Running a Server at Freezing Temperatures?

mw13068 asks: "As a part of a backup solution, I'm thinking of running a backup server in my unheated, unattached garage. I live in central New York State, and the temperatures very often drop below zero degrees Celsius. The computer is a Pentium III Celeron running at 733MHz. Has anyone else tried this sort of thing? If you have, please share your experiences."

Mice cause cancer in computers

[pease1]

Make sure your case is hardened. Every little critter, including mice, will want to live in the warm case. We had a computer in an astronomical observatory dome and mice built their nest on the CPU. The acid in urine from the mice destroyed the motherboard.

Check the specs

[lrdviperscorpian]

Look up the specs on all the hardware. Most have an operating temps guideline. If your within it you should be alright.

Been there, done that.

[Oinos]

I ran a couple of dual PIII 450's in my garage in Minnesota last winter with no problem. I didn't have any room in my small apartment for them so I put them in the garage and used a couple of Linksys WAP11's in bridge mode to get them talking to my cable modem in my apartment. The average temp in the garage was about 5 degrees above zero last winter.

The one thing you need to watch out for though is static. When it gets cold and dry, you don't want to be ripping open your machines in the garage. My machines stayed up from October through last June without any problems.

I wouldnt trust it...

[Bilzmoude]

Unless you know it is secure from rodents and bugs, and you are positive about the operating temperatures of your hardware, I wouldnt trust it.

Also, you will need to check the operating temperatures of your motherboard, processor, hard drive, memory, and any other components, and make sure that they will work in the temperature range you expect. Operating temperatures are much more narrow than storage temperatures. The operating temperature, for example, of the Maxtor DiamondMax 10 is 0 to 60 degrees celcius. I am guessing your temperatures will get well below 0 degrees, so you will need to be certian before attempting it.

Also, what type of humidity do you get in this garage? Are you worried about data security? Can someone just walk into your garage, and fiddle with your server? Can the kids basketball hit it? All things to think about.

constant temperature?

[Naikrovek]

I wouldn't worry too much about it being too cold. if you have a pusher fan, take that out. puller fans (that exhaust air, instead of pulling it in) will have the temp of the air inside the case, rather than the temp of the outside air. lubricants become more viscous with colder temps, so you want you fan to breathe the warmer air from inside the case.

you probably want to make it a smaller fan also, you don't want too much cold air going through. cold is good for CPUs but too much cold breaks solder joints.

if you can control your fan thermostatically i would recommend that. having computer parts get hot, then cold, then hot, then cold, then hot, then cold, due to day/night cycles KILLS solder joints quick. condensation is also a concern with widely varying temperatures. condensation is bad, of course.

as someone else said, rodent-proof the case and check it for infestation often. mice will chew right through sheet metal when they need to. Try mounting it on a wall somehow so rodents can't get to it.

i'm not worried about the below zero C temps, i'm worried about temperature fluctuation. using a smaller than OEM fan will keep what warm air there is inside the case there a little longer, and should keep the insides of the case above 0C constantly.

Re:I would be concerned about humidity

[alienw]

At lower temperatures electrical wires have less resistance and it could do some damage (theoretically of course) to some electronic components.

Bullshit. Wire resistance in an electronic component should be negligible. The resistance change caused by temperature is just about impossible to detect without very sensitive instruments.

In general, electronics do not care about temperature much. Most chips, for instance, are rated from -40 to +70 degrees C. It's the mechanical stuff (hard drives and, to a lesser extent, fans) that you have to worry about. The only electrical problems that could occur would be related to condensation.

Dust isn't that big of a problem

[rhpot1991]

I have had some overheating problems with my athlon xp 1900+ in the past, so I started to keep my windows in my bedroom open 24x7 durring the winter, this solved my overheating problems and the computer seemed to run better as I was sitting in front of it seeing my breath. As far as the dust goes, I used to work in an IT Dept. for a factory that made security doors for mall shops, They had some old computers through out the factory that were used to operate some of the machines. I did maintainance on a few of them and when they were opened there was literally a layer of thick dust covering everything inside, this didn't effect any internal parts, the only thing we ever had to replace on these pc's were floppy drives. I am talking pentium 1 generation boxes here, so I would venture that yours should be pretty safe since your garage should not produce near the amount of dust that this factory produced.

