Is my configuration flawed?

[immudium]

I am completing setup of my first peltier based water cooling system and some of my initial results are not what I expected and I was hoping some of you all-powerful water cooling/peltier Jedi could set me back on the path to the light side of the force.

I've done most of the condensation prevention around the socket and I am using two 120 W peltiers @ 20 V each with an Athlon XP 1600 (1.4Ghz). My CPU temps generally stabalize at about 23 deg C. My system temp is at 32 deg C. I am not overclocking at all yet so all cpu voltages are at default. The board I am using is a Soyo Dragon Plus which should be using the thermistor built into the Athlon XP CPU and not an external thermistor so it seems like the temps should be fairly accurate.

Anyway, I was under the impression that my temps should be much, much lower that 23 degrees. Do you think it could be my water loop setup? I have the line going from the pump to the water block and then from the water block to the radiator and then back to the pump. I noticed that the Eheim 1250 I am using gets fairly warm after its been on a while so could this be keeping my water block too hot? Should I have the water coming from the radiator going straight into the water block or does the heat from the pump have a neglible effect on the heat of the water loop the way I have it? Do you think it could be a question of insulating the cold plate better? There are parts of the cold plate still exposed, but I didn't think that would matter too much (certainly not 23 degrees worth), but I could be wrong. I also have two 7200 rpm 120 mm fans blowing through the radiator (one for intake, one for exhaust), so I don't think it is lack of air movement through the radiator. Anyway, I've listed system components below for thoroghness.

Thanks for any feedback,
Immudium

My Water cooling gear is as follows:
Maze 2-2
2 120W Peltiers (ordered directly from TE Dist) @ 24V each
Super Cube radiator + two 7200 rpm fans at intake and exhaust
Eheim 1250 pump
1/2" fittings
2 X 230W power supplies in series for peltiers

Computer components:
Athlon XP 1600 (1.4GHz) @ 1.75V
Soyo Dragon+ MB
512 MB Crucial DDR
430 W Antec PS
32 MB Creative Labs GeForce vid card

[Brians256]

I'll reply from a purely theoretical standpoint, since I freely admit that I've never used peltiers.

First thing to check before you ordered the peltiers would be to run a simulation of your heatload on the two peltiers. The manufacturers often provide a program or web app to estimate what wattage you need for a certain target temp. I don't think your current temps are bad for the setup that you have.

But.... you might have insufficient clamping pressure. Your voltage seems good, you have a cold plate and water cooling... unless the peltiers aren't getting enough cooling, you may be at the optimum for the equipement you have.

[immudium]

Thanks, definitely some good ideas to try. Maybe a better understandind of the theoretical aspects would help me know better what to expect. I've seen a few simulators around, I'll give one of those a try

[Brad]

you should be around the -10C mark with that setup.

I'd have to say it is going to be mounting, or your thermal interface, something like that.

[Brians256]

Ahhh.... good! Brad actually has experience using peltiers. Me, I'm just flappin' my gums and regurgitating stuff that I've read. (Hmmmm... that extended metaphor gives me a really bad image when I imagine it.)

[decodeddiesel]

Deffiniatly a number of things to try. First of all the 4 screws which provide independant clamping pressure to the TEC should be absolutly as tight as you can get them. You are aiming for app. 300 psi on the TECs.

Second, what is the exact specs of your PSU? You could be putting too much on it and it's not giving the TECs the juice they need.

3rd, you said you had most of your insulation done, you MUST have proper insulation. See the articles on http://www.octools.com for a good tutorial on this. It won't affect your temps much, but condensation can corrode and fry your system, it must not be taken lightly.

Are you using AS on all thermal interfaces? (ie both sides of the tecs and the CPU?) Have you tried tightening the mounting nuts so the springs are nearly completly compressed? just having a little pressure on your CPU won't provide good results.

Also, the temps are NOT comming from the on-die XP diode, but rather from a straight socket probe which will almost always give you phony temps in a WC rig and especially in a TEC rig. If you want to use the on-die diode see this article http://www.voidyourwarranty.net/revi...de/index.php3, or you need to wait untill they support the diode (deffiniatly won't see this in KT266A boards) and buy a new board.

Remember TECs are not "plug and play" no brainer rigs they require a lot of time, effort, money (wait till you see yuour electric bill), and research. Not to mention they have a tendancy to fail at the worst possible times, and can cause a lot of damage. Good Luck to you I hope everything works out well.

[immudium]

Great info. Thanks guys. You're right, I'm finding out that the more I learn about peltier cooling, the less I seem to know. It's taken me more than two months to get to the point I'm at now, but it's been fun and interesting and I plan to keep working on it for as long as it takes to get it right.

