First attempt at creating a silent server

Show off your quiet rig.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
SixedUp
Posts: 16
Joined: Sat Jun 02, 2007 9:44 am
Location: Hampshire, UK

First attempt at creating a silent server

Post by SixedUp » Sat Jul 28, 2007 11:42 am

Some time ago I was given a fully functional (and original) Shuttle SV25. The size of the machine makes it perfect to fit on a bookshelf next to my printer, where I hope to leave it running as a headless Linux server, 24x7.

Image

The problem was that in its original condition, this was a very noisy (and hot) little box, and I wanted it to be as near silent as possible. As far as I can tell, it was one of the earliest machines, with a Pentium 3 1.2GHz Tualatin, with a standard CPU heatsink/fan combination and 3 fans ... a 40mm in the PSU, a 60mm on the heatsink, and another 60mm on the case. None were particularly quiet. The PSU fan in particular was a screamer, so I swapped that for a new Scythe one. The CPU heatsink fan was a high flow rate design, that ran at high speed and produced a lot of air and mechanical noise, so I swapped it for an Acoustifan, but the airflow noise from it was still way too high to be bearable long-term, and the CPU temp jumped 19c (to 55c) at idle ... not a good indicator for likely performance under load on a warm day :)

I also noticed that some of the capacitors on the motherboard were showing signs of leakage (which I understand is not uncommon), so taking all that into account, I decided to “upgradeâ€

bonestonne
Posts: 1839
Joined: Mon Feb 05, 2007 2:10 pm
Location: Northern New Jersey
Contact:

Post by bonestonne » Sat Jul 28, 2007 11:52 am

it looks pretty nice...i wish i had a box that small [would help in my room, full size ATX stands out like no other].

as nice as it looks though, i'd probably work on the inner grill in the front to allow more intake, and i'm not sure i'd keep the fan thats so loud...i might want to try to fit a larger fan elsewhere.

SixedUp
Posts: 16
Joined: Sat Jun 02, 2007 9:44 am
Location: Hampshire, UK

Post by SixedUp » Sat Jul 28, 2007 12:19 pm

Yeah, I suspect a large part of the noise is now down to intake too. Unfortunately, the front panel is overlaid by a thick, shaped (and beautifully finished) piece of aluminium, so cosmetically-acceptable alterations are not going to be easy. I was wondering about making a large number of holes at the front of the floor of the case to try to improve the flow.

However, at the moment I'm considering just pulling the case fan altogether and watching what happens to the case temperature. Since I'm actually putting out so little heat, I'm wondering if I even need it any more.

The other thought is that if I get the CrystalFontz drivebay kit, I can get that with a comprehensive fan controller kit for very little extra $$$, which would allow me to monitor the temps, and control all the fans under software. Lots of options still.

bonestonne
Posts: 1839
Joined: Mon Feb 05, 2007 2:10 pm
Location: Northern New Jersey
Contact:

Post by bonestonne » Sat Jul 28, 2007 12:40 pm

it might even be possible to duct a 120mm fan outside of the case to pull air out of the 60mm hole...that would keep the temps down, and give few problems...aside from mounting, ducting and placing...

i'd just worry about the hard drives...they tend to get warm and stay warm.

SixedUp
Posts: 16
Joined: Sat Jun 02, 2007 9:44 am
Location: Hampshire, UK

Post by SixedUp » Sat Jul 28, 2007 1:56 pm

After 6 hours the temperatures seem really stable, and well within tolerances:

CPU: 26c
Case: 27c
Lowest (boot) disk: 37c
Middle (500GB): 38c
Upper (500GB): 38c

Its as stable as you could wish for at the moment. Really tempted to try turning off the case fan ... :)
Last edited by SixedUp on Sat Jul 28, 2007 2:38 pm, edited 1 time in total.

bonestonne
Posts: 1839
Joined: Mon Feb 05, 2007 2:10 pm
Location: Northern New Jersey
Contact:

Post by bonestonne » Sat Jul 28, 2007 2:06 pm

you should look up the model of your fan [by google works] and find the CFM it pushes...if its below 15, i'd say its hardly worth running at all. that hole in the back could easily make ambient possible in the case.