Re:CPU probably irrelevant

[sakusha]

It's not an issue of hard drives melting, it's an issue of thermal expansion of the platters. Hard drive platters go through a normal amount of expansion because solids expand when heated and contract when cooled. Drive controllers are designed to recalibrate occasionally to check for expansion, to insure the heads are positioned correctly, off-track positioning leads to errors. But I seriously doubt the calibration would work outside the range of temps designed into the controller.
Another issue is lubrication viscosity. Lubricants become more viscous at low temps, if it got really cold, the lubricants in the drive spindle could actually become solid, freezing the bearings and burning out the motor.

How to be absolutely sure it'll be okay

[Engineer-Poet]

Get a case with a thermostatically-controlled main fan (not CPU fan, main fan). Put this in a 5-sided wooden box (hardened against critters, screened on the bottom) and insulate it with construction foam (inside) on four sides and the top. Half-inch foam will probably do. Vent the system fan out the bottom.

What this will do is create a "bubble" of warm air inside the box that is vented when the fan is running and stable when it is off. This will keep your box temperature roughly even. If you are concerned about cold-starting hard disks after a period of off-time, make sure you have a power supply which remains off after a power loss and add a 100 W light bulb inside the box. When you want to power the system back on, switch the bulb on and leave it for an hour or two before you hit the power button, then turn the bulb off again. Do not bring cold hardware into a warm, humid house to warm up - you will get condensation.

As long as you have the bottom of the box screened against critters and otherwise isolated, you probably won't have to worry about static or other environmental nastiness.

Re:I would be concerned about humidity

[Anonymous Coward]

Fans circulate air through the case. Desicant packs aren't going to cut unless you have enough to dry out the whole garage (or possibly the world).

Make a few BIOS settings

[QuantumRiff]

Tell your PC to never turn off hard disks, never turn off fans. (might freeze if they stop, and not start again). Take the floppy out of the machine, and replace the hole in the front with a blank panel. It might be a good idea to do that with the CD/DVD drives as well. Make sure that the back of the case is all sealed up, (ie, no open holes for old PCI devices you no longer have). Lastly, Don't put anything over or close to it. Your going to need it to be able to suck in air, and evacuate the air with the fans. you do not want to be recycling the air (like you would if it was under a blanket) as it can increase the moisture of the air.

Re:Garages

[ivan256]

My stepdad has a garage and I maintain his systems (or at least talk him through it on the phone if I can get away with not going there it). The average lifespan of a machine there is about a year. He used to use DEC VT-100s. Those things lasted 10+ years easily (except the keyboards), but in a PC, he needs new fans every six months or so and a new hard drive every year or so. FOr his current batch I've got him using rack mount equipment since it has built-in air filtration, but he hasn't been using it long enough for me to tell you if that's helping.

We keep his server in a dehumidified space in a rack with doors and air filters over all the openings. That machine seems to be OK...

Running is easy, starting is hard. Think about:

[Myself]

Oil viscosity is the most important factor here. The machine will be fine in steady-state operation, but if it's turned off for more than a few minutes, a cold start will be very difficult.

Cold-starts outdoors will require use of a heater. Blowing a hair-dryer (on low heat!) into the case for a few minutes prior to startup should warm the drives enough to spin freely, but consider this: During the warming period, the hard drive platters are stationary, and may heat asymmetrically. This means their thermal expansion will be uneven, throwing the spindle off balance and making it nearly impossible for the heads to track a cylinder. Depending on how the drive case is built and how the heat is applied, this may not be a concern. (Heat for 5 minutes, let sit for 1 minute, then power on?)

