A question about building my unwisely-extreme hi-performance cooling system

[Faye_Kane_girl_brain]

Okay, I put a thermoelectric (TEC) slab between my CPU and a cheap, one-fan, pre-built water cooler. It melted the pump, which was mounted directly over the CPU. It didn't melt as in "flows down to the floor", but it was visibly distorted and the rotor made noise but wouldn't spin.

So I built my own custom water cooler, and I put a 3x120mm radiator in the loop—BEFORE the hot water gets to the pump.

I also bought an asskick pump, since the little one obviously wasn't flowing enough water over the hot side of the Peltier to keep the gizmo from melting.

You may not have this problem, even if you cool thermoelectrically. But I use a 24v, 60mm 420-watt industrial TEC (It has its very own PSU, which also drives the pump). My goal is to freeze the processor solid (below 0º C) while running prime95. Why? For the only good reason to do anything at all: because I want to.

It may not be possible, but grownups say that about everything fun. I'm going to do it anyway.

So far, though I've only run the thing at 12v because I'm scared it will shoot sparks like the Enterprise control panels when the Klignons attack or something. I'm going to wait until I can afford to upgrade my MB, so I have a spare of everything when it blows up in my face. I have a history of unwise experiments which did that. In fact, I pulled the Peltier out altogether until I have spare components to destroy.

I know that instead, I could just throttle the Peltier back and slow down the clock, but that defeats the fun of being a speed freq. My main concern is that peltiers put out twice as many watts of heat than they suck up on the cold side. Plus, when my grad school fall semester living expense $$ arrives, I'm going to buy a 220-watt AMD 9590, which runs so hot that AMD sells it with an AMD water cooler. Add another 420 watts at 24 v, and I'm wondering if it will overwhelm even MY cooler. I use 1/2" tubing, and am prepared to add another pump and radiator. Even two or three more, if necessary. (I do everything with excessive intensity, otherwise it's not worth doing. "Too much" is NEVER enough!)

My question is:

What is the max heat in watts that I can expect that setup to dissipate?

Also, do you think I'll need a second radiator? There are probably equations about it somewhere, but I'm too lazy too look them up.

☺,

-faye

 

[sneekypeet]

AIOs are only good for 150-200W of heat dissipation. For something like you are trying, you need a custom loop.

In reality, to use the TEC to its fullest, you need 500W worth of radiator, which is slightly more than a 3 X120mm rad will handle comfortably.

Also you may want to ponder what happens if the TEC is left active and the pump fails or lost power in this endeavor. Some gruesome images of the potential ahead of you.
http://www.techpowerup.com/forums/t...hen-a-tec-overheats-look-inside-to-see.38153/

 

[Faye_Kane_girl_brain]

First of all, what's an AIO? It wasn't in Wikipedia and it's too technical for urban dictionary.

> For something like you are trying, you need a custom loop.

??
My whole post was about the fact that I'm building a custom loop. It's not fully completed, but it works anyway, just like the Death Star.

> In reality, to use the TEC to its fullest, you need 500W worth of radiator,

I'll need a lot more than that. If I run it at full blast, the TEC alone is 420 watts.

> which is slightly more than a 3 X120mm rad will handle comfortably.

Really? Is that all? Damn! Okay, I'll get another one in two weeks when I buy my 9790 processor.

► Do I need a new pump, too?

Also, if one pump dies, will the second one still work, or does a nonfunc pump block the flow? Also, can you point me to a table somewhere with all the parameters concerning heat dissipation in PC watercooling, like what effect tube diameters and stuff have?

> you may want to ponder what happens if the TEC is left active and the pump fails

Hey, pondering is what I do best! ...Well, second best☺. I have a solenoid gizmo that cuts off the 120v if its 24v input drops. In also have the system PS slaved to the pump and Peltier PS, so if PWR_GOOD (the orange wire) goes low, the computer stops.

But you're right; there are other failure modes. Do they sell a sensor that will open a 120v circuit if the water flow stops? I'm already running a utility which shuts the system down on overtemp, but I don't trust single factor prevention of horrible disasters. I picked that up when I designed a neutron monitor for a reactor, and it had to be fail-safe, without question, no matter what. BTW, this book about failure modes in complex systems is absolutely amazing. I bought a copy for my boss.

He never read it though, and that's why Three Mile Island happened.

Just kidding.

> Some gruesome images of the potential ahead of you.
http://www.techpowerup.com/forums/t...hen-a-tec-overheats-look-inside-to-see.38153/

Woh! I'm there.

Thanx, Peet!

-faye

 

[FR@NK]

How are you managing condensation?

 

[sneekypeet]

AIO is an All-In-One water cooler like the Corsair Hydro series coolers.

TEC heat is the issue, since the TEC is cooling the CPU, the CPU heat is not dumped into the loop, just the TEC heat is what the loop is cooling.

As far as I am aware, each 120mm radiator section is worth ~150W of cooling potential.