SixedUp
Posts: 16
Joined: Sat Jun 02, 2007 9:44 am
Location: Hampshire, UK

Post by SixedUp » Sat Jul 28, 2007 2:47 pm

It only does 13cfm at 12v. Down to 9cfm at 7.5v. No data for "5v", but I think I screwed up there as the fan is still spinning at nearly the speed it should be for 12v, so I suspect I'm not dropping enough voltage across my resistors ... which may explain why I've still got quite a lot of fan noise. Need to get the multimeter out again :)

bonestonne
Posts: 1839
Joined: Mon Feb 05, 2007 2:10 pm
Location: Northern New Jersey
Contact:

Post by bonestonne » Sat Jul 28, 2007 6:12 pm

is it molex or 3pin?

Image

taken from: http://www.silentpcreview.com/article63-page1.html

at 13cfm, you might as well just put a swarm of gnats in the hole.

jaldridge6
Posts: 319
Joined: Wed May 09, 2007 2:31 am
Location: Hell

Post by jaldridge6 » Sat Jul 28, 2007 6:36 pm

actually, those little scythe fans are quite effective. I had one on a swiftech MCX159 northbridge cooler on a 650i chipset............... without the fan, it heated up to >60C. With the fan, it stayed around 45. So....... 13CFMs gotta account for somethin ;)

My conclusion. An insignificant amount of airflow is almost all that is needed for a lot of components. True passive leads to very hot parts.

SixedUp
Posts: 16
Joined: Sat Jun 02, 2007 9:44 am
Location: Hampshire, UK

Post by SixedUp » Sun Jul 29, 2007 1:10 am

bonestonne wrote:is it molex or 3pin?
I'm driving it off one of the 3-pin headers on the motherboard (hence messing around with resistors, rather than just taking 5v from the PSU loom as you showed). I've no more molex connectors free, and would prefer not to start adding "Y" splitters in now if I can avoid it - there's really not much space left in there for cabling.
... at 13cfm, you might as well just put a swarm of gnats in the hole.
Well, I can feel a noticeable draft coming out the machine, so I guess they're fairly strong gnats (!), but I do tend to agree with you. I think when I have a little spare time I'll write a shell script to monitor the temps and shutdown the box if it gets too hot. Then I can pull the case fan and leave it running for a while to see how it gets on.

Incidentally, the 40mm Scythe fan I'm using on the CPU heatsink is practically inaudible. Its a 4.8cfm unit, rather than the 5.6cfm stock part, so I guess there is some loss of cooling performance but it is significantly quieter. Very impressive.

SixedUp
Posts: 16
Joined: Sat Jun 02, 2007 9:44 am
Location: Hampshire, UK

Post by SixedUp » Sun Jul 29, 2007 4:14 am

Ok, I pulled the case fan. Temps rose steadily, until I decided to stop it. Last reading (after about 15 minutes) was :

CPU: 37c
Case: 40c
Lowest (boot) drive: 52c
Middle (500GB) drive: 47c
Top (500GB) drive: 46c

It's all within specifications, but I'm not comfortable running it that hot - I'd rather have the noise from the case fan :)

SixedUp
Posts: 16
Joined: Sat Jun 02, 2007 9:44 am
Location: Hampshire, UK

Post by SixedUp » Sun Jul 29, 2007 11:14 am

I mentioned earlier that I'd noticed that at "5v" the case fan was producing almost full RPM. Worked back through my calculations for the voltage drop resistors, and found my mistake. I was only dropping the voltage to about 10v --- no-where near 5v. I've made up a new lead with a more sensible drop-resistor in it, and the case fan is now getting 5.5v ... noise is much better now, with just a slight whirring to be heard. Subjectively it's similar in level to when my laptop's fan is running.

After an hours uptime the temperatures now seem to be stabilising at:
CPU: 29c
Case: 33c
Lowest (boot) drive: 45c
Middle (500GB) drive: 44c
Top (500GB) drive: 43c

The drives are still fluctuating a bit; they may even still be creeping up a few degrees. Ideally I'd actually like to see the drives a bit cooler, but to do that and maintain the current noise level I think I would need to fabricate a new rear panel to accommodate something like a 92mm fan, and make a lot of changes to the inlet side of things too.

