Is there a use for Peltiers in Silent computing?

The forum for non-component-related silent pc discussions.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Bob_the_lost
Posts: 136
Joined: Sun Aug 21, 2005 1:52 pm

Is there a use for Peltiers in Silent computing?

Post by Bob_the_lost » Fri Aug 26, 2005 7:33 am

Yeah, the work of a very bored person, but i'm wondering if it would be possible to use a peltier and a large passive heat sink to try and disspate heat rather than a heatsink and fan? There's a shop in the UK that sells peltiers for prices starting at £9.00 or so for a tiny little one, might be worth a try sometime, or am i just getting really confused. Any ideas and has it been tried before?

ckolivas
Posts: 393
Joined: Thu Jan 27, 2005 7:16 am
Location: Melbourne, Australia
Contact:

Post by ckolivas » Fri Aug 26, 2005 7:47 am

Peltiers increase the total amount of heat you need to dissipate therefore they actually increase the amount of cooling you require. ie it's not really a silencing device so much as an overclocking one because of the potential for much lower temperatures.

Bob_the_lost
Posts: 136
Joined: Sun Aug 21, 2005 1:52 pm

Post by Bob_the_lost » Fri Aug 26, 2005 7:54 am

I have been thinking of a use for them some more, imagine cutting a hole in the top of a large case, placing a peltier element in the hole and fixing a heatsink on both sides, the one on the outside of the case being heated by the peltier, the one on the inside being cooled by it.

(Thus reducing the temp inside the case, possibly replacing case fans.)

ckolivas
Posts: 393
Joined: Thu Jan 27, 2005 7:16 am
Location: Melbourne, Australia
Contact:

Post by ckolivas » Fri Aug 26, 2005 7:57 am

Provided you're careful about condensation which is truly dangerous to electronics with peltiers, feel free to try and don't forget to post some pictures and info here! About the only thing I can suggest is to use the lowest powered peltier you can get your hands on. The more power, the more heat and the more danger.

Rusty075
SPCR Reviewer
Posts: 4000
Joined: Sun Aug 11, 2002 3:26 pm
Location: Phoenix, AZ
Contact:

Post by Rusty075 » Fri Aug 26, 2005 8:00 am

I had very similar thoughts while working with the ActiveCool AC4G Thermoelectric cooler.

The upside is that as the passive HS gets hotter, its ability to dissipate heat goes up. Double the dT, and the wattage approximately doubles. It doesn't matter if your HS is cooking along at 90°, so long as the CPU side of the TEC stays cool.

The downsides are multiple though:

1. The extra wattage required by the TEC has to be produced without adding noise or heat to the case. (no easy feat, that's like adding a second CPU)
2. The increase in cooling power of the passive heatsink has to be enough to allow it to keep up with the now nearly double wattage coming from the TEC. Instead of just dealing with a say 75W CPU, it now has to dissipate that 75 plus another 50-100watts of "waste heat" from the TEC.
3. You have to keep the TEC below about 100°, or it will lead a very short life. So you can't just keep cranking up the heat of the HS until it glows. :wink:
4. You need a way to control the TEC, so that you don't have to worry about the CPU getting too cold when the load is low. Condensation is a bad thing. The AC4G is probably on the right track with their system.

I think it could be done, but not easily. Might be a fun project to tinker with.

Bob_the_lost
Posts: 136
Joined: Sun Aug 21, 2005 1:52 pm

Post by Bob_the_lost » Fri Aug 26, 2005 8:10 am

To do this the least expensive bits will be the peltiers if my suscpicions are right, thinking of going for one of the 30x30 ones here: http://www.sciencestore.co.uk/StorePeltier.htm to trial it. Then i'd need a lab style power supply (if i asked nice i could probably borrow one from my old school) and a good thermometer to test it's performance.

For the hot side i could go with something like the XP120, it'd work even better than the test situations as the contact area would be greater.

For the cold side i'd need another heatsink, the scythe ninja possibly, although i've no idea how that would work when inverted! The alternative to these two include for the hot side a great big metal plate on the outside of the PC and several low powered Peltiers connected to passive heatsinks on the inside of the case...

