I’d hate to piss on everyone’s bonfire, but unless I misunderstand the laws of thermodynamics and physics this won’t work at all.
....
If you really want to cool your CPU to below freezing temps without going phase-change....
Ah, you're missing the point a little. The point isn't to cool my CPU, the point is to have a vaguely interesting project, give me something fun to do, and learn something about Peltiers / water cooling etc. As well as that I was unable to find anyone (on t'internet) who's even tried this....
perhaps my original estimation of -30C was a bit drastic, reading the review on dangerden of the 226watt peltier: "Author: CaptainTrap
I have hit 1 C with Danger Dens 226w Pelt and Maze4 Setup!
"
But you'll just end up with a lot of hot coolant, an even hotter CPU and (if you're really unlucky) a burnt out pelt.
au contraire, I'll end up with several hours of fun, some information about whether this is even possible and (maybe) a published article on the front page
I'm not even going to consider connecting this thing to my CPU until I've had a play about with it in isolation. my first step will be to strap the two heatsinks and the peltier together and see if it works at all.
Man, I'm glad I saw this thread before you invest in buying heatsinks, deicist. Smidge is dead on in his post; it won't work for the reasons he stated. I have messed with peltier cooling and they are fun to mess with, but you can't make an effective chiller out of them. A much better and more efficient way to use them is to use the peltier with a waterblock that is designed to work with the peltier element such as the ones that Swiftech makes or Dangerden used to make like I have. They have a 4-5mm copper cold plate to sandwich the peltier element with and the cold plate is actually what contacts the processor. And for cooling your water loop, you will need a big ass radiator to handle the 220 watts of the pelt plus the heat from the processor. If you use an A64 and overclock the hell out of it, you are probably talking at least 125 watts worth of heat. So you will need to get a rather large radiator in order to be able to ahndle the heat. Now that you have a tec-ready waterblock and heavy duty radiator, you will also have some other things to consider, like condensation on the motherboard. You will have to insulate all around the socket and inside the socket, plus the back side of the motherboard behind the socket so you don't have condensation forming and killing your components. Neoprene is a good choice for insulating around and behind the socket. You will also need to get some dielectric grease to put into the socket where the pins stick into it to stop condensation from forming within the socket itself. For a power supply, figure you are going to spend between $100-200 most probably. When I bought my Meanwell 300-24, it cost me around $140 or so and that was better than 3 years ago. Of course, since you are getting a 12v tec (actually it's probably rated for 14.8v, you could probably adapt a 12v battery charger that can feed a constant 25 amp load or so (tec's aren't 100% efficient by any means).
I'm not trying to scare you off, but sub ambient processor cooling gets expensive any way you go at it. If you still want to do some tec cooling, I'll be glad to help you all I can. I really had fun with it when I was doing it to the old P3 system that prime bought off me and I got some really good clocks off that P3S proc for it's day. And it was cool as hell to see the machine folding and the cpu temp reading 0 C and idle temps around -15 C, and that was with a 120 watt tec.:D
BTW TheSmJ, air cooling a high watt peltier element has been done successfully (if loud). Swiftech used to make an air cooled pelt based on the MCX462 heatsink coupled with a high cfm fan.
BTW TheSmJ, air cooling a high watt peltier element has been done successfully (if loud). Swiftech used to make an air cooled pelt based on the MCX462 heatsink coupled with a high cfm fan.
I'm not suprised, but I gathered that he was looking for something quiet.
yeah, I've done some more reading after TheSMJ's post and I'm starting to realise it won't work with that peltier... but a smaller peltier should be able to be cooled by a normal heatsink right? and the cold plate will still chill the water, just not as effectively.... I'm going to press on with this regardless
Well, good luck on it and for now I don't think you will have to worry about using antifreeze or anything else for your cooling medium; water whould do fine for you initial experiments.
Use your same pelt, just start at lower voltages and work your way up intil you reach your heat disipation limits. These things are almost linear with voltage.
Use a water alcohol mixture if you are getting freezing. The low FP and visc more than make up for the low heat capacity.