Where I am now feels like a reasonable compromise; although it's not silent, it is very quiet, and all the temperatures are within specification. I just need to get on with installing all the software now :)

jaldridge6
Posts: 319
Joined: Wed May 09, 2007 2:31 am
Location: Hell

Post by jaldridge6 » Sun Jul 29, 2007 5:18 pm

the majority of the noise is probably coming from the harddrives. those 501ljs are much louder than say, a nexus 120mm fan at 12v.

SixedUp
Posts: 16
Joined: Sat Jun 02, 2007 9:44 am
Location: Hampshire, UK

Post by SixedUp » Wed Aug 08, 2007 7:35 am

I just couldn't help myself ... I had to have another attempt at getting this quieter. And that's when I discovered that with the cover off, the system is actually as close to silent as makes no odds :shock:

I wondered about vibration causing the case to resonate, but the panels are small, and very tighly held in place, and if I damp them (by pressing them in place) there is *no* change to the noise (volume or tone). In addition, the case fan moves *much* more air with the covers off.

So, could it be that all the remaining noise is down to the case having practically *no* air intakes --- see the tiny holes on the front panel? I did some rough calculations, and I reckon the total diameter of my air intakes are less than 20% of my air outlet. Which would mean the air would have to travel through them 5x faster to keep up with the case fan. Would that explain the noise with the covers on, and also the reduced airflow? Have I build a rather poor trumpet, as well as a fairly neat server?! :roll:

So I'm now wondering about putting 20-odd x 10mm holes into the floor of the case, right at the front edge. That ought to balance the inlet / outlet diameters, and allow more air to be drawn through the case too. Anyone know if that's likely to help the noise, or just prove to be a massive dust-magnet?

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

Post by MikeC » Wed Aug 08, 2007 7:47 am

There are probably 2 reasons for it being louder with the cover on --

1) The fan is actually working harder against more airflow impedance.
2) With the cover closed, you have a cavity which has air resonances. The vibration and noise of the HDD and the fan together are "exciting" those resonances. Hence, even though the cover doesn't vibrate, the air inside is vibrating -- which is what you hear. The same phenomenon as when you talk in a small shower.

SixedUp
Posts: 16
Joined: Sat Jun 02, 2007 9:44 am
Location: Hampshire, UK

Post by SixedUp » Thu Aug 09, 2007 4:52 am

Thanks Mike, so less like a trumpet, and more like a drum. I should have realised that.

Anyway, more holes to let more air in should help with (1), and some sort of acoustic damping foam glued to the inside of the covers ought to help with (2).

I'll give that a try tonight :)

SixedUp
Posts: 16
Joined: Sat Jun 02, 2007 9:44 am
Location: Hampshire, UK

Post by SixedUp » Sun Aug 12, 2007 6:00 am

I added a series of 30 x 10mm holes to the front of the floor of the case, which brings the inlet side up to about 80% of the total diameter of the outlet. This turns out to be nearly a 10x improvement over the stock inlets. I also installed some nice soft open-cell packing foam around the sides of the system to help cut down on the resonance of the case.

The foam probably counteracts the extra airflow somewhat, as it does press closely around the drives in particular. However, the overall result is another nice drop in the noise levels, and a small improvement in the HDD temperatures.

Now seeing (stable) temps of:
CPU 30c
Case 33c
Lowest drive (2.5") 37c
Middle drive (3.5") 40c
Top drive (3.5"0 40c

Noise levels are now down to barely audible. I think that (for now!) I'm going to declare it "complete", get it loaded up with software and start living with it for a while. So, a big thanks to everyone who helped me out with this, for providing all your help and advice.

SixedUp
Posts: 16
Joined: Sat Jun 02, 2007 9:44 am
Location: Hampshire, UK

UPDATE:

Post by SixedUp » Tue Sep 04, 2007 2:44 am

So about a week into running the server, the 2.5" Seagate "boot" drive (the lowest in the stack) failed. Lots of unrecoverable errors, and then a complete failure to spin up. Since this held the OS, this took down the whole server, and I lost all my software installation/configuration work. My supplier RMA'd it, and sent me out a new one with only a couple of days turn-around, but I'd lost several evenings work at a stroke. Although I doubt this failure was anything to do with the way I'd used the drive, I set up a bunch of temperature monitoring/logging, and some cron scripts to shutdown the entire system if any of the temps ever went too high.

