Geothermal Cooling?

[Orwellian]

Hello all. I have been thinking about overclocking my intel 2160. Right now its running 34-37 idle and 61-65 stress (only 6 mins in stress test, might go higher) around 22 deg ambient. stock cooler.

First, do I need to address cooling before tinkering?

Second...I'm a heat and air guy that has a bit of experience with geothermals. Since ground temp is a pretty stable 16-17c I was wondering if i could sink a coil in the ground with a water/ethanol mix and pump it to my case into a little radiator in front of a strong fan or maybe a small squirrell cage blower if I can find one. I figure I could scrabble all this together for next to nothing.

anyone done anything like this or is it too Rube Goldberg?

any guestimates at cooling improvement?

any way to control a 110v pump w/ some sort of usb switching device?

am I just dumb for thinkin about doing this?

 

[Rubycon]

That's a lot of work and I would not recommend a straight loop from CPU block to ground. If there is a loss of charge (coolant) you will be in trouble. So a pair of pumps and water-water exchanger would be required.

Pumps can be cycled on an off and that affords easiest control. Use a solid state relay (3-30VDC control, 24-250VAC load) for your pump if mains powered. Iwaki/Walchem is a good brand of small mains powered magnetic driven pump. Variable speed pumps are of special design - one cannot control pumps effectively unless designed for the application and this requires lots of transducers (another place of failure) as well as proportional derivative controllers. It's easier to use a servo valve on a recirculating loop and leave the pump run WOT all the time.

Several years ago I recall an article of a person that did something similar by running copper pipes beneath his garage slab and running cooling water through those to and from the computer system with good results. There's a good deal of coolth available but do expect a gradual rise with 500W (1.7 mbh) or more.

 

[Orwellian]

I wasn't planning on water blocks, just a water to air. that way I wouldn't be in crisis if the pump failed.

should have thought about a dc relay off the PSU, duh.

as for temp rise, I figure thats a matter of loop sizing. I can do a heat load calc to be exact, although I'll prolly just over-engineer. If I saw temps creeping up because of leaving it running for days on end, I could just water the yard over the loop for 30 mins (or just leave it off over night).

I'd probably use a hot water generator pump out of a geo. rated for eons of use, sturdy enough to take unconditioned water, and readily available to me

I'd have to stick with distilled water in a closed, possibly pressurized, loop. I'd be too afraid of static/voltage surge during one of our spring storms if I used tap water in a big copper coil in the ground...

 

[DerwenArtos12]

I'm surprised ruby didn't mention this but, how are you going to get a copper pipe from your computer into the ground? I'm assuming it would have to go through one wall, minimum, which could present a whole new set of issues. Plus, in order to get your water/ethanol mix down to the 16-17C you are definitely going to have to use a radiator to get it to ambient room temps then to your copper running under ground. I don't know where you're from but to get a year-round consistent temp of 16-17C you're going to have to bury it more than a couple feet down which is going to create a lot of resistance and require a big pump.

You'll have to isolate the copper from your computer in case of a big storm as well, it may be mostly underground but, if it's only a couple of feet it could very well still act as a lightning rod. I'll see what I can think up tonight. I'm all for new and crazy cooling ideas.

 

[Rubycon]

A Taco or Grundfos circulator should be plenty.

Originally posted by: DerwenArtos12
I'm surprised ruby didn't mention this but, how are you going to get a copper pipe from your computer into the ground?

That should be fairly easy. Transition from hard copper by brazing female adapters and use barbs/clamps or parker fittings. Quick disconnect fittings should work if they're designed to operate with liquids at lower pressures. (ones for potable water at > +20 psig will likely leak)

Grounding is a whole other issue. A water-water exchanger could be fitted with isolation nipples, grounded and fitted with flashover tubes and the residual spikes clamped with more typical protection.

 

[aigomorla]

Originally posted by: Orwellian
Hello all. I have been thinking about overclocking my intel 2160. Right now its running 34-37 idle and 61-65 stress (only 6 mins in stress test, might go higher) around 22 deg ambient. stock cooler.

