Using a underpowered pelt on purpose

[drsmoothie]

I was wondering if there is any benefit to using a low power peltier.

I have a water cooled setup, and my temps are running at around 30 C idle and 38C at full load. My idea in using a low power pelt is to reduce the the temp a little bit to around 20C. This way, you have additional cooling using a pelt without having to worry about condensation issues as the temp does not go below ambient.

[DSREC]

Nope! If anything it will raise your CPU temp and case temp. You forget that even if you're going to use a low wattage TEC, its still going to pump out alot of heat. not worth it

[resago]

however, what you could do with a 156 or 172 watt pelt, is adjust the voltage according to temp to keep your cpu @ ambient. then you still wouldn't have to worry about condensation. :cool:

[drsmoothie]

How about adjusting the voltage being applied to the peltier? Is this possible?

Thanks.

[Butcher]

Yes I designed a small circuit the other day in fact which does just that. You sit a temp probe on your coldplate and the voltage varies with temperature... this was to be built into a psu though, not an addon unit so I'm not sure it'd be very useful to you, however I also have some designs for temp sensing fan circuits which could be beefed up a bit to use with a peltier. The main problem is cooling the transistors in such a circuit.

[resago]

could you post a schematic?

[platypii]

Originally posted by DSREC:
Nope! If anything it will raise your CPU temp and case temp. You forget that even if you're going to use a low wattage TEC, its still going to pump out alot of heat. not worth it

I've seen people say things like this before, but i don't see any logical reason why it wouldn't help. Preferably answer in a Private message or email since I don't check here often.

[LiquidCool]

As long as your peltier is rated HIGHER than the wattge that your CPU is putting out you will be ok. Try to get about 130%-150% of what your processor is putting out. EG If your processor puts out 50 watts, then you will want a peltier rated around 75 watts at 15 volts. They provide less cooling at 12 volts so you would need an even larger peltier.

The reason it would overheat is that a peltier doesnt "create" cold, it just moves a certain amount of heat. If the CPU is putting out more heat than the peltier can remove then the peltier heats up, which in turn causes the CPU to heat up, which causes blue smoke to be realeased!

[Kevin]

DONT DO IT

[Butcher]

Yeah I'll try to get one into electronic form this weekend, I normally design on a pad with a pencil :)

[Butcher]

Ok the ciruit is ready! Hopefully you can follow my crazy diagram and explanation :)

http://www-users.york.ac.uk/~rec108/circuit.jpg

R1 and R2 form a potential divider, R1 being a thermistor and R2 being either a carefully chosen fixed resistor or a 'preset' variable resistor (the type you change with a small screwdriver). As the temperature varies the voltage to the gate of the MOSFET Q1 is varied, thus causing the resistance across the source-drain path of Q1 to vary. As Q1 is in series with the load RL (fan, peltier, etc.) this varies the voltage across the load.
The whole thing costs next to nothing and is pretty simple to assemble, the only thing to watch out for being the temperature of Q1. Q1 can be any N-Channel MOSFET of suitable power.

[platypii]

I believe all of you who say an underpowered peltier will make it hotter are dead wrong. Your assertion that a 50W peltier (for example) can only move 50W of heat is first of all wrong. That the is the amount of energy it draws -- and therefore adds -- to the system. Your theory that it couldn't move enough heat if it were underpowerred would hold up, on the assumption that the peltier was a perfect insulator on its own, which it definitively is not.

[LiquidCool]

BRO! It's a fact, not an assertion. A peltier acts EXACTLY the same as a compressor based system. If the heat load is to high for the peltier, it WILL overheat and fry your cpu.

Like I said a peltier does NOT GENERATE COLDNESS. It MOVES HEAT. The wattage rating specifies how much heat can be MOVED through the device before it overwhelms it.

A peltier also creates heat as the byproduct of moving heat. so the peltir may move 50 watts but to mave that 50 watts it creates XX amount of heat. The problem is that once the heat load overwhelms the peltier, then all you have is more heat being generated.

Oh well, your going to try it anyway but first go here http://liquidcool.org/main.php?zone=500&keyword=what_is_a_peltier at the bottom of the page is some peltier manufacturer links, read them and learn then try your foolish idea...

[platypii]

I was slightly confused about the wattage rating being its transfer capability and not its power draw, however my point remains that all your theories are assuming that the peltier is a perfect insulator (for the heat not transfered thermoelectrically). Do you believe that you will fry your cpu if you simply unplug your peltier? essentially making it a 0-watt peltier. No. The excess heat not transfered thermoelectrically is transfered via conduction, like a heatsink.

[platypii]

continued...

This way, if i were to put a 56 Watt peltier on my 78 Watt Athlon, 56 Watts would transfer thermoelectrically, while the remaining 22 Watts travel via the less efficent, but still functional method of conduction. I am still open to a reason why this wouldn't be the case, but I have yet to see it anywhere on the internet -- including the manufacturer's sites. To me it just seems like an internet urban-legend. Or an attempt to sell higher rated peltiers?

[LiquidCool]

NOT being rude right now, just had a long weekend. BUT, you are still confused...

The peltier makes a VERY GOOD insulator. It is in fact not anywhere near solid. It is 2 pieces of ceramic with special metal connections in the middle. These connections are about the thickness of paper and maybe 1 mm wide.

