PDA

View Full Version : Submerged Liquid Cooling



mashakos
01-02-2013, 11:20 AM
This is just beautiful:

www.youtube.com/watch?v=Z_X_hgtlJpA

This is possible using the 3M Novec 7000, a non-conductive liquid with a boiling point of 34C.
The above example uses absolutely no fans, so the only sound would probably come from the slight bubbling/boiling. Steampunk PC is here!
An engineer combined the liquid solution seen above with a radiator (no pump), allowing for high overclocks:


https://www.youtube.com/watch?v=uOgFDe7Q3Ak

https://www.youtube.com/watch?v=BUCTEFaunZc

here's the guy's website:
http://www.vaporphaze.com/

trjp
01-02-2013, 01:04 PM
Why is he wearing a coat and hat indoors? Needs less cooling?? :)

Liquid cooling PCs scares me - it's a massively silly idea and that is just bonkers...

Could you keep fish in it too? :)

mashakos
01-02-2013, 01:06 PM
Why is he wearing a coat and hat indoors? Needs less cooling?? :)

Liquid cooling PCs scares me - it's a massively silly idea and that is just bonkers...

Could you keep fish in it too? :)

here's a blade server solution for ya:
http://www.youtube.com/watch?v=N3gCavl2Y6U

what are you, a village IT guy? Scared of everything! lol

EDIT: BTW, 3M Novec 7000 is not H20...

trjp
01-02-2013, 01:11 PM
Quick Google brings up some interestingly deep technical arguments about this - such as this one

http://www.overclock.net/t/1209583/3m-novec-7000-group

Cooling servers this way is daft because you need a LOT of liquid to cool relatively little of the internal area of the server - and servers don't actually run that hot (it's more the heat they dissipate which is the issue, whether that being in the form of pressurised gas/liquid is better is arguable I guess?)

For the mentalists who want to run 4 copies of Furmark 'near silently' tho, it seems ideal :)

That said - I'm not sure saying that it doesn't corrode stuff in 48 hours is a great test - more like 2 years maybe!?

There also talk of the bubbles in the liquid pushing components away from the board - they talk about 'cheap boards' but I guess the designer wasn't planning on them being used by Captain Nemo :)

mashakos
01-02-2013, 01:16 PM
Cooling servers this way is daft because you need a LOT of liquid to cool relatively little of the internal area of the server
actually, this method expends massive amounts of heat in very small areas (500w per litre as I recall), so it's actually the opposite of what you state :P
the problem with implementing it with current server hardware is that most server motherboards are E-ATX or greater in size which necessitates a large volume of liquid to cover the entire surface area of the components. At $300 a gallon it's not a cheap solution.

as for bubbles pushing components (??) just use a horizontal rack setup rather than go vertical. Simple.

Heliocentric
01-02-2013, 01:56 PM
Ah, mashakos, don't go changing.

alms
04-02-2013, 11:51 PM
I look forward to it dropping in price so I can cool my feet down in it during hot summer nights.

CMaster
05-02-2013, 12:51 PM
That's pretty cool.
I can think of several reasons why it's probably not a great way of going about cooling your PC though.

Ernesto
05-02-2013, 04:37 PM
actually, this method expends massive amounts of heat in very small areas (500w per litre as I recall), so it's actually the opposite of what you state :P

That's not what trjp meant, I think. You would have to submerge the whole pcb to cool down a very little area on it. Therefore you need a lot of liquid.

trjp
05-02-2013, 04:43 PM
That's pretty cool.
I can think of several reasons why it's probably not a great way of going about cooling your PC though.

As per the above comment - mainly because the heat dissapation from a PC comes from 2 main TINY areas, so sticking the whole PC in fluid is bloody daft :)

I tend to think liquid cooling of all sorts is a bit bonkers tbh - it just seems like a step-too-far but this one has to have a stack of downsides other than just cost...

Sakkura
05-02-2013, 05:27 PM
As per the above comment - mainly because the heat dissapation from a PC comes from 2 main TINY areas, so sticking the whole PC in fluid is bloody daft :)
Gets rid of the whole worry about leaks though, and no pipes or radiators or pumps to muck around with. Quiet(er) too.

Still, it has its drawbacks. Plus stuff's going to be using less and less wattage, so there'll be less and less focus on fancy new cooling methods, aside from server farms and awesome casemods.

Heliocentric
05-02-2013, 05:32 PM
How do they cool things in space? By what method do they "vent" heat? Cold cathodes i suppose.

Sakkura
05-02-2013, 05:41 PM
How do they cool things in space? By what method do they "vent" heat? Cold cathodes i suppose.
Well, they don't use overclocked AMD processors, let's put it that way. You only have one kind of cooling that works in a vacuum, so you basically just have to use the spacecraft as a heatsink and then manage how heat is radiated and how light is absorbed (put a mirror between the Sun and the spacecraft = way less heat to worry about).