So many bad ideas, so little time!

(Oh and in all likelyhood two new PSUs to cope with the power drain!) :lol:

IsaacKuo
Posts: 1705
Joined: Fri Jan 23, 2004 7:50 am
Location: Baton Rouge, Louisiana

Post by IsaacKuo » Fri Aug 26, 2005 9:05 am

You're talking about mixing up heat pipes and peltiers, which to me sounds too complex to start off with.

In order for (wicked) heat pipes to operate, two things are required:

1. The hot side must be hot enough to boil the working fluid.

and

2. The cold side must be cool enough to condense the working fluid.

This suggests a relatively narrow range of temperatures you're working with. You can't just crank up the hot side to 90 degrees if it means that even the cool side is too hot to condense the working fluid.

My gut feeling is that you'd do better by just going all heatpipe than trying to do heatpipe->peltier->heatpipe as you're thinking.

If I were experimenting with this concept, I'd be thinking in terms of placing the computer components in a full enclosure to block noise. The components would be mounted to one side of the enclosure, while the peltiers would be mounted within the opposite side (sandwiched between large aluminum heat sinks). Thus, if there is condensation on the inner heat sink, it will at worst drip to the bottom of the enclosure where it doesn't threaten to damage any components.

ronrem
Posts: 1066
Joined: Sun Jan 16, 2005 2:59 am
Location: Santa Cruz

Post by ronrem » Fri Aug 26, 2005 1:22 pm

IsaacKuo wrote:You're talking about mixing up heat pipes and peltiers, which to me sounds too complex to start off with.

In order for (wicked) heat pipes to operate, two things are required:

1. The hot side must be hot enough to boil the working fluid.

and

2. The cold side must be cool enough to condense the working fluid.

This suggests a relatively narrow range of temperatures you're working with. You can't just crank up the hot side to 90 degrees if it means that even the cool side is too hot to condense the working fluid.
This is sort of what I was thinking about on another thread,Peltiers NOT right on the CPU/mobo on the outer side of the case or an aluminum intake duct so the hotside does not heat the innards of the case,and the air in the case gets chilled
My gut feeling is that you'd do better by just going all heatpipe than trying to do heatpipe->peltier->heatpipe as you're thinking.

If I were experimenting with this concept, I'd be thinking in terms of placing the computer components in a full enclosure to block noise. The components would be mounted to one side of the enclosure, while the peltiers would be mounted within the opposite side (sandwiched between large aluminum heat sinks). Thus, if there is condensation on the inner heat sink, it will at worst drip to the bottom of the enclosure where it doesn't threaten to damage any components.
On another thread I was also wondering about a new way to employ a peltier. Big issues are condensatin on the CPU or mobo,and getting rid of the heat output. I have some hollow aluminum beams,about 2x4 cross section and thought about a few feet of one turned into an intake duct. Projecting the duct beyond the case,a peltier could be outside the box,and chilling that aluminum,the walls of which are about 1/4" thick,I could add some internal fins,baffles etc. The idea,the peltier makes the aluminum duct cold which chills incoming air,also, that aluminum will "radiate" coldness in the box. I figured on a heatsink on the peltier's hotside-outside the case,as I said,and I'd deflect the PSU exhaust air to cool that external heatsink. I figure I likely need to do some ductwork so the air-inlet is not drawing the heated air coming off the peltier hotside. Too bad I can't just up a sketch.

I also had thought about a case-wall mounted big heatsink with an external peltier,I was thinking though of heatpipes transmitting CPU,chipset heat to the wallmount heatsink then the peltier is chilling that heatsink. The idea in this thread is simpler-at least if you use big but basic copper heatsinks, A couple of $50 heatpipe cpu coolers,like Ninja's would get expensive. I'm thinking more like a pair of finned sinks,say 4"x9"x1"thick,sanwiching the peltier. I have seen some "Heatlane" devices,that are lke 2" x 6",less than 1/2" thick,they are like big flattened heatpipes. In a sandwich a heatlane would distribute the heat output of a small Peltier more evenly to a much larger sink.