Anyway, I have only been using standard silicone based heatsink compound between the peltiers, so I will replace it with some Artic Silver. Also I was a bit scared to put too much pressure on the pelts, but it sounds like the more the better. I am a little shaky on the two 230W PSUs that I am using for the peltiers. They are as old as the hills and I had them just lying around and decided to see how well they would work connecting them in series as per the instructions on this very web site. I tried to follow those instructions to the letter, so I've got the 2ohm 25 W restors and everything hooked up. The only difference was in the wattage of the PSUs themselves, so I will look into getting some newer 250 W or higher units.

It is interesting that no KT266A boards will include on-die support. My Dragon Plus definitely does not have any sort of external thermistor in the CPU cavity (at least not that I could recognize) unlike my Epox 8KHA+ which is definitely obvious. If the Dragon thermistor is external then it is now covered by my conformal coating b/c it definitely is no taller than any of the other components in that same cavity. And they are all pretty flat.

Anyway, thanks again for the suggestions...

[resago]

this might sound stupid, but check and make sure you don't have one of the pelts flipped the wrong way. IE, canceling each other out. this would create the effect you are experiencing.

wire order should be red-black-red-black or black-red-black-red, NOT red-black-black-red or black-red-red-black

also the waterblock should be warm to the touch if both pelts are oriented correctly. my single 156watter keeps the waterblock rather warm to the touch.

also, pump->radiator->block->pump is generally considered best.

and finaly, make sure that whatever method you are using to insulate isn't preventing good contact on the cold plate to the core.

[Brad]

go grab a voltmeter and see what sort of voltage those psu's are putting out. it might be a lot less than you expect.

as2 vs normal stuff won't be 20C differance. it might be 2 or 3, but nothing much more.

peltier pressure is very important. Basically, as much as you can possibly do without breaking your arm.

[decodeddiesel]

Agreed Brad, I think he deffiniatly needs to have a looksie at those PSUs with a Multimeter. You can find one cheap at Rat Shack that will do what you need. At least this will give you a good idea as to what you're feeding those power hungry TECs.

Yes, AS won't make much of a difference form Rat Shack paste, I was more asking if you had a thermal interface. not trying to insult your intelligance but i have heard of people using nothing in the way of paste with the TEC interface.

You may want to try running the system outside of your box with no CPU. Fire up your cooling system and the TECs, see how long it takes frost to form on the cold plate. Should take maybe 20-30 seconds for frost to form. If there's no frost on the cold plate you know there's a problem with your TECs. If you've got frost you know they are both doing what they're supposed to be doing.

[immudium]

I took some of your suggestions into consideration last night and I think I'm making progress. I tightened the bolts on my maze 2 and much to my surprise found that they had a lot more love to give, i.e. they tightened up quite a bit more and I thought I already had them pretty tight. I then put a neoprene shim around the outside edges of the cold plate and also around the socket itself. I also replaced my generic heatsink paste with some artic silver between the cpu and cold plate. I haven't replaced the paste between the peltiers yet, but it is on my to do list.

Results after doing these simple tweaks were 13 degrees CPU temp and 35 degrees system temp. So the system temp stayed pretty much the same while the cpu temp dropped a good 10 degrees C. I did notice that the cpu temp would hit a low of about 8 degrees C and then slowly climb up to 13 degrees where it would stabalize which suggests to me either the pump is adding to the heat after its been on a while or the radiator is being overloaded. I'll have to put my thinking cap on for this one. But either way, this weekend, I plan to rework my plumbing. When I first layed out the water flow it seemed logical to me to have the pump feeding directly into the water block, but the more I've played around with it the more it seems like a bad idea to me.

Anyway, I did check the voltage coming out of the power supplies with a voltmeter. It fluctuates between 20.45 and 20.55 volts. Without the peltiers connected the voltage is about 23.85, so is a 3-4 V drop a bad thing?

I will give the simple frost test a try tonight and check my peltier configuration. I am really going to give myself a good kicking if I got it wrong b/c I tried to be extremely careful about orienting them the right way.

Thanks for the help!

[Butcher]

The rise will be water temps finding equilibrium (what are the water temps btw?) You might knock a few degrees off with a better rad setup or not depends (btw 7200 rpm fans sound excessive, mine is single and can push 235 cfm at 3500 rpm )
As for temps, I don't think -10C is realistic (wtf you get that number brad?). I tapped your numbers into a handy peltier calc and came up with around -5 to 13C as realistic temps (note this takes into account peltier voltage and such as well as cooling system efficiency). -5C is with a damn good setup on the other side. This is assuming those PSUs are actually putting out 20V.
As for rat shack vs artic silver, AS2 is know to have major problems with foiling at low temps, you may do better with generic white goop.

[Brad]

butch, he said under load the psu's are 23.85v, so you might want to re-adjust that temp a little bit.

the 10C is heaps man, many people will pull everything apart for 1C, but with just very small mods you got 10C. Try to tighten up the cold plate some more.

as for rerouting your system, yes, you might get 5C or something if you make sure there are no kinks, or 90 degree bends, and the radiator feeds straight into the block.