I would recommend against doing this. Will it work? Sure. But you are so much better off just buying an old mini-fridge for $20 and gutting it. Peltiers are extremely inefficient, and their only practical use is to provide extremely low temperatues on a small contact point at the cost of about 65% cooling efficiency. For cooling a relatively large volume of liquid, there are far more efficient (And cheaper, if you're a DIY kind of person) methods of doing it.
I understand you already own the peltier, but for the cost of the heatsinks and other materials, you could just buy a used minifridge. It will dissipate the heat from the liquid more efficiently, have more of a coolness factor, and you'll still have the peltier... Which would be better used on the chip itself than for the liquid cooling system.
That reminds me... out refrigerator has a water dispenser on the insde upper roof of the refrigerator half. It takes in regular tap water and cools it off pretty well. It takes a while to cool it (probably overnight), but it works (and I would imagine once a gallon or so is cooled, it would stay cool pretty easily) - would something like that work? It's probably a little too modern to get from a junkyard or something, but just a thought.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited March 2006
Edit: Done some more reading on this, apparently very pure water can be cooled to below 0 degrees without freezing in certain circumstances, but it will still freeze eventually.
I stand corrected...and come to think of it, that is indeed what I learned. Conductivity of water. Pure water is completely non-conductive, it needs particles or dissolved foreign matter in it. Sea water conducts much better than fresh water because of the dissolved sodium chloride and other salts.
Well I'm off work ill at the moment with a cold bug type thing (not the AMD one ) but I should be back tomorrow and hopefully my peltier & heatsink will be sat on my desk waiting for me so I should be able to make a start. I'm planning to write it up piece by piece in this thread as I do each bit, then compile it all at the end into a big write up.
Okay, the peltier and the 'hot heatsink' arrived today so I thought I'd have a little play about with it. I don't have a decent 12v power supply yet but what I do have is a standard ATX power supply. Now this is quite a poor quality one, it only supplies 13A on the 12v rail, but on reading the specs it does supply up to 25A on the 5v rail....which is more than enough for a quick test. Only problem is the peltier had no markings to indicate which side was which (hot / cold) so I came up with a very scientific test. Hold the peltier in one hand, and use the other hand to switch on the PSU. When the peltier gets hot / cold switch off the PSU. This test went swimmingly, one side got warm and the other side got cold, exactly as advertised.
Here's some pictures for you: The peliter when it arrived, undergoing my 5v test, and marked up (x marks hot side)
Now I need to figure out a way to mount the 'hot' and 'cold' heatsinks on the peltier. My idea is to drill mounting holes in the 'cold' heatsink (which is just a finned block of aluminium) and use them to mount the 'hot' heatsink on it, with the peltier sandwiched tightly in-between. Only problem is I've never drilled into metal before, so I'm not sure what I need. I'd like to make properly threaded holes so I can just use the screws already on the 'hot' heatsink rather than mess around with bolts and whatnot. Can anyone shed light on the easiest way to go about doing this?
More pics: 'Hot' heatsink in box, 'Hot' heatsink assembled with fan (note the mounting screws) 'cold' heatsink.
OKay, I got bored again, so I decided to try just sticking the peltier to the hot heatsink with the thermal paste that's on there and holding the cold heatsink on it. There's a definate change in the temperature of the cold heatsink. Now that may sound like a 'well duh' statement, after all that is what peltier's are supposed to do. But considering that in an air conditioned room, running the peltier at 5v and with poor thermal contact between the cold heatsink and the peltier you can still notice that the cold heatsink is getting cold I'd say it's quite successful
At some point I think I'm going to have to look at the interface between the peltier and the 'hot' heatsink. The problem is that the heatsink is designed to sit on a processor and the peltier is slightly bigger than a processor. not sure how to get round this but I'm sure I'll think of something.
more pics: Peltier sat on top of the hot heatsink, and just for reference the power supply I'm using.
Right, I had another 10 minute play around and here's what I learned.
