Ventilating my PC in a cupboard

[Al Reynolds]

Morning,

I want to put my PC in a cupboard to reduce the sound impact of the fans. Before anyone suggests that I get quiter fans, please bear in mind that this is a very small form factor PC and I haven't got the time to go finding exact match parts that will fit in its tiny case!

Anyway, it piles out a lot of hot air, and currently has ventilation on both sides of the case. I'm conscious that if I put it in a cupboard it may overheat quite quickly, so I'm looking for options to ventilate the cupboard.

I did consider venting to the outside wall, but it's upstairs and the wall is rendered, and I'd rather not disturb the render unless absolutely necessary. I have a number of other options:

(i) Vent to and from the room I could just put in grills with quiet fans on the face of the cupboard somewhere - one to suck air in and one to push it out. My worries are that: (a) it will look bad (b) it will sound worse than before (c) it'll just push warm air around the case

(ii) Vent into the chimney breast The cupboard will be next to the chimney breast, so I could potentially put an extractor to take the hot air out into the chimney. Both fireplaces are closed off but the chimney is ventilated by grilles on the chimney breast. The air could just be sucked into the cupbard through gaps round the door or I could fit a grille.

(iii) Vent into the loft The same as above but vent into the loft space. I could conceivably end up venting to the outside via the eaves.

(iv) Vent to/from the loft/chimney breast It occured to me that I could suck cold air in from the loft, and vent the hot air out to the chimney breast, or vice versa.

What do people think?

Al Reynolds

[Ian Stirling]

I'll just mention what I did. I started off with a PC in a knee-hole desks cupboard. This was then lined with carpet, and two foam-mounted fans placed on the back.

This was a good deal quieter.

Then I made an interior cupboard spaced some 1" in, also carpet-lined. The fans were replaced with ones inside the inner cupboard that connected to the outside of the inner cupboard (blowing in and sucking out) through tunnels of carpet, and ducting (again of carpet) to the original vent holes (which were in different places.

Alternatively, have you considered simply putting the PC in the attic?

[mike. buckley]

In message , Al Reynolds writes

First off - what temps is the PC running now? You're probably pretty safe with a CPU temp up to 50 degrees c, so stick it in the cupboard for a bit and see how it goes. Not ideal but might save you some work. I know people who have multiple servers/switches/routers inside an airing cupboard and don't have any problems with overheating....

[Ian Stirling]

Don't forget that you're not interested in what temperature it's at now, but the temperature when it gets to 30C or so inside. (or whatever the max temp is.) Hard drives get lots less reliable as the temps climb.

Many hard drives have thermometers inside to measure the temp, and can report via SMART. Look for disk status monitoring tools.

[Dorothy Bradbury]

You say it is a very small form factor, but exactly what? :-)

o Micro-ATX - CPU can be P3-Cel (30W) to P4-Prescott (115W) o Mini-ITX - CPU can be VIA-C3 (12W) to P4-Prescott (115W)

You say it piles out a lot of hot air, which makes me think >> VIA C3.

Duct/Extractor to Outside = Static-Resistance + Continuous-Rating + Noise-Level o PC-style 12V Axial fan - eg, 120x38, 171x51mm

---- +ve -- Continuous-Rated + Quietish + Cheap

---- -ve -- Low Pressure even from 38-51mm depth fans = Low Airflow

------- you can get away with a 172x51mm eg, Papst 6212NM, but it's not quiet

------- so you are then into noise-proofing the cupboard &/or fan o Extractor Axial fan - eg, shower, toilet etc

---- +ve -- Easy Build-In off the shelf fitment + Acoustic enclosures for loft use

---- -ve -- NOT always Continuously-Rated (some Vent-Axia are) + Not quiet + Cost

Since this is just a single PC, I don't think this solution is worth the hassle. With 16x 1U dual-processor PCs doing CFD it is a different matter re >4kW of heat.

Noise Reduction: o Distance -- dB(A) drops off rapidly with distance

---- use K-V-M cable extensions to relocate the SFF-PC further away

---- the catch-22 is your removeable media - altho external USB enclosures exist o Noise Source -- fans are one source (quiet fans exist), HDs another source

---- targetting the noise source is more effective than treating the effect o Noise Effect -- an enclosure (cupboard) can be useful for reducing noise

---- soundproofing must use Mass (low frequency) & Absorption (high frequency)

-------- TreadAir Carpet Underlay - rubber (Mass) rippled foam (Closed-Cell)

-------- carpet may include nylon - running the risk of static

---- do not fit ventillation fans to the cupboard (re PC / soundproofing / fan / ear)

-------- soundproofing should between the source & your ear :-)

---- every 90-degree turn reduces noise by 3dB(A)

-------- like distance, acoustic baffles work well - especially if lined

Cupboard cooling: o Natural convection

---- wood-slatted-door (prettier than vents)

---- at the rear of the slats, close of all except for the top & bottom o Forced cooling

---- PC Axial fans will be ok here

---- Do not fit to cupboard front (line intake path to absorb fan noise)

---- Do not fit in narrow-end of an enclosure (acoustically this is ugly)

Baseline before/after the design: o Identify Thermal Baseline Before -- CPU & Board Temp + HD Temp

---- onboard monitoring tools (MBM) + S.M.A.R.T. HD temp checker (HDDTemp) o Identify Thermal Baseline After -- objective is HD

[Andrew Mawson]

Al,

I have a small form factor PC (Compaq EN SFF) running 24/7 doing various control functions around the house. It sits in a normal 'wall cupboard' of the tatty melamine variety, and I have installed 9"x3" air vents in the bottom and top of the door. This had proved adequate for several years. Quite a bit of sound deadening was achieved by placing it on a carpet offcut.