Heliocentric
05-02-2013, 05:47 PM
Not strictly true, a cold cathode is a lightbulb the absorbs heat from its environment and uses the energy to transmit light, hence my suggestion.

Sakkura
05-02-2013, 06:21 PM
A cold cathode draws current which generates heat.

Ernesto
05-02-2013, 07:25 PM
The only way to get rid of heat in a vacuum is to radiate it. You can actively cool a part of a satellite but then you would only shift the heat to another place and heat the whole system even more. That's why heat pipes are popular in satellite building. They don't need a power supply and therefore don't add to the heat.
Hmm... slightly OT, except you want a gaming rig in orbit ;)

Heliocentric
05-02-2013, 09:54 PM
slightly OT, except you want a gaming rig in orbit ;)
I was reading/considering about methodologies for a hobbyist probe (launching using an em rail from a weather balloon etc), but even low emission CPU's would cook out the entire system eventually, it was one of the few logical issues that was hard to solve.

Ernesto
05-02-2013, 10:29 PM
Well, heat is only one of your problems. Radiation is pretty nasty too. Random bit flips have bad effect on system stability. Although in a low earth orbit and short term missions this might not be critical.
How about a RaspberryPI? Small, light, low power consumption, GHz-class.
Anyway, have fun with your project :D

Heliocentric
05-02-2013, 10:48 PM
As pre built and robust hardware goes smart phones tech does seem like a good jumping off point. In pure nasa style costs must be minimised :p.

mashakos
06-02-2013, 02:54 AM
That's not what trjp meant, I think. You would have to submerge the whole pcb to cool down a very little area on it. Therefore you need a lot of liquid.

I know what he meant and I answered pretty clearly, at least as clearly as I could.

Current hardware, very big EATX motherboards.
Solution: use very tiny ITX motherboards for servers. Needs less of the expensive liquid.

Ernesto
06-02-2013, 10:51 AM
Hmm... whatever.
So with a boiling point of 34C it would boil constantly on a hot summer day. Ok, maybe british summers don't count ;)

mashakos
06-02-2013, 04:00 PM
Hmm... whatever.
So with a boiling point of 34C it would boil constantly on a hot summer day. Ok, maybe british summers don't count ;)

that's what the condensation part of the setup is for ;)

Sakkura
06-02-2013, 04:39 PM
that's what the condensation part of the setup is for ;)
Well that part wouldn't work with ambient temperatures above 34 degrees.

mashakos
06-02-2013, 04:47 PM
Well that part wouldn't work with ambient temperatures above 34 degrees.
The condensation apparatus, whether a radiator or a peltier condenser, absorbs the heat from the evaporated gas.
There won't be a situation where the liquid has completely evaporated and turned into a cloud of gas stuck in the radiator - at least if the setup is done right.

In any case, air cooling has a delta of 20C or higher (20 above ambient). An air cooled PC would die in a room that has a temperature of 40C (who likes to sit around in an oven anyway?).

Sakkura
06-02-2013, 04:50 PM
They don't have a peltier condenser in there as far as I can tell, it's just air cooled. So it can't go below ambient.

And saying an air cooled PC would die in 40C temperatures is just wrong.

Ernesto
06-02-2013, 05:22 PM
The condensation apparatus, whether a radiator or a peltier condenser, absorbs the heat from the evaporated gas.

Actually, the condensation apparatus doesn't absorb anything. All the energy that is transferred from the vapor to the condensation apparatus has to be transferred further to the surrounding air. But in order to do that, the ambient temperature has to be lower than the temperature of the vapor. If this is not the case the vapor gets even hotter. And that's not what we want, don't we?

Cheers!

By the way: PCs even run in india. They have to. Where else should all the excellent programmers come from :D

mashakos
06-02-2013, 05:34 PM
They don't have a peltier condenser in there as far as I can tell, it's just air cooled. So it can't go below ambient.
I'd go with a fan cooled radiator myself. The important aspect of this system is not the boiling point (the cpu temperature routinely goes above 34C) but the condensation. From experience I know that a radiator and pump loop usually cools water coolant to 10C above ambient with a fan running at 900RPM on the radiator. It goes down to 7C above ambient when the fan ramps up to 1500RPM or so. For situations where the ambient temperature is above 25C the system would have to incorporate a condensation mechanism that operates at below ambient temp. Interesting, I'll have to think about such a system that doesn't involve a peltier condenser.



And saying an air cooled PC would die in 40C temperatures is just wrong.

idle temp would be 60C, load would shoot up to 85C-90C. Wouldn't die immediately but over time...

gundato
06-02-2013, 05:35 PM
At a recent conference, I talked with a few guys who are in charge of data centers/"super computers" at various government labs (and one guy who we were pretty sure worked for Google but wasn't attending in a professional capacity) on cooling and the like.
Submersion cooling is cool (hee hee) but has severe drawbacks.

