This is an idea that’s been lurking in the back of my mind for some time: Why not use a peltier as the “radiator” to cut down the bulk and simplify a water cooling rig? So, I finally scrounged together some parts to “concept test” the idea.
The approach is relatively simple: Cut a hole in a water container, add a cold plate inside, mount a peltier on the cold plate and use a fan/heatsink to cool the peltier. The internal cold plate cools the water, the pump circulates cool water to the water block, and the external heatsink cools the peltier.
Now this does beg the whole question of why do this rather than mount the peltier directly on the CPU? If you want super-cooling, this is not the way to go. If you want ambient +/- 5 C, this could be a way to go. Easiest way to found out (at least for me) is to build one and bench test it.
As the pics above show, the first thing I did was cut a hole in a water container and mount a cold plate inside the container. I scrounged a heatsink and did a little cutting to get it to fit, then used silicon adhesive to make it watertight. I put silicon on the screws also and did not have any leaking problems.
Shown below is the prototype I tested. I used a small submersible pump (the noisy Senfu) with a 72 watt peltier. The peltier heatsink is something I found in my local scrap metal yard – massive but perhaps not the most efficient peltier cooler (more surface area needed). However, for the first run, a working prototype.
The first thing I did was run it on my test rig to see if in fact if could maintain its cooling under load, and it did. I then decided to do a live test with a PIII 600E running at 800 MHz with Prime95. I measured the resulting temps at four locations after temps stabilized, about one hour:
CPU: 32 C
System: 23 C
Water: 24.4 C
Cu Block: 23.2 C
Ambient: 23.2 C
What I found interesting was the time it took to get to a stable temperature – about one hour. This was similar to what I found in using just a pot of water – a long time to reach stable temps; slow, steady climb in temps, them stability. It’s as if the peltier in effect “multiplied” the water volume. It took about as long to get back to ambient after terminating Prime95.
Under no load, the temp declined about 7 C under ambient, taking its time to get there.
An unqualified success? Not really – 32 C under stress is higher than what I could get with a Alpha PEP66. However, this is the first crack at it. With some tuning and a better peltier cooler, I think it’s possible to get this thing working to where it could keep a CPU at ambient and maybe 5 C below ambient.
I’m sure some enterprising souls out there will take up the challenge and build a peltier radiator with better results.
SUMMARY: 50′ of 1/4″ copper tubing and a house fan can substitute as a radiator for CPU Cooling; cost: $11.97 for the tubing.
Another visit to Home Depot and I decide to try out an alternative to the 3/8″ tubing – 1/4″ copper tubing, the size usually used to run a water line to your fridge’s ice-maker. I look around and pick up 50′ for $11.97. I then proceed to meticulously coil it on the front of a house fan (pic below)
Turns out this is not as easy as you might think – 50 feet is a lot of tubing and you have to get it into a preliminary shape before attaching it to the grill; not difficult, just wriggly. Fifty feet looks about right for the average house fan. The only thing I did not like was that I had to use a collar to fit the 3/8″ tubing I use for the rest of the system – I know the water flow will be less than for a 3/8″ system and performance may suffer because of it.
After I get it all wrapped around and secured, I fire everything up (water block on double peltiers – about 100 watts to cool). I wait a while for temps to settle in, and as the pic shows I stabilize at -19.3 C. This is slightly better than with 10″ of 3/8′ tubing (-19.0 C) but really is too close to call – a draw.
All told, it looks like the increase in surface area for the 1/4″ tubing more than compensated for the decrease in water flow rate. I’ve had some emails from folks thinking about shaping tubing into a cooling tower – like the kind you see at power plants. Interesting principle and what’s fun about this tubing is you can bend it very easily into all kinds of shapes to test out alternative designs.
Hmm…cooling tower?
SUMMARY: A large enough water container can act as a radiator – just a question of how large.
Max from Italy sent in a note about how he uses a large (11 liters) water container which acts as a radiator. When you think about it, if you had the Atlantic Ocean as a water source, it would take a very long time to raise its temp if you used it to cool your CPU. Someplace between that and maybe a couple of gallons and you can effectively cool your CPU.