Andrew Mawson, Bromley, Kent, UK

[Clive Summerfield]

For 1 PC I'd line the cupboard with underlay rather than carpet (less risk of dust, fine fibres, static, etc), then egg trays. Then a couple of baffled air vents. Should do the trick with no additional fans.

Cheers Clive

[Dave Liquorice]

This 1GHz Athlon is normally at just below 60C. Wanders of to check, hum, must need a good dusting again it's up at 70C... This machine runs 24/7 with no trouble.

As to the OPs query, suitable quiet fans are normally fairly easy to find. Just look at CPC, RS, Farnell, Rapid etc. Might be worth paying the few quid extra for the temperature sensitive ones. Even knocking 5 or 10% of a fans speed will make a great difference to the noise level.

One PC in a what sounds like a fairly large cupboard, I think I'd try to make sure that it drew it's air from the relative cool of the room and just vented into the cupboard. Why dump this (expensive as it's sourced from electricity) heat outside? It'll help keep your house warm.

[Pete C]

Hi,

Another option could be to fit a quiet CPU fan and have quiet external ATX power supply in the cupboard. The cube would then be fed from an extended ATX power lead.

cheers, Pete.

[luggsie]

A 2 GHz Pentium probably uses 100w power absolute max excluding monitor. Put a 100 w lamp in the cupboard for a day and check the temp rise. If its less than 10 deg C there is unlikely to be a problem in the UK as the ambient temp never gets too high.

john

[Dorothy Bradbury]

Keep the ATX extension lead to ~30cm re remote sensing (but not on all pins), you can go a bit longer if the original PSU lead is a short one - but not infinite.

Frankly keep the solution simple: o Baseline where you are now o Change 1 variable at a time o Do not thro money at the solution :-) o Then establish how the new system compares & adjust accordingly

Treating cause is more effective that effect, although you will find that treadair rubber underlay (and it's clones) is very effective. The acoustic sound-barrier-mat, 5kg/m^2 & acoustic foam cost far more.

Typical coloco facilities run CPUs far hotter than most people realise, but that does not mean you can neglect HD temps - S.M.A.R.T. utils.

[invalid]

In message , Al Reynolds writes

These people are a good port of call for all manner of cooling and noise suppression.

[mike. buckley]

Blimey, that's running hot. I've also got an Athlon (2800, actually

2ghz) and it's below 50C, this is without any extra cooling apart from a more expensive (thus quieter and cooler) PSU. If memory serves my mates CPU cooler died (also a 2800) and the PC was ok til about 85C then it hung.

[Al Reynolds]

Thanks for all the detailed advice people.

For info, it's an MSI Mega PC version 1. The processor is 2.8 Ghz. Part of the problem is that I'm using an nVidia FX5600 Ultra graphics card which also has a fan on it. I don't really want to replace the card. The other thing is that I use a remote with the PC and the IR sensor is on the front of the case, so I'm looking at a glass door on the cupboard.

The discussion reminded me how loud my PS2 was before I put it in a glass-door-fronted cabinet with the video and DVD. Now I can hardly hear it with the door shut. I think I'll build the cupboard with the option to add extra forced ventilation later, and see how things go. I have a roll of unused underlay in the loft, so that'll go in there as well.

Thanks again, Al

[anthony james]

I'm trying to do something similar at the moment. I've a 3.0 P4 in a Shuttle XPC case in my kitchen that i want to keep in one of the base units. My day 1 test involved running the PC at about 70% for half hour or so (ripping MP3's, decoding digital TV and playing mp3s). The temperatures shot up very quickly from around 50C to 65C but what concerned me was that they didnt drop afterwards - a couple of hours close to idle but circulating the same air and it wasnt cooling down significantly. My guess is that you WILL need to ventilate the cuboard. I've not worked out how to solve my own problems yet but will watch this thread.

Re the glass front/IR issue - An IR repeater might be neater than a glass front. Check Keene electronics.

[Al Reynolds]

Good idea! Al

[Pete C]

Really, how does that work? Are some of the duplicated ATX connections used for this or are they commoned together?

cheers, Pete.

