Servers are most often found in climate controlled data centers. This means they aren’t exactly built for creature comforts like quiet operation. Quite the contrary — many server chassis include fans which absolutely scream when the machine is under load. [Whiskykilo] needed to set up a 12 U rack in his basement for working from home. He knew the sound would get on anyone’s nerves, but especially on those of his wife.
To solve this problem, he built a sound isolated rack. The build started with a standard 12 U metal rack frame. This is wrapped in 1/2″ MDF coated with automotive sound deadening material. An outer frame built of 1×4 lumber and another layer of 1/2″ MDF. Isolating the inner and outer boxes made the biggest contribution to quieting down the noisy servers.
Computers need to breathe, so the front and back doors of the rack enclosure include banks of intake and exhaust fans to keep air flowing through the servers. Two AC Infinity controllers keep the fans operating and monitor temperature. These machines do generate some heat – so 64 °F (18C) intake and 81 °F (27C) exhaust is not unheard of. The servers don’t seem to mind running at these temperatures. A Raspberry Pi 3 keeps an eye on UPS operation and displays the data on a 7″ HDMI LCD.
Interested in running a server at home? You don’t have to go to the lumberyard – check out this server made with Ikea components, or this server built from 96 MacBook Pros.
47 thoughts on “Sound Isolated Server Rack”
Exhaust fans are pointed directly at the wall with a few centimetres gap. What I am missing here is before/after comparison of noise and components temperature. I do wonder what is the temperature resistance of materials used to build this? This looks like a receipt for a disaster.
Receipt for disaster? Wow, I really spent too much on this disaster. I hope it’s tax-deductible! :)
Recipe is more used in such a case (haha, case). Anyway, be sure to keep your receipt, of course.
Radiative cooling from metal contributes almost nothing to a typical computer unless it is specifically designed for it, i.e. the components of the computer are bonded to the case as in fanless or industrial PCs.
I remember making my own case for a computer in highschool. I used a standard ATX as a base, bonded wood to the outside, glued some black sound deaddening material to the inside surfaces and left the front and back exhausts intact. It was a 1 to 1 comparison which resulted in absolutely zero difference in temperatures. All the cooling comes from moving cool air in and warm air out.
Is your server chassis warm to the touch? If so that may make a difference. Also if so you already have a recipe for disaster.
I apologize for my harsh comment. Let me write here what my concerns/suggestions really are for this project. For materials used: what I am worried is how temperature/fire resistant those materials are. I know that risk of fire is not big, but if there is fire, are those materials OK to use in such electrical applications? I know that, for example, plastic cases around breaker box are from materials which are non-flammable, or at least they slow the burning.
For actual efficiency: as I wrote, I am missing noise measuring before/after the project. For temperatures: it is good to measure inlet/outlet temperatures, but what really matters are temperatures in the components. I would go so so far that I would write that all equipment in the rack has at least one readable temperature attribute. Which will definitely help with before/after comparison.
As it is written below, most of noise comes directly from the fans (there is, of course secondary source of the noise, vibrations). So in this project noise is coming from the front and from the back. I know how hellish that fans sound can be, trust me, I used to wok with servers full of GPUs and CPUs and tried to make it cool. I also made attempts to make my desktop PC quieter. The way is to silence fans (use ones with better bearing, slow them down, mount them in the way the vibrations are suppressed), or use different ways to cool the components down (submerge the HW in oil, use liquid cooling, other more exotic approaches). That’s for now. Enjoy rest of the year.
The autoiginition point of mdf is somewhere between 425 and 500 degrees F, 218 to 260 C. All the servers I’ve ever worked on will shut themselves off waaaaaay before this temperature. Also, mdf is kind of a pain in the ass to get started burning. The foam however, no clue. Though honestly, I’ve never seen a computer component actually catch fire. I’ve seen brief explosions, but no continuous fire.
You know they invented decimals.
“If anything F’s unit size is better. 180 degrees between freezing and boiling makes a lot more sense than 100 degrees.”
Better at what? Your statment doesn’t make any sense. The only valid argument is that everyone should be using same units. SI units have been accepted by the scientific world and also by most countries in the world.
Most people are neither performing scientific experiments nor communicating temperatures with people living in other countries. Those who do for the most part are already using metric even in the US.
The real problem is this. Can you imagine convincing your grandparents or even your non-scientifically minded peers to replace all their thermometers and only talk about the temperature using some units other than what they grew up with?
I know, it’s bizarre if you only associate with other geeks but most people don’t even want to think. At all! It is highly unpleasant to them. Extended thinking for a norm is kind of like grabbing an electric fence and challenging yourself to see how long you can hold on. How exactly are you going to sell them on re-learning what the temperature and size numbers they have used their whole lives mean?
this is a total valid point, to do so would be crazy, but you need to start to adopt, so a new dual scale thermometer would be nice, also writing out of the si units during tv shows, and for sure start to teach the young at school for si…
As someone who’s moved between countries and systems, this is a total non-issue and something you come to terms with right away. What is a comfortable room temperature? I don’t know type it into Google once and then set your thermostat to 72F here-on forward.