Go look at the people that are running high overclocked, high voltage Thunderbirds. They need to use 2 120watt peltiers because 1 172watt peltier cannot move the heat fast enough.

Like I said though, try it. Thats the only way to get it out of your system. Just do it on a CPU you dont need later :)

[platypii]

Personally, I don't intend to try this because i'm already running a 122 Watt peltier on my Athlon, and am on a bit too tight of a budget to try a low powered pelt on another computer, however from experience I know that a peltier does conduct because I have in the past (accidently) left the peltier powered off, and it didn't fry. The reason it does conduct (though not superbly) is because the individual thermocouples are two layers of metal, which do conduct.

[resago]

stop lying, theres no way a 122watt pelt is keeping a athlon@ 1.5 cool, thats just stupid. :mad:

[platypii]

Do the calculations. the Athlon is rated at 78 Watts of heat. Figure it's running at 125%, and although I know it doesn't work this way exactly, that translates to 97.5 Watts. By my count there's still plenty of leeway.

[drsmoothie]

Alright, I have a new idea. How about using a small (both physically small and small power-wise) peltier and mount it in the middle between the coldplate and the waterblock but also place a rim of copper shim (that is the same thickness as the peltier) around the peltier?

This way, the heat produced by the cpu will still have an avenue to conduct to the waterblock (through the copper spacer around the peltier) and the peltier will provide additional cooling to the cpu core. If you do it just right, I think you can set it up so that the temperature is just above ambient. The whole point of this is to provide additional cooling to the watercooled setup but don't have to worry about condensation issues.

Waddaya guys think?

Jerry

[LiquidCool]

That wont work either. Think about it your thermally connecting what your cooling and what your heating. You will still end up adding more heat than your moving.

It's really not that hard to insulate your processor. Why are you so concerned with condensation? Get some neoprene and a can of conformal coating and have at it...

[drsmoothie]

Michael,

Is there a webpage that illustrates with pictures how to insulate the cpu?

Once the insulation is in place, how hard it to remove it? Does it come off the mobo cleanly?

I also have to insulate the bottom side of the mobo, right?

Lastly (sorry about the barage of questions), for a 1.5 Ghz tbird, how big of a peliter do i need to use? When people use dual peltiers, do they stack them in series?

Thanks.

Jerry

[resago]

I'd want at least 200watts of pelt.

maybe dual 120 or 156 or 172.

If you want to stick to 1 PS (651 enermax)
run 1 156watt @ 12v, AND another one @ 5V.

just look at the pelt curves to add up to the right wattage.

most people use the maze2.2 with pelts side by side.

[LiquidCool]

It's not to hard at all!

Your going to need some type of foam. It has to be a closed cell type, something that doesnt let air flow through. Neoprene rubber is good. Another good solution is the foam that meat comes on, the stuff from the grocery store. It's usually nice and flat, and if you slip the meat guy a $5 he might give you a boxfull of it :)

You are also going to need some vaseline or di-electric grease. You'll use this to pack your processor slot/socket pins.

You dont need to insulate the back of your motherboard. It doesnt usually get cold, unless your doing some really heavy duty cooling...

Depending on how your block mounts, you will have to do it a little different. But all your trying to do is keep the stuff that gets cold from having contact with ambient air.

I dont have my camera right now so I cant show you an example, but here are some other sites that have good instructions and pictures on sealing your CPU up...

Benchtest.com (http://www.benchtest.com/condensation2.html)

I get my camera back in a day or two so I'll whip up something and take some pictures for you...

[resago]

heres how I insulated my cpu.

1)put dielectric grease in the pinholes of the socket (VASELINE)
2)dielectric grease (vaseline) on the bridges.
3) 1/4 thick foam of back of MB ~4"x4"
4) silicon in the empty space in socket (don't make it too high)
5) place the cpu in prepared socket.
6) silicon around the socket and zif bar.
7) silver paste on cpu core
8) place temp probe now if you have one.
9) make a platform of silicon for the WB to contact on mb around the cpu.
10) tighten down waterblock, silicon around edges up to makerblock level.
11) after 24hrs or more, use great stuff expanding foam around water block. If you haven't siliconed a barrier around the waterblock, the foam will creep under and lift the waterblock off the core(BAD).

look here

www.overclock-supercool.com (http://www.overclock-supercool.com)

[Spinoza]

Heat and Power Usage go up with the Square of the frequency... all other things equal.

[Butcher]

Isn't it down with square of frequency hence switch mode power supplies running at much higher frequency than standard AC line frequency (200KHz ish vs 50 or 60 Hz).

[Spinoza]

I was going for the short version. I was referring to CPU heat generation going up with the square of the freqency. This is why raising the voltage is key to big overclocks. Sure.. at stock voltage there's a little play in the activation level of the transistors... but not much... if you raise the frequency more energy goes into heat, soon there's not enough energy left to run the chip and instablility insues.
Yeah yeah.. I know all the physics of this.. I'm going for the non-EE version.

[Butcher]

It's normally reckoned that heat is linear with frequency and square with voltage...

[Spinoza]

My god... You are absolutly right. I had it backwards. Thank you for the correction.

[vfrjim]

I found a great article on Peltiers I think most people will enjoy, sorry if it was posted previosly,
http://www.tellurex.com/resource/txfaq.htm

Read article 45 on underpowering a peltier.

Jim