I also wonder about a "passive" water cool where there is a loop with a waterblock on each end,one block is on the coldside of a peltier,the other is on the CPU-a heatsource. Would the temp range create a flow without a pump? Could it work with a reservoir?

Maybe a rather small low watt peltier could enhance a watercool system enough that a much smaller,quiet,low flow pump could yield great results? It could be interesting to explore the possibilites.

As effective as todays big heatpipe passives are,if a Peltier just lowered case internal temps,say 5-10 degrees that could be major
Last edited by ronrem on Fri Aug 26, 2005 7:19 pm, edited 1 time in total.

Bob_the_lost
Posts: 136
Joined: Sun Aug 21, 2005 1:52 pm

Post by Bob_the_lost » Fri Aug 26, 2005 1:39 pm

Yeah, that makes sense. This is why i asked you lot before i bought any of this stuff :D

For the hot side i could go for an all singing/dancing XP120, which would no doubt work rather well. Or i could try and find or make a large metal heatsink that would cover the top of the case, am i right in thinking that it would be more efficent?

For the cold side heat pipes are out, there's no point to them. Either several small Heatsinks (if more than one peltier is to be used, would work well with the large metal heatsink idea) or a custom built one (again)

Condensation may be a problem, the easy way to get round it would be to mount some sort of tray underneath the heatsink to catch any runnoff :p or i could just turn the voltage down till it's not a problem. I don't think that insulation is appropriate here, as the cooled side isn't a CPU but a wholel lump of metal that has to be isolated (and thereby removing the point of it) :roll:

The case would be sealed, that way the acoustics from the Hard Drives could be reduced more, and if i used any fans inside the case (to move more air past the heat sink) they'd be quieter than quiet, although i'm not sure if i want to do that.

Now, the biggy, do i buy one of the little ones i linked to earlier, or a sodding massive one off ebay? (£17.50 for a 170Watt on BIN, or maybe the same price for a 370W on auction!?)

cotdt
Posts: 295
Joined: Tue May 10, 2005 3:30 pm

Post by cotdt » Fri Aug 26, 2005 1:50 pm

No need for speculation. Passive cooling via peltier has already been shown to work well, especially with low-power CPUs like the 90nm Athlons (overclocked=55W). The trick is to use low voltage (ie. 5V from PSU), but I think the sweet spot would be to use 7V from a 350W Peltier. At 14V full 350W, the efficiency is only 5% giving peltiers a bad rep, but at low votlages the efficiency is far, far higher. There is a squared relationship between voltage and power. I know that a 350W peltier drawing power from the 5V rail eats up about 50W but is pretty efficient so doesn't add too much extra heat. CPU temps stayed at about 20-30C though the large heatpiped heatsink (similar to Zalman TN500) was hot to the touch. There was no need for condensation insulation though insulation would be a good precaution.

Like I said, it's been shown to be effective, but the problem is that the C/W curve is exponential. So it is not versatile, on a CPU that is just 10W higher than intended, it can be easily overloaded. On idle, it can potentially cause condensation, although insulation fixes that problem. By comparison, a watercooling system has a relatively constant C/W, and an aircooled heatsink has a linear C/W that climbs up slowly. So I think that's the biggest problem of peltiers. Electronic voltage control might fix this problem.

I'm pretty sure the XP-120 would be too weak. Usually passive heatsinks are cut to match the dimensions of the side of your case, maybe 40cm by 30cm, and one inch thick. They are about $15-30 and made out of aluminum. I'm pretty sure copper would perform much better as a passive than aluminum, but aluminum seems good enough. Four heatblocks near every corner of the passive heatsink, each connected to the CPU through a heatpipe. A thermosyphon would also work great.
Last edited by cotdt on Fri Aug 26, 2005 2:38 pm, edited 1 time in total.