But another couple of weeks in and I've already seen filesystem corruption on that replacement drive too, along with some strange "clicking" sounds from the drive. The drive temp has never been beyond a peak of mid-40ish celcius (under heavy load) and SMART reports that the drive is working fine, but it's clear that there have been some problems. To say that I'm now suspicious of this replacement Seagate drive would be an understatement. Meanwhile the big 500GB Samsungs have been performing flawlessly. So I did some more measuring, and went back and reread all the specs again. And replaced the Seagate with a 3rd 500GB Samsung :shock:

Space is incredibly tight. So much so that I had to make up some brackets to hold the drives in a tight stack with 1/4" airgap between each of them. The whole lot is then suspended in a similar manner to before. The Pico PSU is fine with the load, but I'm running my AC-DC brick close to its maximum for the 15 seconds or so while the the three drives are spinning up. I've now reinstalled the whole system (OS, Swap and data) onto a combination of raided partitions across the three drives to build (I hope) a completely disk-fault tolerant system. In theory I can lose any one drive and the system should continue running. The extra drive allows me to use RAID5 too, so I now have 1TB available for data rather than just the 500GB before.

Interestingly the temperatures of the drives have dropped when compared to the old arrangement. I think this is because the aluminum brackets that hold the drives in a stack are acting as a simple heatsink, and helping to cool the drives. Case temperature is (unsurprisingly) up a couple of degrees, but still generally in the low-mid 30's celcius. Noise levels are basically unchanged.

klankymen
Patron of SPCR
Posts: 1069
Joined: Thu Aug 04, 2005 3:31 pm
Location: Munich, Bavaria, Europe

Post by klankymen » Tue Sep 04, 2007 5:16 am

AWESOME build!

partially cause I was planning a VERY similar build myself, and was surprised to see how close it is to your machine when I opened this thread.
In case you're interested, though I dont want to steal your thread, I'm thinking about building my own case, for better airflow and even smaller form factor - make a few sketchup drawings: one & two.

Anyway I have a few questions:
As far as I can tell the motherboard only has 2 sata ports, now that you're running 4 (or is it 3? did you take out the laptop drive completely?) drives, have you started using a raid controller?

And as for running RAID5, why is it that you feel the need to use RAID? is it for performance or for reliability reasons?

And also, what size brick are you using? is it passive? I couldn't find any info in your posts. Do you have a power meter to measure the startup draw? allthough I'm assuming you don't reboot much if it's a server.

One more thing that got me thinking, Jetway claims the 1.5GHz model draws 25W, whereas the 1.2GHz model only draws 7W. As I doubt 300MHz makes a big difference, I would tend to the low power model, but I can't find the 1.2 anywhere in germany.... I wonder if the wattage difference really is that significant, or if they're exaggerating.

SixedUp
Posts: 16
Joined: Sat Jun 02, 2007 9:44 am
Location: Hampshire, UK

Post by SixedUp » Tue Sep 04, 2007 8:00 am

klankymen wrote:AWESOME build!
Thanks :)
partially cause I was planning a VERY similar build myself, and was surprised to see how close it is to your machine when I opened this thread.
In case you're interested, though I dont want to steal your thread, I'm thinking about building my own case, for better airflow and even smaller form factor - make a few sketchup drawings: one & two.
I thought about that too, though since I've managed to get this working acceptably I'm not going to go down that route until I need more drives, which I'm hoping won't be too soon.
Anyway I have a few questions:
As far as I can tell the motherboard only has 2 sata ports, now that you're running 4 (or is it 3? did you take out the laptop drive completely?) drives, have you started using a raid controller?
I took the laptop drive out completely. I have 2 native SATA II ports on the motherboard, and a dual channel PATA-SATA converter, which converts the PATA connector on the motherboard into another pair of SATA I ports. I'm using 3 ports for the 3 x 500GB drives. It's not really ideal ... but I couldn't find a mini-ITX board with more than 2 native SATA drives, and cabling with PATA cables would have really obstructed airflow. In theory I could fit a SATA raid card, but in practice there just isn't room in the case, and I'd rather stick with software raid for compatibility reasons anyway.
And as for running RAID5, why is it that you feel the need to use RAID? is it for performance or for reliability reasons?
Reliability. This is going to hold my entire music collection, the whole families digital photo collection, and act as a backup server for all the other machines in the house. It will also run various servers, including my websites and mail server. I really don't want to lose that data ...
And also, what size brick are you using? is it passive? I couldn't find any info in your posts. Do you have a power meter to measure the startup draw? allthough I'm assuming you don't reboot much if it's a server.
Its rated at 12v 8.5A (102w). I thought it was completely passive, but as I discovered when I did a reboot recently, there is actually a small fan in it. I've only heard it come on the once, when it was probably already warm, and I then needed it to spin up all three drives at once. It was bought from LinITX.