First, do I need to address cooling before tinkering?

Second...I'm a heat and air guy that has a bit of experience with geothermals. Since ground temp is a pretty stable 16-17c I was wondering if i could sink a coil in the ground with a water/ethanol mix and pump it to my case into a little radiator in front of a strong fan or maybe a small squirrell cage blower if I can find one. I figure I could scrabble all this together for next to nothing.

anyone done anything like this or is it too Rube Goldberg?

any guestimates at cooling improvement?

any way to control a 110v pump w/ some sort of usb switching device?

am I just dumb for thinkin about doing this?

depends on how big of a coil you guys use.

a keg drump bomb with feed in and might also be another good idea. basically roll up coil of copper inside the drum. coolant feed.

Then the drum has another line to fill and flush.

they key is surface area, and water volumn that needs to be underground.

 

[Orwellian]

Taco is a dirty word

rig is sitting next to an outside wall right now and I own a drill

as for depth, I was guessing around 20" diameter and probably 4-5' down with a coiled copper heat exchanger. I figured I could solder some thin plate copper around the coil as well.

I have a copper heat exchanger out of a tankless water heater, but I think I would need 3-4 of them to get the transfer needed for extended use. it's prolly only 8' of linear copper tubing wrapped around a thin copper cylinder (1/2")

I live in Oklahoma, the 16-17c is year round at depth. Top couple feet changes.

 

[aigomorla]

Originally posted by: Orwellian
Taco is a dirty word

rig is sitting next to an outside wall right now and I own a drill

as for depth, I was guessing around 20" diameter and probably 4-5' down with a coiled copper heat exchanger. I figured I could solder some thin plate copper around the coil as well.

I have a copper heat exchanger out of a tankless water heater, but I think I would need 3-4 of them to get the transfer needed for extended use. it's prolly only 8' of linear copper tubing wrapped around a thin copper cylinder (1/2")

I live in Oklahoma, the 16-17c is year round at depth. Top couple feet changes.

what you really need to do is take this question onto xtremesystems.org/forums

many many many creative people, who have done things like this goto that site to look at new results and products. For special things like this, they would have the best answers.

the keg coil method is something i remember reading off there.

 

[DerwenArtos12]

Originally posted by: Orwellian
Taco is a dirty word

rig is sitting next to an outside wall right now and I own a drill

as for depth, I was guessing around 20" diameter and probably 4-5' down with a coiled copper heat exchanger. I figured I could solder some thin plate copper around the coil as well.

I have a copper heat exchanger out of a tankless water heater, but I think I would need 3-4 of them to get the transfer needed for extended use. it's prolly only 8' of linear copper tubing wrapped around a thin copper cylinder (1/2")

I live in Oklahoma, the 16-17c is year round at depth. Top couple feet changes.

Should work. 4-5' is exactly what I was thinking. Probably do 20' of copper coil. I still think ruby is right on about using a heat exchanger. That way you could keep the noisy super-high head pump ouside and just insulate everything from the coil to the pump and then run a feed and a return line inside to the exchanger. so that you can minimize the lentgh of the loop you're using to your water blocks and keep volume down.

 

[Rubycon]

Originally posted by: DerwenArtos12

Should work. 4-5' is exactly what I was thinking. Probably do 20' of copper coil. I still think ruby is right on about using a heat exchanger. That way you could keep the noisy super-high head pump ouside and just insulate everything from the coil to the pump and then run a feed and a return line inside to the exchanger. so that you can minimize the lentgh of the loop you're using to your water blocks and keep volume down.

Not only that but circulators will superimpose a buzzing on the lines that is greatly amplified if they touch a flat surface. The changeover to tubing will prevent this, however.

 

[Orwellian]

you guys are stuck on the direct water cooling. I was figuring on doing water cooled air, at least for a few weeks to make sure I didn't have any failures or leaks. I'd run the ground cooled water into a radiator and blow that air into case. If it turned out to be reliable and the water temps stayed mostly flat, I could then buy some water blocks for the cpu and gpu and go to direct water cooling.