Bob_the_lost
Posts: 136
Joined: Sun Aug 21, 2005 1:52 pm

Post by Bob_the_lost » Fri Aug 26, 2005 2:23 pm

Hrm, i didn't realise how inefficent they were :oops: This might still end up as a TEWC (Technical Excersise Without Components, and if anyone gets where that is from i'll be rather suprised)

Off to find spec sheets for Peltiers!

cotdt
Posts: 295
Joined: Tue May 10, 2005 3:30 pm

Post by cotdt » Fri Aug 26, 2005 2:40 pm

Try the highest-powered peltier you can find and give it your PSU's 5V rail and see if that works. They should be pretty efficient at low voltages. Don't forget the anti-condensation insulation just to be safe. This is a lot of work, but if you can TAKE UP THE CAUSE we will thank you for it. The idea itself should work =).

It might be easier to do this with a watercooling system rather than with heatpipes, but heatpipes are cheaper and dead silent so of course heatpipes are the best. They are just a lot more work since there are not as many premade stuff for them.

Bob_the_lost
Posts: 136
Joined: Sun Aug 21, 2005 1:52 pm

Post by Bob_the_lost » Fri Aug 26, 2005 3:11 pm

I'm gonna sleep on it, as far as i can find on the web no one has been stupid enough to try this idea, or if they have they were smart and did it quietly so no one laughed at them.

cotdt
Posts: 295
Joined: Tue May 10, 2005 3:30 pm

Post by cotdt » Fri Aug 26, 2005 3:19 pm

This is similar:
http://overclockers.com/tips1211/

But it seriously needs refinement, because there is no pump on the watercooler. However, combined with this idea:
http://overclockers.com/articles1248/

Those two ideas combined with a large passive heatsink/radiator would be what we want. It will be cheaper than any watercooling kit and dead silent. Unlike heatpipes, thermosyphons are bendable and very cheap, and can be any length.

ronrem
Posts: 1066
Joined: Sun Jan 16, 2005 2:59 am
Location: Santa Cruz

Post by ronrem » Fri Aug 26, 2005 7:51 pm

Direct,on-the CPU peltiers do heat the innards of the case-potentially put condensation in all the worst places-but the concept here is to move the peltier away from the CPU in such a way that the hotside is outside and the coldside is off the mobo maybe "refrigerating" the inside of the case,maybe pre-cooling inlet airflow,maybe chilling some type of big heatsink linked to CPU (and whatever else) by heatpipes,by active or passive watercool,thermosyphons...whatever.

If a 3800 X2 can be cooled with one or two 500 rpm 120mm fans that would be pretty sweet. If something with that sort of power can be done totally fanless-wow.. I think,realistically,having at least a PSU fan would work best. Getting rid of the peltiers hotside heat is important. How big would a heatsink need to be to not include a fan? Lots of experiments to be done.

DryFire
Posts: 1076
Joined: Sun May 25, 2003 8:29 am
Location: USA

Post by DryFire » Fri Aug 26, 2005 8:52 pm

another advantage to peltiers is that you have more surface area to remove the heat from (though it may be a lot more). 900mm^2 is a good deal larger then then 100-200mm^2.

What if you ran the cpu at full load all the time (folding) would boot up be enough time to make the cpu too cold?

mathias
Posts: 2057
Joined: Sun Jul 11, 2004 3:58 pm
Location: Toronto
Contact:

Post by mathias » Fri Aug 26, 2005 9:20 pm

I've heard about running peltiers at 5v, but what about at 3.3v?

cotdt
Posts: 295
Joined: Tue May 10, 2005 3:30 pm

Post by cotdt » Fri Aug 26, 2005 10:23 pm

mathias wrote:I've heard about running peltiers at 5v, but what about at 3.3v?
Someone should try 3.3V, but I speculate that anything over 40W will overload it at 3.3V. For a 350W peltier, testing shows that at 5V it overloads at 65W. At 7V it should be able to handle anything, but anti-condensation insulation might be required if peltier is mounted on CPU. Although you can probably somehow combine the 3.3V and the 5V to get 8.3V. 5V and 90nm Athlons seem to get along really well.