Of course you will need a pump to make the coolant flow, that is a must in any loop. As for dual pumps, it is possible, and the only way it would block flow is if the impeller detonated and you had bits of plastic blocking flow, its rare that the armature will bind. Tubing, well I've used 1/2" and 3/8" ID tubing mostly, and there is little difference in performance, its more about personal choice.

Also having the PSU to cut off when the other shows failure is the right way to go about it, this way you should not end up with melted bits as the images in that link showed.

Also Frank has a good point, but I assumed you were already aware that it will be an issue trying to go sub-ambient.

 

[Faye_Kane_girl_brain]

How are you managing condensation?

As a matter of fact, I already killed a motherboard and a CPU from condensation. I had the CPU insulated real good with silicone, but it got so cold that the other side of the MB froze.

I could have given up and run everything at stock. Instead, I sprayed the new MB on both sides with waterproofing sealant.

Then I melted a pump that was built into the waterblock of a prebuilt liquid cooler. I was all downhearted because I had destroyed my spiffy new watercooling system on the first day. I considered giving up.

But what did I do? I ordered a 3-fan radiator, and an asskick pump that I put AFTER the radiator in the loop!

I know something else will probably go wrong. In fact, I am waiting for a backup motherboard to become available next month before I throttle the TEC up to a full 24 volts.

But I HAVE TO overclock, due to a life-or death health issue: if I couldn't fuck with fast computers, I'd just DIE!

--speed freq faye

 

[Faye_Kane_girl_brain]

AIO is an All-In-One water cooler like the Corsair Hydro series coolers.

That's the brand I melted. And come to think of it, it had a two-fan radiator. So you're right, three will not be enough. I got it from frozencpu.com, which I was disappointed to find out, doesn't sell anything that will freeze a CPU.

> the CPU heat is not dumped into the loop, just the TEC heat is what the loop is cooling.

Well, the CPU heat is dumped into the loop, too, obviously. That, and the fact that the TEC is not very efficient is why TECs push out three times as much heat as they suck.

>each 120mm radiator section is worth ~150W of cooling potential.

THANK YOU!

Of course you will need a pump to make the coolant flow

Umm... yeah, I figured that already. Running a loop with no pump is about the only mistake I haven;t made. Yet.

its rare that the armature will bind.

Yeah, but isn't the impeller locked into the motor by a gear? That would block the flow unless the other pump is really good, in which case it would make the rotor in the dead one rotate, and you'd generate electricity!

having the PSU to cut off when the other fails is the right way to go

Thank you, Peet! [curtsies]

> you should not end up with melted bits as the images in that link showed.

Yeah, that guy had incredible bad luck. Talk about "anything that can go wrong, will!"

I figured he'd blow it up by not adding circuitry to shut down the other PSU when one of them fails, or by switching it off by the PSU instead of the pwr strip. But he DID order the fail-safe circuit—the power just happened to go off before it arrived in the mail!

He's lucky his house didn't burn down! When he said "I came home to a house filled with acrid smoke," it made me cringe: that smell is SOOO familiar, and it always means you lost control of a bold, brave experiment which got out of hand on its own, like Frankenstein escaping from the lab and wandering through the village killing people.

> condensation will be an issue trying to go sub-ambient.

...Not to mention, sub-freezing! Yeah, see my reply above about that. I never make the same mistake twice, but god damn if I don't make every mistake once.

-faye

 

[sneekypeet]

Yeah, but isn't the impeller locked into the motor by a gear? That would block the flow unless the other pump is really good, in which case it would make the rotor in the dead one rotate, and you'd generate electricity!

No gear, it is magnetically suspended and powered. If one pump were to die, it may slightly restrict flow more, but the impeller will still spin "freely" in the dead pump.

 

[Steevo]

All of the advice here is good, I would recommend that you use http://www.dx.com/p/dn32-hall-effec...Kve6I4Wj9c4wCPirjrj_caAqDX8P8HAQ#.U-w2G2PDXVI or something like it on the cold side plugged directly into the CPU header and turn on any features that will shutdown the system if the "fan" signal is not seen.

Wired so that the pump and all cooling is on the same PSU so no failure can occur where the TEC is left on when the pump and fans are off, and the system will shutdown if no flow is seen provides safety.

 

[Johan45]

For a high flow pump I use one of these http://www.globalindustrial.com/pro...6inOhYB890CjUtmlwZAIYaAjbm8P8HAQ&gclsrc=aw.ds submerged in a kitty litter container that doubles as my resevoir. When I first started I used it in a cooler with ice water.

 

[de.das.dude]

as steevo partly said, running a TEC isnt as simple as, say just plugging everything to make it work.
Remember TEC are industrial components and they require proper equipment like monitor of voltage and amps and safety shut off devices in case of failure.

Also remember that just like anything else its not 100% efficient. The heat out put on the hot side is always more than the heat influx from the cold side.


If you had a multimeter hooked up to the TEC you would have been able to calculate how much power its dissipating on the hot side and tally that with the dissipative capability of the cooling system and adjust the voltage for the TEC accordingly if needed be.

i think Wile_E had a horrible experience with TECs which melted everything he had.