The reason is that it takes a long time to heat up a couple of gallons of water with a small heat source, and if you are only heating it up intermittently, even longer. Also, once you go over ambient temps, the heat is lost to the ambient air, so cooling is taking place.
To test this out, I went into the kitchen and got the largest pot I could find – holds about 1 1/2 gallons (see pic below). I then placed a submersible pump in the pot and used my portable test rig to see what would happen. This consists of 2 peltiers which throw off about 100 watts. I than ran it until the water temp stabilized – it took 4 hours!
Shown below are the temps I achieved with just a pump and 1 1/2 gallons of water. Water temp peaked at about 98 F and stayed there. This was with an ambient temp of 72 F. Considering that I ran full tilt four four hours, if you are not running flat out it will take longer to hit peak temps.
A number of folks are using picnic coolers and putting ice packs in the water – this is another approach and can get CPU temp below ambient. The point here is that a large water source in a container that is designed to shed heat can be an effective alternative, although clearly space is a big consideration.
Now, if you live by a lake…
This is an idea that’s been lurking in the back of my mind for some time: Why not use a peltier as the “radiator” to cut down the bulk and simplify a water cooling rig? So, I finally scrounged together some parts to “concept test” the idea.
The approach is relatively simple: Cut a hole in a water container, add a cold plate inside, mount a peltier on the cold plate and use a fan/heatsink to cool the peltier. The internal cold plate cools the water, the pump circulates cool water to the water block, and the external heatsink cools the peltier.
Now this does beg the whole question of why do this rather than mount the peltier directly on the CPU? If you want super-cooling, this is not the way to go. If you want ambient +/- 5 C, this could be a way to go. Easiest way to found out (at least for me) is to build one and bench test it.
As the pics above show, the first thing I did was cut a hole in a water container and mount a cold plate inside the container. I scrounged a heatsink and did a little cutting to get it to fit, then used silicon adhesive to make it watertight. I put silicon on the screws also and did not have any leaking problems.
Shown below is the prototype I tested. I used a small submersible pump (the noisy Senfu) with a 72 watt peltier. The peltier heatsink is something I found in my local scrap metal yard – massive but perhaps not the most efficient peltier cooler (more surface area needed). However, for the first run, a working prototype.
The first thing I did was run it on my test rig to see if in fact if could maintain its cooling under load, and it did. I then decided to do a live test with a PIII 600E running at 800 MHz with Prime95. I measured the resulting temps at four locations after temps stabilized, about one hour:
CPU: 32 C
System: 23 C
Water: 24.4 C
Cu Block: 23.2 C
Ambient: 23.2 C
What I found interesting was the time it took to get to a stable temperature – about one hour. This was similar to what I found in using just a pot of water – a long time to reach stable temps; slow, steady climb in temps, them stability. It’s as if the peltier in effect “multiplied” the water volume. It took about as long to get back to ambient after terminating Prime95.
Under no load, the temp declined about 7 C under ambient, taking its time to get there.
An unqualified success? Not really – 32 C under stress is higher than what I could get with a Alpha PEP66. However, this is the first crack at it. With some tuning and a better peltier cooler, I think it’s possible to get this thing working to where it could keep a CPU at ambient and maybe 5 C below ambient.
I’m sure some enterprising souls out there will take up the challenge and build a peltier radiator with better results.
Just read your article about the peltier reservoir. I had that same train of thought about 3 weeks ago which led me to build one. I used a couple of Alpha P3125’s I had lying around. I used a 6x6x4 pvc junction box, cut it out to fit one of the Alpha’s, leaving the entire face exposed for multiple peltier application.
I was using two commercial grade 60W peltiers that I hacked out of a Thermoelectric cooler to cool the sink in my reservoir.
Well, my luck was basically same as yours, except my starting temp averaged 41-42 F. Man, it was great – out – of – the – gate !@!, but not much for stamina. Water temp would level off at about 78-80 F (large swing in temp due to 52watt peltier on CPU).
