Peltier instead of radiator in WC setup?

[R.C.Anderson]

I've just been browsing some watercooling equipment, and an idea came to me, what would it be like to get rid of the rediator out of the watercooling setup, and instead get a peliter cooler, and attach the cold side to a heatsink submerged in the reservoir, and cool the hot side with either a large passive heatsink, or a decent CPU HSF?

I've had a quick googling session on the matter, but have only found people who have tried it with old components (back in the P3-P4 days) is this because they just couldn't handle the higher heat output of the newer chips, or have people just not been interested?

I would love to know, as if I get a watercooling setup, a reasonable peltier is a similar sort of price to a decent radiator from what I remember.

 

[HW91]

no, it would be worse than a rad, there's a reason why people only use water cooling to cool the hot side of peltiers, because otherwise it can be worse.

also the peltiers would also bring sub zero temps to the table so unless you were using methanol(the liquid generally used for sub zero water loops) the liquid could freeze

 

[Tikkabhuna]

You place the peltier on the chip and the cold side cools the chip. Then you have to use watercooling to cool the hot side. So you still have a normal watercooling loop but it is cooling the peltier instead. For a peltier you need a seperate power supply. There's a bit of information about it in the cooling sticky.

 

[JonJ678]

It's just about possible, in that you will indeed remove heat from the water (like a radiator), and the heatsink will have a go at removing heat from the peltier. Beyond proof of concept it makes no particular sense though. Putting your big hsf on the cpu would be far, far more effective and much cheaper.

Peltiers are not cheap. They cost a hell of a lot more than watercooling, and under some circumstances will cost more than phase. They're almost universally a bad idea for computers. That's not to say I think they're a bad thing, but buying one on the basis that you think it'll be cheaper than a radiator is absolutely not the way to go.

 

[HW91]

Peltiers are not cheap. They cost a hell of a lot more than watercooling, and under some circumstances will cost more than phase. They're almost universally a bad idea for computers. That's not to say I think they're a bad thing, but buying one on the basis that you think it'll be cheaper than a radiator is absolutely not the way to go.

you can get 400w peltiers for £10 off ebay, add in the right PSU and it will cost less than £80

considering phase change systems can cost about £600 pre-built

 

[PaulyD]

Nice to see you still about Jon

Thought of you when i saw this thread!

 

[R.C.Anderson]

You place the peltier on the chip and the cold side cools the chip. Then you have to use watercooling to cool the hot side. So you still have a normal watercooling loop but it is cooling the peltier instead. For a peltier you need a seperate power supply. There's a bit of information about it in the cooling sticky.

Yea, I've seen that, however my thinking was, if it was a decent enough it would be able to chill the whole cooling loop, so the gfx card, and NB is also cooled by it, rather than having to buy multiple peltiers.

Another think I thought of was to use a peltier as well as a radiator, in an order similar to this: waterblocks>radiator>reservoir with peltier>pump>waterblocks, as this would allow the radiator to take out most of the heat (as it would in a normal setup) but then the water would be able to be cooled to slightly below ambient temperatures (I would do some calculations, and get a micro-controller to control the peltier to keep the water above the dew point (Electronics is my thing, so making the micro-controller circuit, and programming it would be enjoyable, and doable)) This would also allow a lower powered peltier to be able to be used. Any opinions on this?

I've also just checked prices on peltier cooler, and I can get my hands on a 200W for around £25 (from farnell, so avoiding ebay) and the PSU isn't a problem, as if I get a new build, I'll be buying a new PSU for it, so I'll be able to use my 450W corsair that I have in my current machine.

 

[JonJ678]

you can get 400w peltiers for £10 off ebay, add in the right PSU and it will cost less than £80

A 400W peltier from the bay means, if it's rated honestly, it'll move 400W through a temperature gradient of zero when running at maximum voltage. The point to a peltier is to maintain a temperature difference between the two sides, the Qmax value (400W) isn't the wattage you can hope to move with it.

I was budgeting approximately £300 for a prototype, not including power supply. That was using pelts running at 6V, moving approx 40W each through a gradient of 20 degrees. There's a build on here by Pneumonic somewhere which shows what peltiers involve.

It's a bit different if you can find a peltier that can comfortably deal with the wattage from the processor, and decide to go cpu -> peltier -> waterblock directly. If you find a peltier that can move 250W through 20 degrees do let us know, I don't believe one exists.

Nice to see you still about Jon

Thought of you when i saw this thread!

Cheers man. I'm about, sort of. Hoping to post more often once life calms down slightly and I can get into a manufacturing lab again.

waterblocks>radiator>reservoir with peltier>pump>waterblocks, as this would allow the radiator to take out most of the heat (as it would in a normal setup) but then the water would be able to be cooled to slightly below ambient temperatures

I've also just checked prices on peltier cooler, and I can get my hands on a 200W for around £25 (from farnell, so avoiding ebay)

so I'll be able to use my 450W corsair that I have in my current machine.

Interesting that you're a circuits guy, that's a gaping hole in my knowledge at present. I was relying on a simple on/off approach which would be rather worse than microprocessor control.

The wattage a radiator removes is linearly proportional to the temperature difference between water and air. C/W values are available online. If you try to bring the water below ambient, the radiator will now be working to warm up the water. As such, a radiator in the chilled loop is brilliant as a means of preventing the water going too far below ambient but a really bad idea if you want it to go sub ambient.

