Water-cooled peltier system acting strange

[bob]

Here's a list of what I used to make my water-cooled pelt system:
- Hayden transmission cooler - 5" x 12 1/2" x 3/4" (basically, 2 120mm fans covers all the fins)
- Maxi-Jet 750 - mag-drive pump going about 180+ gph
- BE Cooling Cu block
- Frost72 peltier running at 13.8V
- 1.5"x1.5"x1/4" Cu cold plate
- Tripp Lite PR-7b - 13.8V, 5A continuous, 7A max
- 2 80mm Sunon fans attached to the cooler (blowing out), running in series off the Tripp Lite
- Artic Silver thermal paste
- distilled water/Redline Water-Wetter mix
- misc - surgical tubing, vitamin bottle resevoir, home-made clip

What's the above attached to:
- everything above sits inside an Inwin A500
- block is attached to Asus S370-DL sloket insulated with neoprene front to back, GE Kitchen & Bath
Silicone, and dielectric grease
- in the sloket is a Celeron II 600 SL4PB (cC0 stepping) running at 1.01GHz (112fsb) at 1.80V
- AX6BC R type mobo which according to MBM 5.04 uses the GL518SM-1 sensor chip
- the rest of the equipment doesn't apply to the problem, at least I don't think so

The problem:
According to MBM, the water-cooled pelt system is doing worse than the Gorb when I run at 1.01GHz. At full load I see an avg temp of about 35 degrees C with the Gorb. Whereas with the water-cooled pelt system, I see an avg temp of about 40 degrees C. In both cases, I had my case cover off. (Btw, I calibrated MBM based on what SiSoft Sandra Pro 2001 reads.)

Question:
WTF is going on? Aren't I supposed to get low temps, maybe even below zero with a water-cooled pelt?

Plea:
If anyone has an explanation and/or solution to my problem, I would greatly appreciate your info.

 

[Eriksson]

Calibrating mbm with Sandra is not a good idea in my opinion. Sandra is frequently way off in temps. I got Aopen AX6BC board and I have actually had strange temp readings both with mbm and sandra. You have to keep in mind that the temp reading is done in your Asus S370-DL sloket !!

What voltage rating is on your peltier.? Is it running at optimal voltage at the TERMINALS? Does your waterline/cooler setup feel hot? Is the back of your slocket hot/cold (might give idea about cpu temp)?? Do you see temp variations under different cpu load?

 

[DrMac]

Did you check the temp with just the waterblock system and without the pelt?

Take the pelt out and check that your system runs cooler with the water cooling system alone. Your waterflow may need to be adjusted, i.e. your water may be flowing too fast or too slowly (unlikely!) to absorb the heat from the CPU. There is an optimum rate of flow to allow for heat transfer from one component to the next.

Also, you seem to have changed two variables at once, so you can't tell which one is affecting your results.

The use of a peltier device for cooling a CPU is not quite as straightforward as it may seem. The peltier actually produces heat, which must be safely and effectively dissipated, or else your system could possibily run even hotter than without it, which yours apparently does, or perhaps even fry your CPU. A peltier doesn't really get cold, it just transfers heat from one side of it to the other, and if that heat is not subsequently transferred to somewhere else, you have a total heat (ie temp) gain, and not an overall cooling effect.


...and remember, colder doesn't always translate into higher speed.

 

[bob]

To Eriksson:
- For monitoring temps with the Aopen mobo, do you think it would be wise to purchase something like
CompuNurse?
-The specs for the Frost72 is 16.3V max at 6.1A, 62 watts.
-Waterline/cooler are warm to the touch.
-As for voltage at the terminals, I'd have to get back to you on that.
-Ditto with the slocket back temp, I'd have to pry off the neoprene siliconed to the slocket.
-Ditto too with the different cpu load, I'm currently using the comp for school work.

To DrMac:
- Testing with the water-cooling system haven't occured to me. I used to have a Al block with no pelt, I got the Cu block in hopes of using a pelt to lower temps. I'll have to get back to you on how the Cu block fares without the pelt.
- About your heat gain theory, I think you might be onto something. Perhaps the heat is not being taken away quickly enough. Here's a thought. I tested out the cooling system unattached to the cpu and I get frost on the cold plate. Doesn't that mean the water-cooling is working properly and the cold side goes below zero? Anyway, I'll rip off the pelt off the block and try it out like that. Btw, how do you safely take a superglued pelt off a block? (I only applied at the corners.) Once I do get it off, I suppose I'd have to lap the block.

To both:
Thanks for your info and I'll get back to your questions when time permits.

 

[bob]

Quickie update:
- I popped off the cold plate/pelt/block to find subjectively find out what the temp on the cpu felt
like. Anyway, I popped it off and lo-and-behold the cpu was cool (like when you stick your hand in
a fridge, not the freezer part ... something close to that) and so was the socket itself. (Eriksson - I didn't
check the back of the slocket, but I took off the cpu and pulled out the neoprene plug underneath it and the
pcb there was cool.) What does that mean?

 

[Eriksson]

Well what I was getting at.. the back of your slocket is the closest spot to the cpu when running. If you can feel it is cold your cpu must be too.

If your cpu and pcb feels cold after you popped off the gear, it means either your peltier was not turned on at the same time as the comp. (it is easy to cool non powered cpu). Or your actual cpu temps are quite low

The water flow is NOT critical is this situation, if you got any considerable flow at all. Your cooler should be few deegrees above ambient, and I guess the "feel warm" temp is around 40c which is ok. If your cooler is 60+ c° then you need to do something rational.