I would want to test everything and make sure the ground loop could handle the heat load before I invested in water blocks. If it was stable I could test it for direct cooling by dropping a 500w element in the water circuit (taking the computer out of the system) and leave it on for a few days and check to see if I have any temp rise from ground warming (doubtfull).

picture in my head:

~20' of 1/2" soft copper tubing coiled, leaving 2-3" in between coils
braze 3-4 sheet copper strips a couple inches wide up the sides of the coil cylinder
have several other copper sheet strips running through the coil across the diameter and extending out the sides of the coil several inches

adapt to 1/2" PEX piping before leaving ground level and extending the PEX into my computer room.

I would have a "T" on each line before the computer case in which I could put these nifty ports in that allow you to shove a spike thermometer into the water stream without opening the circuit.

from the open side of the "T"s I'd probably stay PEX into the computer case. I'll lose a little flexibility but I have all sorts of PEX stuff and I trust those fittings at any concievable pressure I'll be running. Connect to a radiator type HX and blow that sub-ambient air into the case.

I figure I could get rid of any possible condensation on high dew point days by sealing the case/radiator/blower setup off and just recycling the air if the ground cooling is efficient enough (that may be stretching)

 

[aigomorla]

Originally posted by: Orwellian
you guys are stuck on the direct water cooling. I was figuring on doing water cooled air, at least for a few weeks to make sure I didn't have any failures or leaks. I'd run the ground cooled water into a radiator and blow that air into case. If it turned out to be reliable and the water temps stayed mostly flat, I could then buy some water blocks for the cpu and gpu and go to direct water cooling.

I would want to test everything and make sure the ground loop could handle the heat load before I invested in water blocks. If it was stable I could test it for direct cooling by dropping a 500w element in the water circuit (taking the computer out of the system) and leave it on for a few days and check to see if I have any temp rise from ground warming (doubtfull).

picture in my head:

~20' of 1/2" soft copper tubing coiled, leaving 2-3" in between coils
braze 3-4 sheet copper strips a couple inches wide up the sides of the coil cylinder
have several other copper sheet strips running through the coil across the diameter and extending out the sides of the coil several inches

adapt to 1/2" PEX piping before leaving ground level and extending the PEX into my computer room.

I would have a "T" on each line before the computer case in which I could put these nifty ports in that allow you to shove a spike thermometer into the water stream without opening the circuit.

from the open side of the "T"s I'd probably stay PEX into the computer case. I'll lose a little flexibility but I have all sorts of PEX stuff and I trust those fittings at any concievable pressure I'll be running. Connect to a radiator type HX and blow that sub-ambient air into the case.

I figure I could get rid of any possible condensation on high dew point days by sealing the case/radiator/blower setup off and just recycling the air if the ground cooling is efficient enough (that may be stretching)

ahhhh so instead of using the water directly on your computer, you want to use a radiator to vent cold air into the server room? So kinda like a geothermal AC that spits out 16-17C air all year long?

Is it possible to go down deeper below the frost point and get colder earth? 16-17C is like 60F. You can do that at night with the windows open.


If you can take coolant down to 10-15C, you might run into some condensation issues, but since nothing electrical besides the pump will be in contact i dont see a problem. Your ambient would be much cooler. Only problem now is how you intend to vent the cold air? If you use a radiator, like a MCR320, it would be able to vent that cold air however i dont know how long it would stay cold. An MCR320 is capable of disapating around 400-450W of heat depending on fan. That means it could pull a lot of cooling off the earth over a extended period of time.

I havent worked with geothermal so i dont know how steady the temps stays.
How steady is the temp down there exactly? i know if you dig deep enough its enough for to supply cooling and heating for a large building. But something on that small of a level.. how effecient would it be?