[immudium]

Brad, there's no danger in me snapping the Maze2 bolts by pulling too hard if I'm just using a simple Allen wrench is there? I could probably torque them a bit more until my fingers start bleeding, but I've destroyed lesser bolts without too much effort, so I'm probably overly paranoid, I guess.

Also, just to clarify, by my measurements, the PSUs are putting out about 23 V with just the load resistors on them and not the pelts. With the peltiers attached (is that what you mean by under load?) the voltage drops 3-4 volts to about 20 V.

[Butcher]

TECs are rec. for 200psi, on dual 40x40's that's about 1000lb force on them, no danger of too much there, as for the maze, if you strip the threads from the block that's too much force.

[Miss_Man]

How is your radiator handling the heat? Maybe you can add in a water temp monitor before the block so that you know how well the heat is taken care of by the radiator. Once you up the voltage and speed, I'm doubtful that the radiator can keep up. Also, are your peltiers fed enough amperage, you need around 13A for them at 20V.

[Butcher]

If they're at 20V, they're getting 13A, you can't have not enough amps for a given voltage - it just drops to less voltage if the PSU can't handle the current draw - hence the 23.8V to 20V drop.

[Miss_Man]

So thats how it works? Now I get it.

[Butcher]

Amps and volts are directly dependent. If the psu can't supply enough amps it doesn't and the voltage drops proportionally. Also the psu tends to overheat

[Brians256]

Butcher has alluded to a basic electronics principle. With DC current (it is not as cut and dried in high frequency situations), current is equal to voltage divided by resistance. Always.

So, for every DC circuit, there are three unknowns: current, voltage and resistance. Supply two of those values and you can do simple algebra to find the third.

[Butcher]

I was deliberately not quoting that because it doesn't quite hold for semiconductor junctions, peltiers are not ohmic devices so the R isn't constant at different voltages.

[Brians256]

You are correct. I was too emphatic in saying that Ohm's law is always true. I know that it is not an ohmic device, but, it's close enough to do some ballpark figures. Temperatures and differing current densities affect the resistance.

[futRtrubL]

I think you have two problems, well one problem, which has been talked about, and one bug.
The problem is the voltage drop on the PSUs, this can be fixed with better rated PSUs.
The bug is board not using the on chip temp sensor. If the mobo sensor is on the pcb them it is covered with conformal coating as you said. This would insulate it from the CPU and so your readings would be closer to ambient than expected. You said that insulation wasn't complete, if that's still the case try touching the cold plate.

Edward

[Brad]

how much more overclockability have you got with the extra temp drop? it might be a lot more than your temps would tell you. This is because the temperature drop is a lot more than what your diode will let you know

[immudium]

Just thought I would share the results of some of my weekend tinkerings. I reworked my plumbing a bit so that the radiator feeds directly into the waterbock now instead of first passing through the pump. I was also able two twist the bolts of the maze2 a couple of notches tighter and luckily without significant blood loss and/or any other kind of damage to myself or the block.

After doing the above, I got another 6 or 7 degree drop so that I am sitting at about 7 degrees C after 30 minutes. I think I still have some room for improvement, but I am somewhat more satisfied with the results thus far based upon the results of the little pelier calculation program I am using (writtem by Miles Asvachin from this forum). I don't know how complete my input parameters are, but according to it, the min temp I should be expecting should be around 6 degrees.

Anyway, I have not idea how some out there are consistently getting in the -10 degree range and beyond, but another reason I think my temps are OK is that I am still unsure how accurate my cpu temp reading is. I have been querying the forums at anandtech and amdmb concerning the thermistor configuration of the Dragon+ but have been unable to dig up any definite answers. Regardless, if I peek underneath the layers of my neoprene insulation I can see some definite frost on the corners of the cold plate.

Anyway, concerning overclocking my Athon 1.4 (1600+), I haven't done any extensive testing as far as stability and stuff goes. As yet I have not unlocked my Athlon XP though I have the silver laquer and everything to do it. However, I have been able to successfully boot into win2k at 10.5 X 150 = 1575 with default voltage and have been able to boot as high as 10.5 X 170 = 1785 at 1.85V however at that point I think I am beginning to push the limits of what my Crucial ddr memory can handle. Despite this, with the added voltage and speed, I only saw about a 3-4 degree rise in cpu temp which I think might be encouraging.

Also, one last note about PSUs. I really need to thank you guys again for the great info you've given me on this thread. I never really knew before what to look for in a power supply and I have been noticing for the first time what some of the PSUs I am using are actually capable of. For example, my two old 230W PSUs that I am using for the peltiers are both rated for 9 amps, however some of my bigger 300W and 400W PSUs used to actually power my systems are only rated for 6 and 8 amps respectively which I find strange, but good to know nonetheless. I don't know if there are actually AT PSUs rated for 13 amps, but I know what to watch for now and I'm sure I could at least find a couple of 10A ones like those used in the article.

Immudium