Of course, bringing it back indoors for startup would be an even worse idea, as moisture would condense on the cold metal. Whenever you bring hardware in from the cold, put it in a tightly closed plastic bag first, and leave it in the bag until it comes completely up to temperature. The relative humidity inside the bag will drop as it warms up, avoiding condensation concerns.

Fans are a bigger concern, as they don't generate much of their own heat like drives do. While it's likely that you won't need much cooling, a CPU fan is almost guaranteed to still be necessary. Look into tip-magnetic-driven (TMD) fans, whose design gives them more torque to overcome stiction at startup. Find one with ball bearings and replace the lube with a light machine oil.

I don't think dust is such a big concern, if the case provides air filtration, as any server case should. Just get the thing off the floor, out of dust-bunny territory. If the power supply fan is thermostatically controlled, airflow should be kept to a minimum and dust entry will be negligible. It still never hurts to pop the case off every few weeks and check. The poster might even have an air compressor in the garage! :)

Optical drives might be tricky, as they don't spin constantly. Luckily their motors are amazingly torquey and should have no problem spinning up even with cold bearings. If you can position the hard drive directly below the optical drive for heat sharing, so much the better.

These suggestions should keep you running to below freezing. If you get much below that, electrical characteristics of components start changing significantly, and you might have all sorts of weird problems. Look at the temperature-versus-value curves of various passives, and you'll see what I mean. Even clock crystals resonate faster because they've physically shrunk.

This hasn't addressed the first pressing question: Why? Except for acoustic noise, I can think of no reason to put a machine outside during winter. Consider that every watt of electricity you use gets turned straight into heat. Putting your electric heater outside simply means the energy gets wasted, rather than heating your house and lightening the load on your furnace. If you're paying for the energy anyway, why not keep it inside where it does some good?

Re:Sealed Case

[NarcolepticTerrorPoo]

Much like a car's water cooling system you want the air to circulate in a closed loop until it's reaches a certain temp and then you want to open the loop and dump the heat.

You also want the circulation within the case to be good so that you don't end up with hot or cold spots (cooked/frozen).

You'd also want to use a heat exchanger to preheat the incoming air to above freezing so that you don't get instant condensation from the inrush of sub zero moist air and make sure there is a fail safe in case the control system dies.

It's the condensation...

[go$$amer]

Rapid temperature change is what you need to watch for - I used to run 486-PII machines in unheated buildings in Minnesota all the time, ambient temps over those winters and in my area (central) got as low as -40 (that's Fahrenheit and Celsius - the scales cross there...) no troubles that I can recall.

and I only had to worry about dust from the shop - BTW, under no circumstances put your box near anything that grinds metal! That's a real quick kill.

Re:try junkyards

[TykeClone]

Garage north of the mason-dixon during the winter usually have temperatures much, much less than 50F. In northern Iowa, the temperature inside the garage can get down to 0F (don't even think about how cold it is outside!)

Re:CPU probably irrelevant

[itwerx]

...it's an issue of thermal expansion of the platters.

Also a consideration in tape drive head alignment.

Re:Not a good idea.

[Anonymous Coward]

Hot CPU + cold air = condensation. Water inside a PC=bad.

Wrong.

Warm air can hold more moisture than cold air. In really cold air, there is almost no humidity as it would have condensed already. The warm surface is at a higher temp than the air, so if it would condense at the temp of the CPU it would have already condensed out of the air.

Warm air + cold surface = condensation.
Cold air + warm surface = nothing.

Re:CPU probably irrelevant

[wayne606]

I wouldn't suggest shutting down the CPU fan no matter what. It can be very cold a few inches from the CPU and the heat sink too hot to touch... Without a fan your system will turn itself off within a minute or two (if you are lucky)

HDD Operating specs

[Anonymous Coward]

Typical operating temperatures for hard disks is 5C to 55C

And they need to acclimate to the environment before you start them up..