 

[Orwellian]

I guess I'm not describing my vision well, the cold water radiator would be inside my computer case (or a small addon to computer case) along with a blower. I just didn't want to go straight to water blocks on the cpu and gpu without some run time on the system. Biggest thing would be to have close to 60f ambient air to use as cooling. In the summer I run the A/C at ~76f and furnace at ~72f in the winter. This would just supply better air inside the case for cooling. If it worked well, I would then go to direct water cooling to components.

Its not the depth that does the cooling, once you get a couple feet down the temp is gonna stay constant for thousands of feet. The only reason big buildings (and houses) dig so far down is to get the surface area. You can get the same effect by putting the geothermal loop horizontal in a ditch that is only a couple feet deep. You just need a reaaaallly long ditch. We did a couple houses in the early days of geos like that. Now everyone just puts the loops vertically in ~150' deep wells, number of wells depends on the size of the house, usually 3-5 unless you are doing really big stuff or the type of ground in the area has really bad heat conductivity. Also Geos just use water circulated in the ground to cool or heat a refrigerant circuit that pretty much just like the one in your air conditioner now. They aint as fancy complex as people think they are.

 

[aigomorla]

You have more knoweldge about the cooling part in this better then i do, So im gonna start with this: as long as you can keep the coolant temp constant.

I dont see why this wouldnt work. I mean you'd be constantly getting air cooled near 60F.

Well, it would be slightly higher because you cant assume 100% coldness transfer.


As far as i know there is no way to control a 110V pump on AC, maybe a light dimmer? Also id hate to see the pump required for something like this. To be honest, i thought my pump was "industrial" but your going to need something a lot more beefier.

But assuming you have a good cool transfer, i think your exhaust air would be somewhere around 65-70F dependant on fan speed.


its so much more efficent if not better to apply that water directly on the blocks.

Heres my cooling @ 16-17C ambients. Coolant temps was around 20.1C on the return.

http://i125.photobucket.com/al.../aigomorla/Temps-2.jpg

these are loaded temps btw. Normally i wouldnt say compare with these... but seeing how you can keep your coolant ambients lower then mine... And you'd probably have a pump with MASSIVE headpressure controling it, id say you'd have one crazy cooling setup.

Probably would even rape mine in capacity depeding on how well you could control that water temp, and headpressure output on the pump.

 

[Rubycon]

Forget about trying to control the speed of the pump. Use a bypass valve. Its position can be controlled by a servo which would get its signal from a temperature controller.

 

[DerwenArtos12]

I don't want to thread jack here but, seeing as you're an HVAC guy I was wondering if I could get a little input from you in my thread about building an AC unit into a case.

 

[Rubycon]

Originally posted by: DerwenArtos12
I don't want to thread jack here but, seeing as you're an HVAC guy I was wondering if I could get a little input from you in my thread about building an AC unit into a case.

Post specs and pics of your (cooling) hardware would help greatly.

 

[DerwenArtos12]

Originally posted by: Rubycon

Originally posted by: DerwenArtos12
I don't want to thread jack here but, seeing as you're an HVAC guy I was wondering if I could get a little input from you in my thread about building an AC unit into a case.

Post specs and pics of your (cooling) hardware would help greatly.

It would, wouldn't it? I forgot that I forgot to do that. I redesigned the case on paper last night and I'll get it into paint this morning, right after taking pics of the AC I tore apart.

 

[Orwellian]

why do I need to control the pump? can't it just be full bore all the time its on? or is that not very engeneery of me?

 

[Rubycon]

Originally posted by: Orwellian
why do I need to control the pump? can't it just be full bore all the time its on? or is that not very engeneery of me?

Yes the pump should be running WOT (Wide Open Throttle) all the time. If flow needs to be controlled, a simple valve to increase head pressure at the discharge can be used (as long as the pump type is designed for this - motor amps go DOWN as head pressure goes up as most centrifugal pumps do). Otherwise a recirculation loop is needed with a bypass valve that allows for reduced main loop flow without changing the head pressure on the pump. Both methods provide good control of flow through areas of dwell.