I didn't measure the actual power draw at the plug yet - I keep meaning to borrow a meter from a friend and do that. However, I did calculate all the component draws from their specification sheets, and sized the brick from that. Eg, each 500GB drive draws 12w in normal operation, but draws 25w (mostly on the 12v line) during spinup. 12w for the motherboard, 2w per fan (max), and so the operating draw should be around 50w now. During spin-up it should be a little over 90w.
One more thing that got me thinking, Jetway claims the 1.5GHz model draws 25W, whereas the 1.2GHz model only draws 7W. As I doubt 300MHz makes a big difference, I would tend to the low power model, but I can't find the 1.2 anywhere in germany.... I wonder if the wattage difference really is that significant, or if they're exaggerating.
I was looking at the 1.2GHz passively cooled J7F4 board too, but there have been some real reliability problems with that board. I know of several retailers who are not currently selling them because of that. In the end I bought the 1.5GHz fanned J7F4 instead, and it's been very reliable. That motherboard comes in two versions using different C7 processors. One draws 25w (uses the "desktop" version of the C7), and the other draws 12w and uses the "normal" C7. I got the 12w version. It's never going to be a high performance system, but that's not what I needed ... and by keeping the power-draw low it's a lot cheaper to run it 24x7

klankymen
Patron of SPCR
Posts: 1069
Joined: Thu Aug 04, 2005 3:31 pm
Location: Munich, Bavaria, Europe

Post by klankymen » Sat Sep 08, 2007 9:46 am

OK, I see what you mean about the desktop board, the model number has a D at the end: J7F4K1G5D vs J7F4K1G5.

One question, how are you managing the power wires? AFAICT the pico-psu has 2 molex connectors and a floppy, are you using some kind of splitters? or are you running one of the drives off the floppy wire?

I have a vague recollection there is such a thing as a 1xMolex-->2xsATA adapter, are you by chance using such a thing?

SixedUp
Posts: 16
Joined: Sat Jun 02, 2007 9:44 am
Location: Hampshire, UK

Post by SixedUp » Sun Sep 09, 2007 12:12 pm

klankymen wrote:OK, I see what you mean about the desktop board, the model number has a D at the end: J7F4K1G5D vs J7F4K1G5.
Its not clear to me how much extra performance you get from the desktop version, but for my usage, the "normal" C7 is perfectly adequate.
One question, how are you managing the power wires? AFAICT the pico-psu has 2 molex connectors and a floppy, are you using some kind of splitters? or are you running one of the drives off the floppy wire?

I have a vague recollection there is such a thing as a 1xMolex-->2xsATA adapter, are you by chance using such a thing?
Yes. The picoPSU provides 2 x molex and 1 x floppy connector. The floppy connector is being used to power the PATA-SATA converter, and then I have 1 x Molex-single SATA power adaptor, and 1 x Molex-double SATA converter which are powering the three drives. The fans are all being driven off the headers on the motherboard by way of voltage-drop resistors. I'm still considering ordering the CrystalFontz display/keypad that I mentioned at the start of this thread, but if I do, I can set some links on it so it draws its power from its own USB port.

BTW, if you are seriously thinking of doing something similar to this, you need to know that the wide-input voltage versions of the PicoPSU have less ultimate capacity on the 12v line than the strict 12v input models. I only found this out when I started reading the specs, but this is really important if you are pushing it to the limit during drive spin-up. Its confusing, because its hard to find really high power 12v power bricks, whereas higher voltage high-power bricks do seem to be more easily available ... but actually won't help you

Post Reply