500RPM fans? I'm pretty sure a large heatsink can do it passive with good temps. Similar to the Zalman TN-300/TN-500, but much cheaper and portable. The peltier can be mounted onto the large heatsink rather than the CPU to avoid condensation worries. Hey maybe a large heatsink won't even be required and a Syche Ninja will do, but this really needs to be tested. The problem with this seems to be ridding the heat buildup inside the case. And even a quiet fan blowing the case heat ruins the whole purpose of passive cooling. So I think the heatsink should be outside of the case.

In other news, Procooling.com announced that their next focus will be silent computing. With their large expertise, SPCR will get a boost.

Bob_the_lost
Posts: 136
Joined: Sun Aug 21, 2005 1:52 pm

Post by Bob_the_lost » Sat Aug 27, 2005 1:03 am

Ok, more thought has gone into this. A small, low powered rig uses something like 150W of power, a peltier would be more than able to remove that much heat, the internal case temperature would then stabalise somewhere around ambient. If the temp inside the case were to be lower than that outside heat would start to move inwards, and it'll become less efficent. So with a 370W TEC you'd have enough power to disipate the heat of even a high end system.

Now, what would the point of this be? The only reason to use a peltier to cool the inside of the case would be so you can reduce the airflow, or to make a quiet system run more stably by lowering the temperature of the components. It'd be able to keep a sealed box system at room temperature (roughly) and since all openings could be sealed noise would be minimal.

Sod it, let's have a go. Worst case i'll have a spare peltier to play with and a new PSU + CPU heasink. (Oh and a mulilated case :lol: )

Bluefront
*Lifetime Patron*
Posts: 5316
Joined: Sat Jan 18, 2003 2:19 pm
Location: St Louis (county) Missouri USA

Post by Bluefront » Sat Aug 27, 2005 3:58 am

I'm not sure if I'm following your plan correctly. You'll have a big, passive heatpipe/heatsink inside the case, attached to a peltier cooler on an outside case panel. On the other side of the peltier will be another big passive heatsink outside the case.

If this is what you're planning, I think Issac covered the outcome. I really don't think the inside heatsink (which is not attached to the CPU), will ever get hot enough to work well. Heatpipes in particular don't start working till they get really hot......and unless directly attached to a heat source, will be less efficient than a plain finned heatsink.

I've tried placing a finned heatsink on the inside of a case, directing the CPU fan right at the heatsink. This inside heatsink was directly(thermally) attached to another heatsink outside the case. The whole setup worked somewhat....the outside heatsink did get warmer, but not enough to make it worth the trouble. I doubt a peltier between the two heatsinks would make it work much better.

Maybe worth a try however.....

Bob_the_lost
Posts: 136
Joined: Sun Aug 21, 2005 1:52 pm

Post by Bob_the_lost » Sat Aug 27, 2005 4:45 am

Yeah, still thinking on this one, but the internal heatsink would be very cold, CPUs can be kept in minus temps after all, so the heatsink inside would be at say -10, assuming the ambient temp in the case was 30 then you've got a 40 degree temperature difference. Add that in with a fan and you'd get a notable cooling effect.

However i am edging towards a much smaller, less ambitious use for a peltier, for a silent machine, more later when i've got a plan that makes some sense.

SoopahMan
Posts: 50
Joined: Fri Aug 19, 2005 6:22 pm
Location: North Hollywood, CA, USA
Contact:

Post by SoopahMan » Sat Aug 27, 2005 9:23 pm

Peltiers are really making the rounds after that Utah science fair article hit Slashdot... .

As long as I have the eyes of some people who know something about Peltiers can anyone riddle me this? Would it be possible to use the massive heat created by a GPU to generate electricity? Peltier chips have 2 modes as I understand it, one pushing power to create a differential, and the other taking advantage of a differential to create power. Wouldn't a GPU be an opportune case of dramatic differential and opportunity to create power? And doesn't this conversion of heat to electricity reduce the total heat in the system? Wouldn't that provide some cooling to the GPU?