Your peltier might need a little more voltage to deliver max effect,
I cannot give you any advice on how to accurately measure your cpu temps However mbm 5 (uncalibrated could work) Trying your system with the copperblock alone is a good idea, and use several speed / voltage combinations, primarily to see if the cpu temp reading makes sence.

Good Luck and keep asking

 

[Big Lar]

Sounds to me like it could be a coupla things, 1st, Do you have a way of getting a temp of the actual coolant while running? I believe I saw you say the Res. is an aspirin container?? Deffinately not big enuff, also, the Bench supply of which you speak, may not be giving you enuff amps for the Frost, I run a single Ice71 nowadays, and have a Radio Shack 13.8v 10Amp supply, and it barely does the job for me, in fact a good 300 watt atx supply will give you a tad better amperage then the bench supply.

 

[proze]

sounds to me like the pelt isn't getting near enough juice.

 

[bob]

To proze and Big Lar:

Lack of power, eh? Hmmm ... I figured that the power supply I got would be ample to supply the Frost72. I don't remember my physics much, but I figured that the Frost would want about 5 amps at 13.8V, whereas it wants 6.1A at 16.3V. Anyway, I was looking at eBay and saw a 12A power supply, do you guys think that would be enough to power the Frost properly?

 

[Eriksson]

I would not take it for granted that your psu is to small. Many ppl think that peltiers deliver max heat at max voltage, it simply is not so. The max voltage rating defines how much voltage the unit can take without damage.

Look at the voltage Dr Rom is using here and his comment on voltage.

http://www.overclockers.com/tips326/

I found that my peltier gave best cooling at 12-13V even though it is rated 15V max.

Tip: measure the voltage at the peltier terminals, if it is 13v or higher your psu is big enough. If your psu cannot deliver enough current for your tec the voltage will be lower.

 

[bob]

I tested the pelt's terminals and I get 13.5V which, according to Eriksson, means that the Tripp Lite is ample. Also, I had a chance to pop off the pelt and am now running water-cooling only and am running a tad bit cooler. I don't trust MBM completely, but it says I dropped about five degrees Celsius.

On another note. I was talking to my friend who basically has the same setup that I do and he said he had the same problem. His crappy *** AMD certified hsf worked better than his watecooled pelt system. His system consists of the same radiator, cross drilled block and inline water pump less powerful than mine. He thinks, possibly, that the Hayden oil cooler is the problem. That would be peculiar, because the radiators sold for water cooling are just variations of the Hayden or looks like a cut in half Hayden. Anyway, what are your thoughts on this?

 

[Eriksson]

Ok good your tec is getting enough voltage I think. Lets make clear that you MIGHT get LITTLE better results with higher voltage, but this is NOT your problem.

It is easy to find you if your Hayden cooler is the problem. Since water has very high thermal capacity, it takes some time to heat up. If you cooler is too small and cant remove the heat away, water temp will gradually climb.

Therefore when you first fire up your system, water temp will not be a problem, it will become one in few minutes if your cooler is not big enough.
This means you get few minutes to see what the peltier can with low water temps. Right!?

So unless your cpu temp is low right after boot and then climbs your cooler is not the problem.

Btw your problem is too small peltier. It is rated 62 W, and It cannot transter enough heat. Celerons running at G+ with high core voltage put out probably 40W+, and given the losses due to heat resistance in materials it is too much for your tec.

Check this by putting your pelt back on and see what happens when you run your cpu at stock speed and voltage. It will be giving away much less heat, and your temps should drop a lot.

 

[bob]

Apparently I only get a couple of degrees lower at stock speed and I also dropped the voltage by 0.2V from the stock voltage of 1.7V. As for the pelt wattage, if the pelt is rated at 62watts at 16.3V, at 13.5V it would have a wattage somewhere in the 50s. Do you think I should get a beefier pelt, 74watt maybe? And if I do, do you think the psu can handle that, because I don't want to get a new psu. I got it someplace like the RipShack, damn thing cost me $55. Should have bought one off eBay, more powerful and cheaper. Anyway, thanks for the continued advice Eriksson.

 

[Eriksson]

Well I hope I did help you, not just confuse you.

Actually I am shooting in the dark, regarding the size of peltier you need. Since I have not studied the celeron heat output, you can get a lot of info on the main page here regarding peltier size. I guess if you get 70-80W unit it will need more power than you psu can deliver.

I find it a little strange that your cpu temps didnt drop more at low speed/voltage, quite strange. I think you need to check your install carefully, did you put thermal paste on your peltier, both sides? This is a must do, also good clamping pressure is needed for proper heat transfer. And make sure the coldplate is sitting flat on the cpu core. I got lower temps with single 68W peltier on a 700 Duron at stock speed voltages. And Durons pump out more heat than celerons.

Good Luck

 

[Leo]

Big Lar:
Some thoughts:
Heat dissipation of the peltier rises with the second of the voltage. Heat absorbation of the coolant rather linear (at least the heat flow through the thermal resistance between peltier and coolant rises proportional with the temperature, if you assume a perfect cooling).
There is (somewhere) a point where the heat cannot be removed any more. The only solution is to reduce the thermal resistance between the semiconductor-metal-boundary (of the peltier) and the coolant and that means to use an bigger transverse section for the heat transfer ergo a bigger peltier.

 

[dasouthernocer]

You could try taking the peltier setup off of the CPU. Then run it independently and if the peltier starts collecting ice it should be working properly. You may need a new peltier I'm not for certain on that one. My melcor peltier is a 72Watt @14.4V and it is supposed to draw somewhere like 8amps right now on a varible PSU it's running @ 15V and it barely draws 6 amps. So you may not have to replace your PSU if you had to you could get an old 300Watt AT to power it fairly cheap.

-dasouthernocer