Give your grandparents more credit. The vast majority of non-technically minded people in the world have had to endure some incredible changes in their lives as they have in many cases changed miles to kilometers, adopted new currencies, and since you are talking about grandparents pretty much the entire english speaking non-geek grandparents in the world outside of the USA did the F to C conversion too.
Don’t use others as an excuse for not making a change, especially when that excuse is a belittlement of their intelligence.
Ah, you realize that the thermometer doesn’t change. Just add a metric scale to it. Here in Canada, it’s common that thermometers have both F and C scales. Same for speedometers. Etc.. Common usage is in what one knows, defaults to C, provides an easy means to learn. Analog devices simply have both scales.
In practice, electronic devices with a digital setpoint are usually switchable between C and F, and they tend to use ‘even’ values. So selecting F tends to provide more set points across its selectable range. As in:
F: 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, etc.
C: 18, 19, 20, 21, 22, 23, etc.
So I find myself setting devices to F, but think in C.
Metric is much easier to judge how close you are to freezing, room temp, hot, too hot, boiling.
425 and 500 degrees people who put a man on the moon, 218 to 260 people who have not.
Fahrenheit’s scale is literally based off of the dude’s own body temp–inaccurately measured–and the lowest temperature in Danzig in the winter of 1708. Then it was retconned to be the temperature of some absurd mixture of ice and brine that has no scientific application, thus it could be somewhat difficult to reproduce instead of actually impossible without a time machine.
It’s fully stupid. It’s stupid all the way through. Solidly and utterly stupid.
*yawn* it has higher resolution than C… it is an intuitive scale for our environment… and just as easy to re-create in a lab (shock of the f’ing century… ice water is 32 and boiling water is 212… give or take an impurity or elevation difference). Metric proponents love to repeat BS talking points that have been drilled into their heads by other metric proponents. I can work in either system, and have spent half my career overseas… I’ve seen a staggering amount of order-of-magnitude errors from qualified (metric trained) engineers working in metric. I’ve never seen such a mistake in traditional units exactly because that system does not rely on a decimal point that moves around all willy-nilly.
The decimal point makes the resolution irrelevant. Just creating accuracy to 0.5degC makes C more accurate.
There’s nothing intuitive about it. Intuitiveness comes from familiarity with a scale. I see zero sense in what anyone says about F.
It’s easy to recreate in a lab now, … it’s been redefined based on the same principles as C, just 180deg separation instead of 100. (Which makes C more intuitive :-P)
F proponents love to repeat BS talking points that have been drilled into their heads by other F proponents.
I also work in both systems, and order of magnitude errors are dwarfed by general unit conversion errors as orders of magnitude can be understood, but a system which you don’t understand by its very nature can’t.
If you’ve never seen a mistake in the F system, you’re lying or not paying attention.
But really it’s just a scale. There’s only one argument in benefit for a scale, and that is to minimise complex conversions. Metric does that naturally, abolishing F extends that since all it serves is to add one more complex conversion.
Lots of Americans have switched over to metric for much of their work on their own–myself included–and even working inside of a country with stupid measurements it saves time, money, and effort. Everywhere else in the world successfully changed over, but for some reason it’s unreasonable in the USA. Why is it that we’re so special?
Two types of countries – metric users and those that went to the moon.
The irony of this is that NASA uses the metric system.
The extra deliciousness of this is that at least one epic failure (Mars Climate Orbiter) was due to NASA incorrectly converting between the metric system and the USA special system.
The old Gigabyte K8NSNXP (A fairly high-end board in the early 2000’s) had dual power system daughter card that was both not required for normal operation and infamous for catching fire. I had a friend nearly burn his house down with it after I waned him not to install it and he disregarded my warning, This occurred not once, but twice. (He apparently didn’t learn the first time and wanted to hear me say I told you so, a second time.) Still, this kind of thing is uncommon.
As far as foam goes, a quick google returns an issue of Polymer International from June of 1976 that cites a study about the oxidation and ignition of flexible polyurethane foams which concludes a temperature of 190°C (external) or 250°C (Internal) would be required to induce combustion in newer foams with aged foams being notably more stable.
Don’t throw water on burning polyurethane foam. It’s worse than using water on a grease fire.
Some buddies and I had a MUD running in high school on an old Pentium 2-era PC. We had to stop playing when it literally caught fire one day.
Modern PCs probably do shut off before then. Now I’m recalling the old classic video of removing a heatsink from a running Pentium 4 vs an AMD machine while the Quake 3 timedemo was running.
