Cooling Technologies & Ideas (Thread Split)

[Tesselator]

This is a side topic we got into from another thread split out for us by the awesome WildCowboy - a mod here [Thanks].

The discussion includes cooling tech for Workstations, servers, and overclocked systems. It kida started weird (as most OT posts do) with the quote below but got a little interesting later on.

http://www.youtube.com/watch?v=ctI3YZHbWX0 Engineered for future expandability with MB replacement. Customers are taken care of too!

Feel free to jump in on any point. Also I think that keeping the jargon generic or explaining technical terms and issues will allow more people to participate.

 

[Dr.Pants]

http://www.youtube.com/watch?v=ctI3YZHbWX0 Engineered for future expandability with MB replacement

Liquid-cooling-system?! Wasn't that the bane of the G5?

 

[Tesselator]

Liquid-cooling-system?! Wasn't that the bane of the G5?

I guess so. Beats me. But I saw a photo of Apple's WC system and about busted a gut laughing. I saw Plumbers' Tape (hose clamps) on all the connections instead of proper fittings. So that's just pure incompetence on Apple's part. Anyone who's ever done any gardening can tell you that stuff actually causes leaks over time!

NOT FOR USE IN COMPUTERS YOU CARE ABOUT! ​

 

[nanofrog]

I guess so. Beats me. But I saw a photo of Apple's WC system and about busted a gut laughing. I saw Plumbers' Tape (hose clamps) on all the connections instead of proper fittings. So that's just pure incompetence on Apple's part. Anyone who's ever done any gardening can tell you that stuff actually causes leaks over time!

They used those stupid spring hose clamps. These things do leak (becoming more likely if reused), and are a PITA to mess with (busted knuckles in tight places). Then add in the fact the coolant chemically dissolved/ate the metals used. Apparently a rather bad design/implementation on Apple's part.

 

[Dr.Pants]

They used those stupid spring hose clamps. These things do leak (becoming more likely if reused), and are a PITA to mess with (busted knuckles in tight places). Then add in the fact the coolant chemically dissolved/ate the metals used. Apparently a rather bad design/implementation on Apple's part.

Bad Apple relaying their LCS system to a car company!!!! Still need to replace my coolent, I'm being lazy.

However, aren't a load of "speciality" liquid coolents generally conductive/corrosive as well as the particularily harsh version in the G5 LCS?

 

[Tesselator]

They used those stupid spring hose clamps. These things do leak (becoming more likely if reused), and are a PITA to mess with (busted knuckles in tight places). Then add in the fact the coolant chemically dissolved/ate the metals used. Apparently a rather bad design/implementation on Apple's part.

Ah, right... Spring clamps... I remembered it was something dumb and cheapskate-ish.

And what's this? Coolant/material reaction/corrosion? LOL!!! Really? Man that takes it. LOL!!!

Anyway, liquid cooling if done right is a very good thing - not something to be leery of.

 

[nanofrog]

However, aren't a load of "specialty" liquid coolants generally conductive/corrosive as well as the particularly harsh version in the G5 LCS?

AFIAK, most of it's ethylene glycol based. It just needs to be buffered to help with corrosion of the metal, but without it, it gets ugly. It certainly did with automobile radiators in days past, as I recall a few my dad swapped out when I was a kid. Nowhere near the frequency now as then.

Silicon oil is an option, as it's chemically inert and electrically non conductive (for immersion methods), but is more expensive. Though not drastically so.

Anyway, liquid cooling if done right is a very good thing - not something to be leery of.

Yup. Generally it's quite safe, as well as efficient at moving heat off the components.

 

[Tesselator]

Silicon oil is an option, as it's chemically inert and electrically non conductive (for immersion methods), but is more expensive. Though not drastically so.

750ml of it goes for $5.00 here. I use it to make my electric and acoustical guitar collection shine pretty. It's clear though. If you want colored you'd have to add pigment or dye of some kind. Mmmm violet!

 

[nanofrog]

750ml of it goes for $5.00 here. I use it to make my electric and acoustical guitar collection shine pretty. It's clear though. If you want colored you'd have to add pigment or dye of some kind. Mmmm violet!

About the same as here then. I located some at $17USD for a gallon not too long ago. Of course, that didn't include shipping.

 

[Tesselator]

About the same as here then. I located some at $17USD for a gallon not too long ago. Of course, that didn't include shipping.

I wanna find something that's even lighter though. I wonder what's lighter and still non-conductive? These are the times when I wished I had have paid more attention in chem classes.

 

[nanofrog]

I wanna find something that's even lighter though. I wonder what's lighter and still non-conductive? These are the times when I wished I had have paid more attention in chem classes.

What was the viscosity?

I'm used to stuff that's close to water, not mineral oil.

 

[Tesselator]

What was the viscosity?

I'm used to stuff that's close to water, not mineral oil.

Yeah, it's about like water. Pretty close I think. It reminds my of 3-in-1 oil or gun oil.

I was thinking of something more the consistency of contact cleaner or diethyl ether.

 

[nanofrog]

Yeah, it's about like water. Pretty close I think. It reminds my of 3-in-1 oil or gun oil.