First I found another 'cold' heatinsink. It's off an old CPU and it's exactly the right size for the peltier to sit on. So I attached that to the setup (no thermal paste though) and left it running a while. The coldsink got quite...well, cold. Then I decided to have a little test. I put some water in the bottom of a container, placed my setup in the water and left it running. The water got noticabley cooler. I don't have a thermometer though, so it could be subjective. Mental note: get a thermometer before testing a cooling solution, it might come in handy. After all this sensible testing I started to tire of using 5v, so I decided to jump up to 12v. remember my power supply is rated for 13A at that voltage, the peltier is rumoured to need 20A or so. Should be fun.... after running for 5 minutes the PSU hasn't melted so I guess that was a successful test
so, it's nearly home time now, I'll come back to this on monday with (hopefully) a thermometer and do some proper tests,but for now the cold sink is getting cold and the heatsink seems to be dealing with the heat coming off the hot side of the peltier (for 5 minutes at least) so I reckon it's looking good
more pics: my new 'coldsink', the 'Frankencooler' complete setup, and my 5v water test.
First off, I started to notice that while the middle of the peltier was getting cold, the outside edges were staying decidedly warm. After looking at the heatsink I'm using on the hotside I noticed that the contact area is very small, which was hindering the cooling of the hotside, that heat was then transferring to the cold side. Not good. Luckily I had another option for my 'hot' heatsink, one I rescued from a Dell desktop. It's perfect, nice large contact area and several heatpipes. After setting this up as my 'hot' heatsink the cold side now gets (and stays) cold. Also, I'm now using the large akasa heatsink for my cold side. The small contact area doesn't seem to be as much of a problem on the cold side, the whole heatsink gets nice and cold very quickly. I've had it running for about half an hour or so now and the cold side heatsink is really cold to the touch. I've also found a reservoir (an old CD spindle / cover thing) for my coolant, which the 'cold' heatsink fits in perfectly. Next I'm going to get a digital thermometer with a probe and test the setup with the peltier on and off to see if it makes any difference to the temperature of the coolant.
more pics: Poor contact area on the akasa heatsink, the new dell heatsink, system setup next to reservoir.
well I finally tried to cool some liquid down with it....
I went and bought a digital thermometer at lunch, it's one that's supposed to be used in aquariums, has a sensor in the main body for the ambient temperature, and another on the end of a cable for the liquid temperature. I decided to have 2 bodies of liquid, a control one which would just stand at room temperature and my test one, which will house the cooler. So I filled up my two spindle cover reservoir things with water, setup my cooling in one and took some starting temps. Here's where I realised my first mistake. I'd filled both reservoirs from a cold tap, the test one first, then the control one. So the control one was a good few degrees cooler than the test one. What I should have done is filled them both with warm water.....but anyway, starting temps (all temps in centigrade), ambient temp was 25 degrees
Test: 14 degrees
control: 12.5 degrees
then an hour later I checked again.
test: 15 degrees
control: 16 degrees
then half an hour after that
test: 15 degrees
control: 18.5 degrees
The test one seems to be holding steady at 15 degrees while the control one is still rising... I guess it'll stop at 25 degrees (the ambient temp). So my cooling system can hold a small reservoir of water at 10 degrees below ambient. The next test I guess is to apply a heat source to the system (a processor & waterblock), circulate the water and see what happens.
Looking at the setup for my first experiment I noticed that the coldest part of the coldsink wasn't actually making contact with the coolant. See, the coldsink has a large metal base which is connected by a large heatpipe to the finned part of the 'sink. In my first experiment this large base was above the level of the water, which was good because it kept the water away from the peltier. But I was looking at the setup and thought that it'd be cool if I had some sort of lid on the coolant to keep it away from the peltier whilst still allowing the coldsink to make contact with it. So I needed a flat round piece of plastic the same diameter as the spindle case with a round hole in it the same size as the coldsink contact point. Luckily, the base (top?) of the second spindle case fitted this bill perfectly after a bit of modification.... So on to my second experiment. The setup now looks like something you wouldn't mind sat near your pc, I just need to seal the lid in there with some hot glue or something. After some testing (with warm water this time) I found my setup is able to keep the coolant at 12 degrees below ambient with the fan on it's slowest setting. It's basically inaudible now. I've also found an old socket A setup (a 700MHz athlon) which I can use for testing. So, does anyone want to donate a used socket A waterblock and a pump for further testing?