Submersion cooling in particular works REALLY well for things that generate a lot of heat and are almost never touched. Why? It has the same "problems" as liquid cooling cranked up to 11. If you have to regularly remove and inspect components (so not us, but definitely "industry"), you need to redo the seal on the tubes every time. While not a horrifying ordeal, it definitely adds to the difficulty and lowers efficiency.

As mentioned, submersion cooling cranks that up to 11 because now you either have to drain it or (if you have the right kind of connections and the right kind of liquid) reach under-liquid and fiddle with crap without making a mess (so probably still draining it :p).

So it basically boils down to ease of maintenance VS costs and more and more these facilities are being built from the ground up with cooling in mind. Over the next decade or two we might see a shift to submersion, but the big issue is the long-term viability (as has been discussed).

Of course, for the consumer it is overkill. Nifty, but overkill.

mashakos
06-02-2013, 05:36 PM
Actually, the condensation apparatus doesn't absorb anything.
I didn't elaborate but that is obviously what happens :D
Either passively or helped along with a fan.




By the way: PCs even run in india. They have to. Where else should all the excellent programmers come from :D

pretty sure Indians don't swelter in the humidity, idle temps with air conditioning are usually below 20C.

Ernesto
06-02-2013, 07:14 PM
I don't get your last reply. By idle temps you mean room temperature or CPU temperature? Because if it's a CPU temperature I don't see that happening unless cooled with a Peltier element, which is really pointless. And if it's a room temperature it's a pretty steep gradient from below 20C inside to 30C outside and I doubt that it's common in India (it'd be pointless too, imho).
To the original topic: It's certainly an eye catcher. And that's all it is. Because it's also very impractical.

Sakkura
06-02-2013, 07:14 PM
idle temp would be 60C, load would shoot up to 85C-90C. Wouldn't die immediately but over time...
Psh, with a good air cooler you don't get a 20C delta at idle, nor a 45-50C delta at load. At least not with a typical CPU (an overclocked FX 8350 might be a problem).

mashakos
06-02-2013, 08:01 PM
I don't get your last reply. By idle temps you mean room temperature or CPU temperature? Because if it's a CPU temperature I don't see that happening unless cooled with a Peltier element, which is really pointless. And if it's a room temperature it's a pretty steep gradient from below 20C inside to 30C outside and I doubt that it's common in India (it'd be pointless too, imho).
To the original topic: It's certainly an eye catcher. And that's all it is. Because it's also very impractical.

oh sorry, I meant room temperature. Can be much lower than 20C with central air cooling (in the range of 13C)

mashakos
06-02-2013, 08:04 PM
Psh, with a good air cooler you don't get a 20C delta at idle, nor a 45-50C delta at load. At least not with a typical CPU (an overclocked FX 8350 might be a problem).

well you're not going to need fancy cooling on an i3 or atom processor so we're talking about cpu's with 125W TDP or higher (which end up in the 180W + range with overclocking)

MiniMatt
08-02-2013, 01:59 PM
I vaguely recall many many years back seeing a Cyrix 6x86 running in veggie oil.

Can't help but think that the age of massive cooling is coming to an end however. Massively clocked 180w CPUs only need substantial cooling because they're turning 90 odd of those watts not into pretty pixels but heat. A (physically impossible) 100% efficient CPU could have no heatsink/cooling whatsoever and run full pelt at ambient room temperature.

Progress in the last few years and what looks to be the trend for the next few years looks significantly skewed toward efficiency. So in the future you'll still have 180 watt CPUs, but 170 of those might be used for explosions and only 10 wasted as heat. At that stage any old passive heatsink off a 386 chip may well prove sufficient cooling.

bigc90210
26-02-2013, 12:32 PM
I vaguely recall many many years back seeing a Cyrix 6x86 running in veggie oil.

Can't help but think that the age of massive cooling is coming to an end however. Massively clocked 180w CPUs only need substantial cooling because they're turning 90 odd of those watts not into pretty pixels but heat. A (physically impossible) 100% efficient CPU could have no heatsink/cooling whatsoever and run full pelt at ambient room temperature.

Progress in the last few years and what looks to be the trend for the next few years looks significantly skewed toward efficiency. So in the future you'll still have 180 watt CPUs, but 170 of those might be used for explosions and only 10 wasted as heat. At that stage any old passive heatsink off a 386 chip may well prove sufficient cooling.

Cooling with oil is funky by by no means a feasible method (unless you like replacing your hardware... frequently). The oil causes the tiny rubber seals on the caps and the likes to swell and pop. all of the oil submersion project logs on bit-tech are dogged with issues of rubber seals going on motherboards over time and popping the motherboard / other components. Its a good proof of concept, though its not a long term solution for a home user by any means