I was thinking of something more the consistency of contact cleaner or diethyl ether.

You're NuTZ.

Too low a boiling point so it's not good for heat transfer, and highly flammable!

I can see the label now... WARNING: COMPUTER MAY EXPLODE... And many posts about the smell before its poor imitation of a Phoenix. Gets the rapid burn part right, just not the resurrection.

 

[Dr.Pants]

Low boiling point is a pretty large factor - sure, as vapour pressure increases the gas/liquid equalibrium gets pushed back towards liquid - but lets just leave that for sturdy metal pipes and not cheap hose clamps Not to mention thermal expansion of a material is also another factor - considering how liquids do not compress well/at all, and if the expansion coefficient is high enough....

Let's just say the computer should be on the floor in all circumstances

 

[Tesselator]

Well, I didn't mean to actually use diethyl ether. Just something more toward that consistency - yet more stable.

And I would think the heat exchange would by MUCH more rapid, no? Think of ammonia in refrigerators. I know they're using a compression -> evaporation cycle to produce cold temperatures but while it's a liquid doesn't it shed it's heat (thermal exchange) very rapidly? It seems the radiator in a computer's liquid cooling systems would be enough to cool it back down. And I don't think ammonia needs much pressure to remain a liquid - the expansion valve in cheaper units is just a pinch in the pipe.

Think of the liquid nitrogen coolers there are now. I think they work because the liquid is so light that heat exchange is extremely rapid. Now, most of the systems I've seen run it open and allow the evaporation itself to cool the liquid - but it wouldn't require much pressure to force it to retain it's liquid form and just run it through an air-cooled radiator. Right? I would have to look it up but I imagine it would be right around a gram per square centimeter - or less. About like a party balloon on it's death-bed the day after the party.

It requires a tremendous amount of pressure to liquify these gasses but many don't require much at all to maintain their liquid state.

So I guess it would be no more pressure than the warm water is already causing in the water pumps today?

But I don't know very much about this topic. I'm left with only guesses till I go look something up - which hasn't happened yet. I probably am NuTz - I just haven proven it to myself yet - or haven't been shown the proof.


In any case something a little lighter than water and less viscous than silicon oil would be nice.


.

 

[DoFoT9]

im afraid your nuts tess...

 

[TheSilencer]

Liquid cooling ftw! Just look at the overclocking world record database.
Overclocking World Records, scroll down for the Xeon

LN2 just seems a bit extreme, tho the Peltier thermo electric cooling times were also and I enjoyed that, at least at the PC side.

 

[nanofrog]

And I would think the heat exchange would by MUCH more rapid, no? Think of ammonia in refrigerators. I know they're using a compression -> evaporation cycle to produce cold temperatures but while it's a liquid doesn't it shed it's heat (thermal exchange) very rapidly? It seems the radiator in a computer's liquid cooling systems would be enough to cool it back down. And I don't think ammonia needs much pressure to remain a liquid - the expansion valve in cheaper units is just a pinch in the pipe.

Think of the liquid nitrogen coolers there are now. I think they work because the liquid is so light that heat exchange is extremely rapid. Now, most of the systems I've seen run it open and allow the evaporation itself to cool the liquid - but it wouldn't require much pressure to force it to retain it's liquid form and just run it through an air-cooled radiator. Right? I would have to look it up but I imagine it would be right around a gram per square centimeter - or less. About like a party balloon on it's death-bed the day after the party.

It requires a tremendous amount of pressure to liquefy these gasses but many don't require much at all to maintain their liquid state.

So I guess it would be no more pressure than the warm water is already causing in the water pumps today?

But I don't know very much about this topic. I'm left with only guesses till I go look something up - which hasn't happened yet. I probably am NuTz - I just haven proven it to myself yet - or haven't been shown the proof.


In any case something a little lighter than water and less viscous than silicon oil would be nice.


.

Two stage cooling systems. I've seen them, but they're considerably more expensive than what we use now. I'd think you're familiar with them as well, as most I'm aware of are used to cool lasers (basically a rackmount 134a A/C system).

Hmm... two stage nitrogen cooler. Imagine the evilness I can get into... //aquiring Grinch like grin, rubs palms, and sprouts horns...

 

[Tesselator]

Two stage cooling systems. I've seen them, but they're considerably more expensive than what we use now. I'd think you're familiar with them as well, as most I'm aware of are used to cool lasers (basically a rackmount 134a A/C system).

Hmm... two stage nitrogen cooler. Imagine the evilness I can get into... //aquiring Grinch like grin, rubs palms, and sprouts horns...

Heheh who was it in the TV series The Adams Family who was always blowing themselves up? Uncle Fester no? Hehehe...

I know! We could use hair tonic! At least no one would complain about the smell on a spill. J/K of course but there's got to be something that's light, inert, electrically non-conductive, and with excellent thermal exchange properties - 2 or 3 time that of water. <shrug>

 

[TheSilencer]

Heheh who was it in the TV series The Adams Family who was always blowing themselves up? Uncle Fester no? Hehehe...