I did find something you may want to try with yours; keep the water level in the reservoir about 1/4″ below the top of the heatsink & route your return water to splash across the top of the exposed
cold sink fins. As you may have experienced, the cold sink never stays “fully charged” because the heated water neutralizes it.
But if you keep the top portion of the sink fins exposed to air….they will get colder….thus allowing a “blast cool” (so to speak) when the return water runs onto it. This little mod was good for approx 2 F for me (again, my water was being very well heated by the peltier on my CPU).
You may realize more out of it, if you’re just using the water
block on your CPU (from the temps you posted, I would guess that you
are). Staying along the subject of the return water, something I thought of trying but never did; putting a diffuser screen on the end of the return hose to aerate the water as it exits & onto the cold sink??
Would you believe the second Alpha P3125 that I had cooling off the
hot side did an excellent job of it?? I thought, at one point, I
needed a bigger heatsink & fan on the hot side so I swapped to a 4×6
sink & 120mm/105cfm fan – louder than hell & did a worse job!!
Just
thought I’d drop you a line on something you could try out for yourself. I would like commend you on the piece you did about using the metal pot for reservoir. I’m currently using a black iron stone pot (6qt) and have a 5×15″ transmission cooler laying across the top of the pot & my return water running across the face of the cooler (causes the water to trickle into the pot).
My water temp is constantly 75 F (+/- 2) – and that’s with the same 52W peltier on my CPU!
Finally got my water cooler all figured out and, after seeing your compact radiator/peltier article, I just had to send it so you could see my compact radiator setup.
Being the blast furnace that it is, I wanted to water cool my K6III 400@472, so I got online and checked some prices. WOW!! Are those things expensive! So I built my own.
Since I was in the forums capping on the drilled type blocks, I figured I’d better do an internally finned block. After 2 prototypes this is what I came up with:
The Lexan was to cut down on condensation because I was running ice water and a peltier trying to get some more MHz. I got it to post at 660 but couldn’t get the video stable over 500, so I went back to straight water. I will solder a copper lid on it shortly.
I wanted a remote radiator to move the heat from the AMD space heater off the desktop but it had to be small to go where I wanted it to, so I came up with this:
I experimented with different lengths of tubing (10-50 feet) and found 30 feet of tubing did the best job. For my late night frags, I point it toward the window and it gets close to the inside ambient temp when I’m playing AHL!!
I am sold on water cooling and I am making one for both my server (Cel366a @550) and my gaming rig (Athlon 650@715). The whole mess cost me less than 75 bucks at Home Depot and copper from Online Metals. I predict the next one will be about a 6 hour project now that we have a working design to follow.
FINALLY!! A way to deal with the California heat when overclocked 🙂
NOTE: Surlyjoe has started a website – it’s new so don’t look for a lot of material yet, but drop by HERE.
Chip Eckert runs Overclock-Watercool and has some neat water cooling stuff – I have the Coil Radiator and will test it out shortly – Joe.
Are you familiar with “swamp coolers”? This is an inexpensive evaporative Air conditioner folks use in drier climates, and in motor homes?
Being a long time camper & traveler, I am….
Anyway…I cut the toe off an old sock to make a tube. I put the tube over one of my Silver Towers (pic below). I adjusted the legs on the fan to give about an inch clearance below it, and sat the whole thing in about a half inch of water in a shallow dish.
The fan has 1-1/2″ adjustable legs on it and is the regular 92mm fan
that comes on the “Silver Tower”. The sock is a special type designed
specifically for overclocking that I’m going to retail for $150 (kidding). I DID make sure it was part cotton so it would wick well (I can’t wait to invest in the sock market!!!).
My 366 @550 – 2.1v CPUs are now happily churning away at SETI and Prime95 without peltiers or any other active cooling at ONE degree above room temp.
It may be ugly, but 7 degrees is 7 degrees…especially for free.
Be the first to comment