[Dorothy Bradbury]

Typical desktop PC: o Remote sensing is on the 3.3V rail

---- since 3.3V is quite a low voltage + power cables have length & voltage drop

---- and spec is +/- 5% which isn't a lot in voltage terms o Implementation for desktop PCs is via pin 11

---- the power pins are commoned off the PSU PCB to distribute load per cable/pin

---- a few (Tyan dual CPU?) boards used to melt their ATX skts under high load o Remote sensing is also required at the SATA 3.3V connector too IIRC

---- even the 2.5" SCSI 10k-rpm Savio is still 5V (I think)

Industrial PCs: o Remote sensing can be on all of 3.3V, 5V & 12V rails

---- these still average across the ATX connector o Remote sensing can get more sophisticated on mini-PCI multi-U arrays

---- these combine outputs from many std ATX PSUs re o/all current draw

---- they manage redundancy without expensive/unobtainable redundant PSUs

will have details for both ATX 2.1/2.2 & BTX.

Basically the spec just requires 3.3V remote-sensing. I've never seen a cable length limit - but have seen 1.5m 1U PSU cables (oddly!). If you take a standard ATX PSU, which has quite a long lead, some E-ATX super server cases required a 30cm extension to be added to reach the power socket.

Never tried an 'overly-long' cable so can't comment on what happens, I suspect the voltages would need careful monitoring during burn-in load testing. Would help if the leads (wire) were a decent thickness - some are, and some are a bit skimpy.

[Pete C]

I see. If i was to extend an ATX power cable I'd use some chunky speaker cable for the high power connections, looks nice and a few mm2+ is easily available. The lower power and signal connections would be handled with standard 4 way flat speaker cable.

All the plugs would be moved down to the far end, using some heatshrink which would look neat. Might need multiple cores or very thick cable for the high power connections, I wonder what sort of cable a standard ATX supply uses...

May be worth measuring leakage currents to see if they are excessive when combined. Also test the wiring or equipment with a Megger or something similar.

If it's intermittent earth leakage, connecting different equipment or circuits into an isolation transformer or maybe even a UPS would help narrow down the cause.

Failing heater elements, damp and even spideys can be a cause of tripped RCDs.

cheers, Pete.

[Dorothy Bradbury]

Do this to an ATX extension cable - rather than your PSU cable: o Ebay probably has a few - either 9" or 12"/30cm o Various online sellers do too -

do them

Someone on Ebay sells any-length of any-colour of any-rating cable. You may want to try a terminal block on the near-PSU connection, just until you know what cable length doesn't affect PC stability.

Double heatshrink the joints & watch length of solder run into the stranded cable core re mechanical inflexibility at that point.

Split-load CU using 2 RCDs, 2 butyl rubber cables on the tripping-RCD o 1ft long - radial to living-room skt other side of the CU o 15ft long - radial to landing skt, subject to vacuum cleaner attack :-)

We suspect the latter, landing skt - also linked to a new PVC-T&E ring :-) o I know the MCB has 3 wires - 1 butyl & 2x PVC-T&E o I suspect the skt has 2 butyl (1 run to disused skt) & 1 PVC-T&E

I guess they linked the radial to the ring for redundancy & that 1 disused skt.

The electrician friend did a few checks: o Opened the landing skt -- no RCD trips since o Powering each MCB in turn -- the 1ft living-room butyl MCB tripped the RCD

Sounds like cumulative leakage on 2 cables, due to insulation creep o Both are the original radial run, in butyl, to 1 socket only, in 4mm^2 o Fat cable so there's not a lot of space even in a double surface pattress

If it trips again, I'm going to repeat the tests again: o RCD instantly reset = turn off all MCBs, turn on 1 by 1 & note if RCD trips o RCD will not reset = turn off all MCBs, turn on 1 by 1 & note which trips RCD o RCD & MCB trips = hurrah, at least it shows which circuit

It could conversely be insulation pressure in the CU itself on the butyl.

Immersion is on the same RCD circuit (3 circuits), 10yr+ old element. A lot to be said for RCBOs even tho just 3 circuits on that RCD.

Frankly I think the future will be: o Verify it's a 2-PVC-Ring linked by 1-Butyl radial

---- then disconnect that radial at both ends

---- and use the 2-PVC-ring thro that landing skt o Replace that 1ft butyl radial to living room skt

---- combined that removes 2 of the 3 butyl rubber cables

The living room skt 1ft should be a 3-min change, but everyone runs: o CU is other side of the wall, displaced 12" horizontally o Cable runs 12" thro a 1-brick-wall into alcove before the fireplace o 3-5" into the alcove it goes thro the wall (behind quarry tile I guess) to skt o Above the skirting-board skt is a skirting-board chromed gas pipe to the fire

Measured the skt to living room door & same the other side re position, then measured the alcove depth v total fireplace depth - 10" = double-brick.

Electricians mumble "it's not straight thro", or "there's a chimney cavity", but the floor is solid, and there's no bermuda triangle here. Then again I'm not so sure IEE 16th onsite would like a new cable behind quarry tile skirting, tho some 25x25mm galvanised may change that (over head or on skirting :-)

Since we don't know which, it's wait for the RCD to trip again & discriminate. Historically, it hasn't and that is why the cables never got done.

Nuisance trips, then someone looks at it, then it never comes back. I think Butyl suffers insulation creep? All springy but "ductile" stuff. Rubber :-)