fan PWM from temperature, and 73 degrees sits inside the range it considers acceptable, so it never asks for more. Switching DSM from Quiet mode to Cool mode bought only 60 RPM and 0.3 degrees. The eight degrees is a new thermal equilibrium, not a cooling failure.
How does fan desynchronisation make noise carry?
When two sound sources sit close together at almost the same frequency, the combined amplitude swells and fades at the difference between them. This is the same phenomenon a piano tuner listens for when two strings are slightly apart. With fans, the frequencies involved are the rotation rate and the blade passing rate, and the modulation you end up hearing follows the speed difference.
Fans do not leave the factory mismatched by 40 RPM. They drift apart as bearings wear and dust loads up unevenly, which means this problem gets worse gradually over years and you acclimatise to it. I had stopped noticing when it started. I only knew it had been happening once it stopped.
The practical consequence is that a pulsing 48 dB is far more intrusive than a steady 48 dB. Steady broadband noise masks into the ambience of a room within a few metres. A slow amplitude swell keeps drawing attention back to itself, and it loses much less energy passing through a wall. That is why the fix worked despite the meter saying nothing had changed.
Which fan should you buy for a Synology NAS fan replacement?
Not the one I bought. Before any Synology NAS fan replacement, measure the thickness of the fan you are removing first. If it is 25 mm, buy a 25 mm replacement.
| Spezifikation | NF-A9x14 (what I fitted) | NF-A9 (what I should have fitted) |
|---|---|---|
| Thickness | 14 mm | 25 mm |
| Maximum speed | 2200 rpm | 2000 rpm |
| Maximum airflow | 50.5 m3/h | 78.9 m3/h |
| Maximum static pressure | 0.92 mmH2O | about 2.28 mmH2O |
Note that the slim fan spins faster yet moves 36 percent less air, and has well under half the static pressure. Static pressure is the figure that matters in a NAS, because the air has to be forced through a dense stack of drives rather than across an open space. That single specification explains the whole eight degrees.
The deeper point about this Synology NAS fan replacement is that the noise was never caused by the stock fans being loud. It was caused by two worn fans falling out of step. A fresh pair at the original thickness would have fixed the pulsing exactly the same way, kept the full airflow, and probably run cooler than the originals did, because the originals were filthy. The quiet was free. The eight degrees was a mistake I made at the checkout page.
Where can I find more official resources?
If you are planning a Synology NAS fan replacement, Noctua publishes full specification sheets for both the 14 mm and 25 mm versions of the 92 mm fan, including airflow and static pressure curves, which is the data you need before committing to a thickness. Synology documents the fan speed modes in DSM and what each one targets. Your own unit hardware installation guide will tell you how the rear panel comes off, which varies between chassis.
I filmed the swap itself, including the sound level readings taken the same way before and after. The audio is the original recording and is not normalised, so the pulsing in the first section is what my family was actually hearing across the room.
For the migration side, the Virtual Machine Manager documentation covers protection plans, replication and switchover, which is what made the maintenance window possible at all.
I recorded the migration itself in real time with nothing cut, so you can see exactly how long each stage of a protection plan switchover takes.
This second recording shows what the surviving node displays while the other one is powered off, including the protection plans turning red and the cluster reporting itself frozen.
What are the limits and risks of a Synology NAS fan replacement?
Opening the chassis will affect your warranty. Synology does not sell these fans as user-serviceable parts on most models, and if the unit is still under warranty the correct route is a support ticket, not a screwdriver. I accepted that trade on a five year old machine. You should make that decision deliberately rather than halfway through the job.
In any Synology NAS fan replacement the fan headers are four-pin PWM, and the speed signal matters. If DSM cannot read a tachometer from a fan it will raise a fan failure alert and may run the remaining fan at full speed. Fit fans that report speed properly, and check after booting that both are being reported.
Plan the downtime around everything on the host, not just the virtual machines. Mine cost eight security cameras 23 minutes of recording that I had not accounted for, and in a two-node cluster the surviving node freezes rather than failing over, because high availability needs at least three nodes. That is normal, but it looks alarming if you have not seen it before.
Do not assume a thinner fan is a free upgrade because it fits. The mounting pattern is the same at every thickness, so nothing will stop you, and the consequence only shows up in temperature logs days later.
Measure before and after, and log continuously rather than taking single readings. A single spot reading of fan speed tells you almost nothing because the controller moves constantly with load. The eight degree penalty only became visible once I binned thousands of samples by system load.
Finally, remember that ambient temperature sets the floor. These readings were taken at 27.8 degrees in a Taiwanese autumn. In a hotter room, or in summer, the same Synology NAS fan replacement would land eight degrees above a higher starting point, and the margin I am comfortable with today would be thinner.
Verweise
- Noctua NF-A9x14 PWM specifications for the 14 mm airflow and static pressure figures
- Noctua NF-A9 PWM specifications, the 25 mm version used for comparison
- Synology Knowledge Center on fan speed modes in DSM
- Virtual Machine Manager protection plans, used for the zero to seven minute migrations in this Synology NAS fan replacement
- Synology DS1621xs+ product page for the chassis specification
Häufig gestellte Fragen
Does a Synology NAS fan replacement make the unit quieter?
Not necessarily in decibels. In my measurements the sound level average went from 48.2 to 49.0 dB(A), which is slightly louder. What changed was that the two fans stopped running at different speeds, which removed a slow pulsing in the sound. The room became subjectively silent even though the meter did not move.
Will a Synology NAS fan replacement void my warranty?
Opening the chassis generally does affect warranty coverage, and Synology does not list case fans as user-replaceable parts on most consumer and plus series models. If your unit is still in warranty, raise a support ticket instead. I made this swap on a five year old machine that was already out of coverage.
Can I fit a thinner low profile fan if it has the same screw spacing?
It will physically fit, and that is exactly the trap. A 92 by 14 mm fan moves about 36 percent less air than a 92 by 25 mm fan and has less than half the static pressure. In my case the CPU settled about 8 degrees Celsius hotter at every load level. Match the thickness of the fan you removed.
Why do two fans in the same NAS make a pulsing sound?
Because they are not turning at exactly the same speed. Two sources at slightly different frequencies combine into a sound whose loudness rises and falls at the difference between them. My old fans differed by about 40 RPM, giving a swell roughly every 1.5 seconds, which is the range human hearing finds most intrusive.
How much downtime does a fan swap need?
The physical work took me about 27 minutes and the machine was off for 23 minutes. The larger planning question is what runs on that unit. Moving virtual machines to another cluster node took three to seven minutes each, but packages running on the host itself, in my case Surveillance Station with eight cameras, stop for the full outage.
Does switching DSM to Cool mode compensate for a weaker fan?
Barely. When I changed the fan speed mode from Quiet to Cool, the fans gained about 60 RPM and the CPU temperature dropped around 0.3 degrees Celsius. The fan was not near its rated limit either, so this was not the fan running out of capacity. DSM simply considers that temperature acceptable and does not ask for more.
Is 73 degrees Celsius too hot for a NAS CPU?
For the Xeon D in this chassis it is well within specification, and DSM continued to report thermal status as normal. Drive temperatures stayed between 34 and 40 degrees. The concern is not today but headroom: these readings were taken at 27.8 degrees ambient, and a hotter room shifts the whole curve upward.
What should I measure before and after a Synology NAS fan replacement?
Log both fan speeds, CPU temperature and system load continuously rather than taking single readings, because the fan controller moves constantly. The gap between the two fan speeds is the number that predicts whether you will hear pulsing. Take a sound level reading before you dismantle anything, because you cannot go back for it afterwards.