Nice job, but no indication of how quiet new set up is c.f. the old?
BTW, cooking your ups battery like that will really shorten its life.
A bit of carpet padding on the wall, behind the cabinet, will help a bit with that hard reflection from the gloss painted (block)wall.
Be sure to put the bare side of the padding out, not the coated side.
I’d also say to put at least 10″ space between the wall and the exhaust of the box.
Less restrictive air flow is generally less noisy, along with better for the hardware temps.
Tighten the gaps around the gauges, etc, will help reduce noise a little too.
Why is there just about no article here where someone doesn’t say the equivalent of “you’ll shoot your eye out with that”, even when it’s obvious that the commenter knows less and is less skilled than the hacker presented in the article? Sure is tiresome. Even if a thing has some moderate risk (unlike the current thing) – hey, getting out of bed is risky, and adulting – you should try it – means learning how to manage risks, even becoming competent at what you’re doing to effectively reduce the probability of trouble. Gheesh. Since everything gives you cancer or burns your house down, and being a race car driver or skydiver kills you even quicker, you might as well just vegetate? Could it be more than luck that some live to tell the tale?
It’s the forum effect. People who can’t do, comment. Yes, I fully realize the hypocrisy of what I’m saying.
I would prefer you get to keep your fingers, eyes, blood, and that I don’t have to pick them up to rush along with you to emergency… if you survived.
These comments ARE ubiquitous, boring, irritating, but priority enough that signs are made because we bore of having to stand next to everyone at a hackerspace to continually remind them.
Honestly what we need is an ultrasonic sensor that detects “operator present” and plays the warnings audibly. Nobody reads signs.
Keep up the boring warnings. The peer pressure is a worthwhile example to those that “just don’t get it till they get hit”.
Warning: if peer pressure exceeds 200 kilopascals, evacuate the work area and vent atmosphere into space.
Seat-belt warnings coming to a hacker-space near you.
Yes, my apologies, unpopular and boring. Much less exciting than watching someone adjusting their oxy/acetylene torch then turn to check the O2 pressure but now holding flame of torch on acetylene hose. Much less exciting than walking into a xray suite to see my newest tech literally crawling inside the bottom of a cabinet with metal tubed vacuum to clean out the dust around 480v three phase mains input transformer while the system is idling and the resulting excitement when I pushed emergency stop. Even a grade school incident, “Brains” decided to reproduce Ben Franklin’s kite in thunderstorm experiment… … and Dad demonstrating incorrect use of a fire extinguisher in garage when welding on a lawnmower without bothering to remove fuel tank.
Unpopular and boring till AFTER an incident like the fire at a hackerspace this summer. Then everyone wants to talk, even gets it’s own article! That sounds great, but safety is BEFORE, not AFTER.
Do blow this post away if you wish, if feels uncomfortable for even me but safety is very much a before thing, not after, so those saying safety comments are boring and yadda yadda are right, wanna just talk after? And here we are… before.
Not after or using a pedestal. Just actual safety, and it’s a pursuit not a happy coincidence.
Biomed, I think it is more an exhaustion from all the panic over soldering without a ventilator, or ludicrous concerns over handling elemental mercury. Or getting silicosis from playing in glass beads. It is the misinformed opinions on the dangers of various gases or chemicals. Or crazy over-estimations on the momentum of projectiles. Or being upset that some adult who is building a home made flying machine is willing to risk life and limb by testing without an ejection seat. It all seems very unhackerish.
Then there are the very knowledgeable contributors who point out unfamiliar dangers or inform in other ways, which is much appreciated. And features on lab and shop safety are great, especially of they cover cool safety hacks and new safety equipment. For example, the discussions from the saw-stop contraption.
I was going to make some more cracks out getting up in the morning, taking off your sleeping helmet and putting on your shower helmet, then the eating helmet with hot coffee shield, etc. etc. But could not make it clever enough to contribute to the gestalt.
Given than the noise is 95% fan generated (a 5% slack for the hard disks), how enclosing those fans and then adding fans is solving this? It would be a lot more engineer oriented to attack the origin of the sound. Use low speed fans, use big fans. Fanless? Not unheard of. Oil bath, neither. Or go full bananas and make it a sound isolating box and a couple (lot) of peltiers to move the heat. Also, 18C inbound, 27C outbound? Are you in Santa’s Home? Here on a day without AC on it would be like 35C inbound/60C outbound.
Not the quality/creativity we use to see here.
His company probably wouldn’t be happy if he tore their 200k USD server apart to install custom cooling. Support and warranty goes down the drain if you do any modification on such device. Also they don’t have more silent version to offer. So only option is to make external casing with improved airflow and noise reduction.
Most people don’t get the fan noise part. Nor that noise travels through the air mass, even when the air mass is being moved along by fans.