The general advice is not to use atx power supplies with peltiers, for fear of killing the psu. If nothing else, be aware that your 200W pelt is probably rated at 200W, 15.4V. If you run it at 12V, you need to downrate the 200W considerably.

 

[Freddie1980]

Delete - sorry should have remembered you wanted to combine this with water cooling (suggested buying the CM V10).

 

[R.C.Anderson]

The wattage a radiator removes is linearly proportional to the temperature difference between water and air. C/W values are available online. If you try to bring the water below ambient, the radiator will now be working to warm up the water. As such, a radiator in the chilled loop is brilliant as a means of preventing the water going too far below ambient but a really bad idea if you want it to go sub ambient.

the idea I had was that the water going into the waterblocks would be sub-ambient, however the water coming out of won't be for example:

Reservoir with peltier (10C)>Pump>Waterblocks (30C)>Radiator (25C)>Reservoir

That's just an example (no idea what any of the temperatures would actually be) with the temperatures being the coolant temperature as it leaves that comonent.

The general advice is not to use atx power supplies with peltiers, for fear of killing the psu. If nothing else, be aware that your 200W pelt is probably rated at 200W, 15.4V. If you run it at 12V, you need to downrate the 200W considerably.

I've actually just noticed that it's rated at 24V, so I would either buy/make a seperate 24V PSU, or I would switch the 12V output of a PC PSU upto 24V, but the electrical side isn't what bothers me at the moment, it's whether there would be a sufficient performance increase from this?

 

[Its_Me]

the idea I had was that the water going into the waterblocks would be sub-ambient, however the water coming out of won't be for example:

Reservoir with peltier (10C)>Pump>Waterblocks (30C)>Radiator (25C)>Reservoir

That's just an example (no idea what any of the temperatures would actually be) with the temperatures being the coolant temperature as it leaves that comonent.



I've actually just noticed that it's rated at 24V, so I would either buy/make a seperate 24V PSU, or I would switch the 12V output of a PC PSU upto 24V, but the electrical side isn't what bothers me at the moment, it's whether there would be a sufficient performance increase from this?

The Delta-T of the water temps in/out of the rad in/out of the waterblock will be a lot lower then those you have mentioned.
Anyway, the simple answer is NO, don't bother.
There are some rather old and long threads about the use of TEC's, that go into great detail and calc kicking about if you use the search function.
I used a TEC on various CPU's, with the first been a SL2TV, last viable was a AthlonXP 2700+.
Direct Die was by far the most efficient use, sadly the loadings started to negate their use.
TEC cooling is viable for loop cooling, but the efficiency/cost of the system for the relatively small (if any) benefit is crazy.

BUT! (an alternative)

If you want to mess with super cool loop temps, buy a used Aircon unit (THIS type of unit), and strip the evaporator & dump the evap head into a large reservoir. Then put your computers cooling loop rad submerged into that.
Chill for a couple of hours so your large res water is sub zero, and then fire up your system.

Condensation is a killer, so if you do not want to use lagging/sealing of your board - check relative humidty is LOW and only run till your pipes start to sweat.
You will see build up of condensate very very fast, and it will nuke your board !

A similar even easier way(and safer) is to use a large Res filled with ice, but the temps will not be as low and the capacity (i.e temp rise over time) is low.

The advantage of these is that you do not have to split your computers loop, meaning its ideal for just messing about with a crazy overclock/benchmark.

I can't stress enough how much damage condensation will do to a system, I have lost count of the number of SlotA systems I killed even after spending hours sealing them up/use of neoprene.

 

[Solaris]

Also, keep in mind that if the cooling of the peltier element fails (e.g. fan failure or pump failure in case of watercooling), the results will be more disasterous that if a conventional cooling system fails. Even if your CPU has a thermal protection that will cause it to shut down if the temperature gets too high, the peltier element may still kill it by continueing to heat it up long after it has shut itself down.

 

[Its_Me]

Also, keep in mind that if the cooling of the peltier element fails (e.g. fan failure or pump failure in case of watercooling), the results will be more disasterous that if a conventional cooling system fails. Even if your CPU has a thermal protection that will cause it to shut down if the temperature gets too high, the peltier element may still kill it by continueing to heat it up long after it has shut itself down.

I tested this a long time ago, in a closed loop system, thermal transfer in the water loop occurs enough to stop the cpu from been completely overheating.
I even tested to destruction in an open loop where it boiled off the water in the block completely and melted the polycarbonate waterblock cover , thus destroying the waterblock, but the CPU was fine.
These on 220w TEC @ Vmax/ Imax.

 

[Solaris]

Oh fair enough, good to know...

 

[Its_Me]

Oh fair enough, good to know...

yeah its one thing that was a concern, yet it always boiled down (no pun intended) to condensate issues killing hardware.

 

[JonJ678]

I even tested to destruction in an open loop where it boiled off the water in the block completely and melted the polycarbonate waterblock cover , thus destroying the waterblock, but the CPU was fine.

That is very useful information indeed. Thank you for running the test, and for posting the result.

Reservoir with peltier (10C)>Pump>Waterblocks (30C)>Radiator (25C)>Reservoir

Oh. So you've never actually watercooled anything then? Start with that, not with peltiers.

it's whether there would be a sufficient performance increase from this?

And this means you haven't looked at a voltage/current vs dt graph for a peltier yet. You really should do, it'll make a lot of this far clearer.