Anyone? :O)

shoebox9
Posts: 76
Joined: Mon Jan 31, 2005 3:50 am
Location: Australia
Contact:

Post by shoebox9 » Sun Aug 28, 2005 3:29 am

An idea I'm itching to see if anyone's done is-

Use a large undervolted peltier with a water block on each side- one side cools the CPU etc, the other side goes to a large passive radiator (ie a car radiator).

Why not just leave out the peltier altogether? Well the radiator would work heaps better when hotter, meaning lower temps for the same passive cooling. (Am I right here?)

The only problem I can think of would be condensation caused by cold pipes etc, inside the case. PLEASE soneone give us a large peltier with two temp probes (ie one for ambient, and one in it's cold plate) and a control unit.

By adjusting the voltage to the peltier, a cold side temp of say 3 degrees above ambient could always be maintained. Attach a heatsink or water block of your choice to the hot side, & yahoo- off we go!

Cheers,
Shoebox9

cotdt
Posts: 295
Joined: Tue May 10, 2005 3:30 pm

Post by cotdt » Sun Aug 28, 2005 2:43 pm

Here is some data for a 226W peltier. At 35W heat load, 9V gives you -11C while 15V gives you -14C, so the temperature difference is very small but the difference in wattage used is huge. The efficiency of low voltage peltiers is quite good. 370W peltiers can be extrapolated. Of course don't expect such low temps when passive cooling with 5V. The heat dump from the peltier is probably around 90% of its total wattage.

226 watt / 15.2 VMax / 24 IMax
=====================
05 volts: 24.5 watts; 7.89 amps
06 volts: 35.2 watts; 9.47 amps
07 volts: 47.9 watts; 11.05 amps
08 volts: 62.6 watts; 12.63 amps
09 volts: 79.2 watts; 14.21 amps
10 volts: 97.8 watts; 15.78 amps
11 volts: 118.4 watts; 17.36 amps
12 volts: 140.9 watts; 18.94 amps
13 volts: 165.3 watts; 20.52 amps
14 volts: 191.7 watts; 22.10 amps
15 volts: 220.1 watts; 23.68 amps

120 watts / 24.6 VMax / 7.9 IMax
======================
05 volts: 3.5 watts; 1.60 amps
06 volts: 5.5 watts; 1.92 amps
07 volts: 8.0 watts; 2.24 amps
08 volts: 10.8 watts; 2.56 amps
09 volts: 14.1 watts; 2.89 amps
10 volts: 17.9 watts; 3.21 amps
11 volts: 22.1 watts; 3.53 amps
12 volts: 26.7 watts; 3.85 amps
13 volts: 31.0 watts; 4.17 amps
14 volts: 36.4 watts; 4.49 amps
15 volts: 42.2 watts; 4.81 amps
16 volts: 48.5 watts; 5.13 amps
17 volts: 55.2 watts; 5.45 amps
18 volts: 62.3 watts; 5.78 amps
19 volts: 69.8 watts; 6.1 amps
20 volts: 77.8 watts; 6.42 amps
21 volts: 86.2 watts; 6.74 amps
22 volts: 95.0 watts; 7.06 amps
23 volts: 104.3 watts; 7.38 amps
24 volts: 114.0 watts; 7.70 amps

SoopahMan - yes you can use it to generate electricity but what for? your GPU will be hot! and GPU temps are more important than even CPU temps for gaming. peltiers can be used as a voltage source like batteries, but here we are solely interested in cooling right?

shoebox - two peltiers will be too cold! People have gotten colder than -40C from that! here at SPCR that is not our purpose!

To answer: why not leave the peltier out? Well for CPU's like the Dothan that can handle 100C, yes a peltier is not needed. But desktop CPUs should never go above 65C, and we all know that passive cooling works best when temperature differential is high. The higher the temperature differential the better, way better. Also, similar setups to what SPCRians are interested in have already been tried and showed to work well.