Nor decouple, decouple, decouple…
He even left the grills on the fans, so the air movement is making more noise as it leaves the box.
He’s done better than most by having an inner and outer box, but then didn’t fully decouple them… and missed on decoupling the ‘blue’ from the ‘green’. An air gap between the ‘green’ and the ‘grey’ does wonders for buffering noise.
The output could be channelled internal to the box lined with absorbing media, so there’s no line-of-sight air mass and more length/time for noise absorption.
Dito for an absorption/damping media on the wall behind it to minimize reflecting the sound.
With the noise of fans out the back reduced adequately, he may notice that adding damping media on the outside of the box stops its vibrations from being transmitting into the air from the box surface.
It’s in the basement. Get that box up off the floor. You never know when your basement is going to flood.
Naturally there are already quiet racks. Usually found more in music studios, or equivalent. I didn’t want to spend the premium, but they’re out there.
Angled cardboard, finger foam, a little aluminum and wood can work wonders.
Cool build… if you need something like this in a commercial setting (i.e. not your dime), you can buy something similar ready-made (Search for “Quiet Rack”… Startech has as product that looks like a less hefty version of this build, but it costs north of USD$1000).
If you only need a single server, then look into a “small office / remote office” style server (e.g. the Dell T630, its motherboard design is nearly identical to the board used in their 2U and 1U servers, but it uses massive CPU heat sinks and large quiet fans).
What am I missing here. Servers are noisy because they have a lot of fans to keep them cool. If you put the server in a bigger box, and put enough fans on that box to keep up the airflow the server is designed to have, you are going to have just as much noise, and you your cooling went from being done with server quality hardware with redundancies, to consumer quality hardware. I guess I am not seeing the gain here.
depending on the servers and the air coming in the fans shouldnt rev any higher, some servers are overbuilt for cooling anyways, I can at least see this reducing the higher pitch noises that high air flow fans produce which the lower pitch sounds are easier to deal with
Servers are noisy because they use (in a 1U server) 40mm fans that spin at thousands of rpm. Swap the fans for low speed units or increase the size of them and you can drop the noise levels massively.
I replaced the stock fan in a dell T410 (dual X5675 CPU’s under a medium load) with a same diameter 5v powered unit and the server lives in my living room. The internal temperature sensor reads only a degree or two warmer than than temperature of the room (about 22 Deg C). Obviously, the servers monitoring complains about the rpm being lower than 750 but that is not an issue (I push the temperature and load average to the LCD which masks the error about the fan).
Chill the air in the basement so the servers are always drawing in cool air. Then they shouldn’t ramp up the internal cooling fans to warp screech speed.
One place I worked had a single east facing window in the server room, about 3×4 feet. They’d painted over the glass with white paint. Despite AC running 24/7 it would still get nice and toasty in the room in the mornings. I was walking behind the racks one morning and went past the window. Yowza! Hot hot hot. I got a sheet of 3/8″ urethane foam insulation, foil faced on both sides and cut pieces to fit into the inside of the removable wooden window screen frame on the outside of the window. Glued them in with silicone and after reinstalling the insulated screen the morning temps in the server room were several degrees lower.
Has anyone attempted to 3d print or manufacture some form of muffler for the fans themselves?
I’m actually looking to do about the same. For my part, fans are not what makes the most noise. 15k SAS drives are always spinning (meant to be) and makes a high frequency vibrating noise. Yea, you could replace those with SSDs, but used SAS 300gb drives are like 10$, 10-15 time less than consumer SSDs, 1Tb 30C) then it’s unbearable !
So my approach will be passive filtered intake at the front, with angled maze to remove noise. Active fan at the back with a rheostat to control speed (5-50CFM) and temperature controller like STC-1000 to start/stop fan. For noise blocking, I have some used under floating floor mat, it is made of fabric with a plastic side (that should work to prevent particles of it to get in my servers).
The idea of a dual containment is good, will reproduce it and put noise absorbing insulation (Rockwool) in between.
The exhaust port will be a 3″ flexible plastic tube (like laundry dryer) that will open up at top front. The air mass in the room is in the center and back exhaust will not be efficient if the rack is against a wall.
The alternative would be to buy a half rack (250-700$) and add insulation to the room. It would cost a lot more but look more professional. A full rack will not fit as the insulation needed on the ceiling would lower it too low to fit a 80″ full rack.
Also, the link to the build is dead.
My post was abridge, here is what’s missing at the end of first paragraph :
1Tb 30C) then it’s unbearable !
2nd try :
Put 16-32 of them on servers and you get a LOT of noise, even worse when you read/write in the raid arrays !
The servers them self are not that loud, if room temp is less than 20C, it’s just a bit more than a loud PC. When the fans kicks in at full speed (room temp more than 30C) then it’s unbearable !
Please be kind and respectful to help make the comments section excellent. (Comment Policy)