Re:Mice cause cancer in computers

[dgsoftnz]

Just terrible. I have that happen all the time. I now know to make sure that all of the card slots in the back have either a card or a blanking plate, otherwise the mice realy do make a mess

Re:CPU probably irrelevant

[sakusha]

Oh yeah, I forgot, there's one other component that will fail under severely low temps: barrel capacitors. They're generally filled with a semi-liquid paste that can freeze at low temps, unless you've got mil-spec computers like the guy who described his aircraft maintenance computers that are rated for operating temps down to -70F. Look at some of the overclocker websites with experimental liquid nitrogen cooling, they take great pains to cool only the CPU chip, if they cool the whole motherboard, the capacitors freeze and fail.

Been there, done that

[holviala]

I have an old IBM Aptiva P2/350 in a closet that's open to the outside and has very little insulation and a leaking door. It's been there for the last three years and it's still working fine 24/7 - the only times it's been down have been when the power has gone down.

So, what's so special about this one?

I live in Finland. It's cold here. It has survived weeks of -30C with NO problems. The only things that have borked are a CPU fan (which had no bearings to begin with) and a brand new Seagate hard drive which lasted for exactly two weeks - the replacement has worked fine for a year and a half now.

I've been thinking about replacing the machine with a nice 4U industrial PC that I have laying around - it's just that the Aptiva has proven to work in extreme conditions so I'm not sure if I wanna replace it. Ever.

An electronic engineer responds

[Anonymous Coward]

"At lower temperatures electrical wires have less resistance and it could do some damage (theoretically of course) to some electronic components."

Highly unlikely, unless winter temperatures approach absolute zero, since normally the resistance of copper changes by parts per million/degree (teperature stability is one reason copper is widely used, even though there are better conductors). Even if the copper became superconductive there'd still be no damage, because the current flowing through any given PCB trace is determined by the circuit configuration, not the resistance of the trace itself (note that in the circuit design phase the resistance of the circuit board is usually taken to be zero. RF emission and trace impedance* are generally only considered during the PCB design phase).

There's actually a greater chance that the semiconductors would stop working because of the extra energy required to force electrons across the semiconductor junction. Or the electrolyte in certain types of capacitors might freeze (bet you wish they still used low-temp polychlorinated biphenols). Or the hard drive spindle could sieze, or the fans die (not that they'll be needed). And, of course, the backup battery on the mobo will simply stop working at some point, just like your car battery does in winter. Basically, there are plenty of things far more temperature sensitive in a computer than a few strands of copper.

"It's an old machine anyways but try to check first if there isn't some temperature that it could reach that could be too low."

Most consumer grade parts are rated to handle down to about 5 degrees centigrade (I don't recall what that is in Archaic American Numbers; I think its something like 6 1/2 furlongs to the hogshead); any manufacturer relying on inherent internal heating to push that lower is fudging the figures in a dangerous way (think PMPO ratings...)

*Resistance is only dominant with DC. For AC, or pulsed DC systems such as computers, inductive or capacitive components can be far more important (too high capacitance can cause your nice square clock signal to turn into a flattened triangle; excellent for generating timing errors), which is why we refer to the sum of inductive, capacitive AND resistive elements, and call it "impedance" (if it helps, you can think of impedance as "frequency dependant virtual resistance", and it too is measured in ohms).

A capacitors' impedance (virtual resistance) drops as frequency increases; so PCBs designed for high frequencies must have as low capacitance as possible between tracks (the simplest capacitor being two pieces of wire seperated by air, the value determined by the surface area of the conductors, their distance apart, and the dilectric constant of air). Hence, short PCB traces with better-than-air insulators (multilayer fibreglass) are needed to push bus speeds to their current point. This explains why its possible to overclock a monolithic CPU more than the motherboard its connected to: timing errors caused by parasitic capacitance.

Ask an engineer a question and you'll get the answer you wanted, plus a few dozen you didn't.