Wow, I can't believe I missed this thread.. awesome work deicist! Definitely write up something for the front page! Even if this turns out to be a failure, it is very, VERY interesting material!
hmm, it's not so much that it's a failure, just that I need some watercooling stuff to proceed and I haven't got any free cash to buy any with at the moment so the project has sort of stalled... at the same time I don't want to write up what I have so far because it'll kind of put a lid on the project.
hmm, it's not so much that it's a failure, just that I need some watercooling stuff to proceed and I haven't got any free cash to buy any with at the moment so the project has sort of stalled... at the same time I don't want to write up what I have so far because it'll kind of put a lid on the project.
Oh no no, I didn't mean that I thought it was a failure, but I was just saying that regardless of the final product (once you are finished setting it up and tweaking it of course) it'll make awesome reading
Comments
Ah, you're missing the point a little. The point isn't to cool my CPU, the point is to have a vaguely interesting project, give me something fun to do, and learn something about Peltiers / water cooling etc. As well as that I was unable to find anyone (on t'internet) who's even tried this....
I have hit 1 C with Danger Dens 226w Pelt and Maze4 Setup!
"
and thats on water cooling the hot side of the peltier.
http://www.dangerdenstore.com/home.php?cat=30
au contraire, I'll end up with several hours of fun, some information about whether this is even possible and (maybe) a published article on the front page
I'm not even going to consider connecting this thing to my CPU until I've had a play about with it in isolation. my first step will be to strap the two heatsinks and the peltier together and see if it works at all.
I'm not trying to scare you off, but sub ambient processor cooling gets expensive any way you go at it. If you still want to do some tec cooling, I'll be glad to help you all I can. I really had fun with it when I was doing it to the old P3 system that prime bought off me and I got some really good clocks off that P3S proc for it's day. And it was cool as hell to see the machine folding and the cpu temp reading 0 C and idle temps around -15 C, and that was with a 120 watt tec.:D
BTW TheSmJ, air cooling a high watt peltier element has been done successfully (if loud). Swiftech used to make an air cooled pelt based on the MCX462 heatsink coupled with a high cfm fan.
I'm not suprised, but I gathered that he was looking for something quiet.
Use a water alcohol mixture if you are getting freezing. The low FP and visc more than make up for the low heat capacity.
I understand you already own the peltier, but for the cost of the heatsinks and other materials, you could just buy a used minifridge. It will dissipate the heat from the liquid more efficiently, have more of a coolness factor, and you'll still have the peltier... Which would be better used on the chip itself than for the liquid cooling system.
Here's some pictures for you: The peliter when it arrived, undergoing my 5v test, and marked up (x marks hot side)
More pics: 'Hot' heatsink in box, 'Hot' heatsink assembled with fan (note the mounting screws) 'cold' heatsink.
At some point I think I'm going to have to look at the interface between the peltier and the 'hot' heatsink. The problem is that the heatsink is designed to sit on a processor and the peltier is slightly bigger than a processor. not sure how to get round this but I'm sure I'll think of something.
more pics: Peltier sat on top of the hot heatsink, and just for reference the power supply I'm using.
First I found another 'cold' heatinsink. It's off an old CPU and it's exactly the right size for the peltier to sit on. So I attached that to the setup (no thermal paste though) and left it running a while. The coldsink got quite...well, cold. Then I decided to have a little test. I put some water in the bottom of a container, placed my setup in the water and left it running. The water got noticabley cooler. I don't have a thermometer though, so it could be subjective. Mental note: get a thermometer before testing a cooling solution, it might come in handy. After all this sensible testing I started to tire of using 5v, so I decided to jump up to 12v. remember my power supply is rated for 13A at that voltage, the peltier is rumoured to need 20A or so. Should be fun.... after running for 5 minutes the PSU hasn't melted so I guess that was a successful test
so, it's nearly home time now, I'll come back to this on monday with (hopefully) a thermometer and do some proper tests,but for now the cold sink is getting cold and the heatsink seems to be dealing with the heat coming off the hot side of the peltier (for 5 minutes at least) so I reckon it's looking good
more pics: my new 'coldsink', the 'Frankencooler' complete setup, and my 5v water test.