The idea is to have the peltier far from the CPU, maybe even outside the case, so that we don't have to deal with annoying anti-condensation insulation. Voltage-controlling the peltier is another very good alternative, similar to automatic fan controllers. Or if you run Folding@HOME, SETI@HOME, etc., then you don't need anything (but it might still cause some worries).
Last edited by cotdt on Sun Aug 28, 2005 2:53 pm, edited 1 time in total.

shoebox9
Posts: 76
Joined: Mon Jan 31, 2005 3:50 am
Location: Australia
Contact:

Post by shoebox9 » Sun Aug 28, 2005 6:02 pm

Thanks for the above lines of enquiry. Very inspiring.

I'm now planning to use a peltier to cool a case water loop, and passively cool the peltier using a large radiator. 2 pumps, one peliter, no fans. To do this I'd insert the peltier into a normal water loop, with a water block attached to each side.

So I'd get: cooler water (potentially too cool without insulation) and hotter (better/more efficent!?) radiation from the radiator. Also, by adding a normal water cooling radiator inside the case on the cold side of the water loop, I'm hoping to produce internal air conditioning, & make the case totally air tight, apart from PSUs, which will need more thought.

I'll post a report of how I go using these ideas, a 350w peltier, and an x2 4400+, once I get it all completed and evaluated. I'll use a variable voltage PS so I can find & stay at the sweet spot for my system.

Question- If I am piping off the heat using w/cooling, is there still any advantage having the TEC outside the case? The downside would be having a w/pump outside the case as well. ?? I've never even w/cooled before, so I don't understand how placing the TEC inside the case would contribute to condensation- since both sides would be having the heat/cold wisked away by the water.

ronrem
Posts: 1066
Joined: Sun Jan 16, 2005 2:59 am
Location: Santa Cruz

Post by ronrem » Fri Nov 18, 2005 6:38 pm

shoebox9 wrote:Thanks for the above lines of enquiry. Very inspiring.

I'm now planning to use a peltier to cool a case water loop, and passively cool the peltier using a large radiator. 2 pumps, one peliter, no fans. To do this I'd insert the peltier into a normal water loop, with a water block attached to each side.

So I'd get: cooler water (potentially too cool without insulation) and hotter (better/more efficent!?) radiation from the radiator. Also, by adding a normal water cooling radiator inside the case on the cold side of the water loop, I'm hoping to produce internal air conditioning, & make the case totally air tight, apart from PSUs, which will need more thought.

I'll post a report of how I go using these ideas, a 350w peltier, and an x2 4400+, once I get it all completed and evaluated. I'll use a variable voltage PS so I can find & stay at the sweet spot for my system.

Question- If I am piping off the heat using w/cooling, is there still any advantage having the TEC outside the case? The downside would be having a w/pump outside the case as well. ?? I've never even w/cooled before, so I don't understand how placing the TEC inside the case would contribute to condensation- since both sides would be having the heat/cold wisked away by the water.
The idea that hotter makes a radiator work "better" is somewhat true,as well as very misleading. Let us say water at 180f is going to the radiator. Room temp is 75 f. The radiator is able to lower the temp to 145 F. a 35 degree drop. Now lets say the system is sending water at 130 F to the radiator which drops the temp to 110 F,a 20 degree drop. You could say that is less drop or less efficiency-but what COUNTS is that the water circulates back at 110,not 145. More heat just means more you want to get rid of,irregardless of efficiency.

As I understand it,effectively getting heat away from the hotside is essential so the coldside can do its cold thing. Bringing below ambient coldness in contact with a hot CPU is where the condensation gets significant. I think letting the coldside just chill the in-case air is less apt to have condensation side effects. You DO NOT want to use the same water loop to do coldside>CPU and hotside>heat exchanger. Much of your cooling effect would be lost,you'd have the hotside and the CPU in the loop making heat,with the coldside and radiator trying to lose that heat,and your gain would not be much. You COULD use 2 pumps,2 seperate loops,but that adds up cost. By doing a WC loop of only the hotside and the radiator,you can offload heat enough that the coldside can get cold. You mount a sizeable finned heatsink on the cold side,or mount it to an aluminum intake duct,letting that cold aluminum (or copper?)radiate chill like a block of ice in a picnic cooler.