I know! We could use hair tonic! At least no one would complain about the smell on a spill. J/K of course but there's got to be something that's light, inert, electrically non-conductive, and with excellent thermal exchange properties - 2 or 3 time that of water. <shrug>

How about oil? It sure is kinda ... uhm ... crazy ... but works. You can chill the oil way under freezing point of water with thermo electric cooling or cooling compressor and submerge virtually the whole computer.

Youtube, build mineralic oil cooled PC

 

[Tesselator]

How about oil? It sure is kinda ... uhm ... crazy ... but works. You can chill the oil way under freezing point of water with thermo electric cooling or cooling compressor and submerge virtually the whole computer.

Youtube, build mineralic oil cooled PC

Thanks, I hadn't seen that one yet. I saw a total submersion in salad oil on Tom's Hardware tho. Submersion isn't convenient tho, as I'm sure anyone can see.

The problem with oil of any kind (in a pump & radiator setup in place of water) is that while it requires a lot of energy to heat up it also requires a lot to cool back off making it slow in both directions of the thermal exchange. It's why water or still water is better and more commonly used (I think). I just want to go one better than water is all. Also with the right properties we could flow it across the HSI directly (or use very thin copper foil) and do away with big heavy copper blocks.

Monitor the actual core temp of your procs and you'll notice the the temp increases almost immediately in ratio with usage activity (%). With a very thin liquid much closer to the silicon this could likely be compensated (cooled) more rapidly and potentially allow higher overclocking.

Anyway, it was just an idea...

 

[TheSilencer]

Thanks, I hadn't seen that one yet. I saw a total submersion in salad oil on Tom's Hardware tho. Submersion isn't convenient tho, as I'm sure anyone can see.

The problem with oil of any kind...
.
...Also with the right properties we could flow it across the HSI directly (or use very thin copper foil) and do away with big heavy copper blocks.

Anyway, it was just an idea...

Agreed, oil cooling sure is not really the number one choice.

Hm, big heavy copper blocks, how about silver? It does transfer heat better!

Very surprised the Quad 2.66 only cost about $300. Wonder what the C2D costs? Perhaps $150 for the 2.26 MacBook processor. If yes, then Apple could offer a better computer across all price points. They could do away with the low end models, but they have us, at least me, trained to their product differentiation and their price jumps (of about $300) to move up to a better model. But with the Pros is bigger price jump of $500.

The remarkable thing of this table is the growing "side cost" margin while the machine is the same type, same motherboard, same CPU board, same RAM, same case, same heatsinks...Just everything is the same except for the jump from single-CPU to dual-CPU but even within these the "side costs" grow for no reason.

The other thing is that in 2008 they offered also the high-end 3.2GHz while we stuck today with the 2.93GHz, not with the 3.2GHz high-end Gainestown chip for dual-CPU systems and instead of the 3.2GHz/3.33GHz high-end Bloomfield single-CPU chips we only have up to 2.93GHz.
At least they didn´t start with the 1.86GHz Gainestown at the same price.

 

[nanofrog]

Thanks, I hadn't seen that one yet. I saw a total submersion in salad oil on Tom's Hardware tho. Submersion isn't convenient tho, as I'm sure anyone can see.

The problem with oil of any kind (in a pump & radiator setup in place of water) is that while it requires a lot of energy to heat up it also requires a lot to cool back off making it slow in both directions of the thermal exchange. It's why water or still water is better and more commonly used (I think). I just want to go one better than water is all. Also with the right properties we could flow it across the HSI directly (or use very thin copper foil) and do away with big heavy copper blocks.

Monitor the actual core temp of your procs and you'll notice the the temp increases almost immediately in ratio with usage activity (%). With a very thin liquid much closer to the silicon this could likely be compensated (cooled) more rapidly and potentially allow higher overclocking.

Anyway, it was just an idea...

What I recall with the silicon oil, was what Cray did. Full immersion, and an exchange system (massive A/C system). Quite efficient cooling wise, but not cheap, including the extra electricity. Not by the concept, but due to the size of the system.

A small scale system that would be needed wouldn't be that expensive, when you figure in economy of scale (around the upper end of current water cooling costs?).

As for thick blocks, I was under the impression it had as much, or even more to do with the machining techniques (CNC mill) on copper. Get too thin, and the bits tear through (too little metal left to handle the friction & lateral forces).

If enough quantity are produced for cost reasons, perhaps precision die cast would do the trick?

In the mean time, current water cooling systems will have to suffice, and swap out the coolant as you're thinking. I'd have to go back and look, but I do think currenly available silicon oil would be the best ATM. Otherwise, something else would already exist, or at least being worked on (coolant). It's not like the thermal issues with semiconductors is new or anything...

 

[thepawn]

What I recall with the silicon oil, was what Cray did. Full immersion, and an exchange system (massive A/C system).

Cray used Flourinert. http://en.wikipedia.org/wiki/Fluorinert

 

[nanofrog]

Cray used Flourinert. http://en.wikipedia.org/wiki/Fluorinert

For some strange reason, I was thinking silicon oil. Good catch, BTW.