First off, I started to notice that while the middle of the peltier was getting cold, the outside edges were staying decidedly warm. After looking at the heatsink I'm using on the hotside I noticed that the contact area is very small, which was hindering the cooling of the hotside, that heat was then transferring to the cold side. Not good. Luckily I had another option for my 'hot' heatsink, one I rescued from a Dell desktop. It's perfect, nice large contact area and several heatpipes. After setting this up as my 'hot' heatsink the cold side now gets (and stays) cold. Also, I'm now using the large akasa heatsink for my cold side. The small contact area doesn't seem to be as much of a problem on the cold side, the whole heatsink gets nice and cold very quickly. I've had it running for about half an hour or so now and the cold side heatsink is really cold to the touch. I've also found a reservoir (an old CD spindle / cover thing) for my coolant, which the 'cold' heatsink fits in perfectly. Next I'm going to get a digital thermometer with a probe and test the setup with the peltier on and off to see if it makes any difference to the temperature of the coolant.
more pics: Poor contact area on the akasa heatsink, the new dell heatsink, system setup next to reservoir.
Looking good so far, deicist.
I went and bought a digital thermometer at lunch, it's one that's supposed to be used in aquariums, has a sensor in the main body for the ambient temperature, and another on the end of a cable for the liquid temperature. I decided to have 2 bodies of liquid, a control one which would just stand at room temperature and my test one, which will house the cooler. So I filled up my two spindle cover reservoir things with water, setup my cooling in one and took some starting temps. Here's where I realised my first mistake. I'd filled both reservoirs from a cold tap, the test one first, then the control one. So the control one was a good few degrees cooler than the test one. What I should have done is filled them both with warm water.....but anyway, starting temps (all temps in centigrade), ambient temp was 25 degrees
Test: 14 degrees
control: 12.5 degrees
then an hour later I checked again.
test: 15 degrees
control: 16 degrees
then half an hour after that
test: 15 degrees
control: 18.5 degrees
The test one seems to be holding steady at 15 degrees while the control one is still rising... I guess it'll stop at 25 degrees (the ambient temp). So my cooling system can hold a small reservoir of water at 10 degrees below ambient. The next test I guess is to apply a heat source to the system (a processor & waterblock), circulate the water and see what happens.
Looking at the setup for my first experiment I noticed that the coldest part of the coldsink wasn't actually making contact with the coolant. See, the coldsink has a large metal base which is connected by a large heatpipe to the finned part of the 'sink. In my first experiment this large base was above the level of the water, which was good because it kept the water away from the peltier. But I was looking at the setup and thought that it'd be cool if I had some sort of lid on the coolant to keep it away from the peltier whilst still allowing the coldsink to make contact with it. So I needed a flat round piece of plastic the same diameter as the spindle case with a round hole in it the same size as the coldsink contact point. Luckily, the base (top?) of the second spindle case fitted this bill perfectly after a bit of modification.... So on to my second experiment. The setup now looks like something you wouldn't mind sat near your pc, I just need to seal the lid in there with some hot glue or something. After some testing (with warm water this time) I found my setup is able to keep the coolant at 12 degrees below ambient with the fan on it's slowest setting. It's basically inaudible now. I've also found an old socket A setup (a 700MHz athlon) which I can use for testing. So, does anyone want to donate a used socket A waterblock and a pump for further testing?
Oh no no, I didn't mean that I thought it was a failure, but I was just saying that regardless of the final product (once you are finished setting it up and tweaking it of course) it'll make awesome reading