With this,a Venice,a Scythe 2005 or Ninja,a psu with a single 120mm or 140 mm fan should be plenty.

shoebox9
Posts: 76
Joined: Mon Jan 31, 2005 3:50 am
Location: Australia
Contact:

Post by shoebox9 » Fri Nov 18, 2005 8:03 pm

Thanks everyone for your contributions.

I had worked on the ideas further, was just in the process of starting to build my system (which had developed to a 10cm thick walled poured cement case with one open side facing down, and all wiring & plumbing in place before pouring the cement. Case would be tilted/jacked up for PC access) when adding a new employee to the business necessitated my changing offices.

Guess what- my new office has a wash room with a built in cupboard directly through the internal cement brick wall from where my desk will be! Now I'll take the easy way out & simply put every cable I might ever need through the wall, and bog it back up again.

Sort of shame to abandon such a fun project- in time maybe the day will come when I'll need to pull the design out again...

Cheers,
Shoebox9

Dragon Puppy
Posts: 13
Joined: Sat Apr 24, 2004 6:34 am

Post by Dragon Puppy » Mon Nov 21, 2005 8:46 am

I havent read all the postings , but if somebody wats to play with an peltier calculator theres one at http://www.kryotherm.com/Software%20Kryotherm.htm for their peltiers.

Playing with it I found that 172W 24.6V pelt with an 50W heat load:

1 peltier @ 12V = power draw 64W + cooling 45W = heat to hot side 110W
2 pelts(parallel) @ 6V = power draw 32W + cooling 50W = heat to hot side 82W

So power draw decreases to an quarter with every halving of volt like cpus?
Is this correct? Or don't I understand this calculator? :oops:

ronrem
Posts: 1066
Joined: Sun Jan 16, 2005 2:59 am
Location: Santa Cruz

Post by ronrem » Thu Dec 01, 2005 12:14 am

I actually stumbled across a review of a commercial case that shares the idea I'd posted earlier,using a peltier to cool the incoming air. From what I understand it DID lower internal temps-but less than I'd hoped. Also it uses extra electricity. I don't think it was the best possible peltier setup,nor am I sure that review showed it's best potential.

I'm starting to think more about a Palermo/Sempron 3000 with a Scythe NCU 2005,and a PSU sporting a lowspeed 140mm as the only system fan.
Ought to be real quiet and pretty cheap too. If I could afford an X2 I'd be real tempted to experiment with a Pelt cooled intake.

SoopahMan
Posts: 50
Joined: Fri Aug 19, 2005 6:22 pm
Location: North Hollywood, CA, USA
Contact:

Post by SoopahMan » Thu Jan 18, 2007 9:29 pm

I've read up more on Peltier chips and found the primary reason they generally aren't used: they're inefficient.

If you have this big GPU heating itself up towards disaster and put a Peltier chip on it, what you're implicitly not doing is putting a heatsink and fan on it - and it turns out most Peltier chips are pretty inefficient at capturing heat and turning it into electricity. I've read that it's something like 30% that they capture, whereas a heatsink/fan combo can capture far more when the GPU heat is well above ambient. Peltier chips peak in efficiency in that same scenario, but they're still accomplishing so much less than a heatsink/fan combo in that critical cooling situation that they're hard to justify.

There's one company trying to get Peltiers in on GPUs and CPUs that tried to address the 30% issue by adding an insulating layer to trap more heat in on its way across the Peltier chip (without the layer, most of the heat passes across the chip before it's converted). That does increase efficiency significantly but at the same time introduces a major problem: You're insulating your GPU! Few things could lead to disaster faster.

It sounds like the ideal scenario for Peltiers is in very narrow bands of temperature possibilities, where you can always be sure the part needing cooling will be hotter than ambient, but fairly consistently hotter and not in desperate need of cooling to avoid failure. One example that comes to mind is most hard drives. GPU's and CPU's vary load so widely that they just really don't suit this definition, although they do have a higher tolerance for trapped heat (~125C) than hard drives (~55C).

Anyway - get those Peltiers more efficient at removing heat from the part and they'll make sense for electronics.

Post Reply