Beyond water cooling

trilogysc

n00b
Joined
Jun 18, 2004
Messages
7
I'm heading into another computer build, and this time I want to go beyond water cooling - if I can satisfy myself (with your help) that my ideas are sound.
I'm thinking of using HydroFluoroEther (HFE from 3M, called FLUORINERT) to totally immerse my motherboard and pci/agp cards.
My questions for this thread is for anyone with experience using the liquid or for anyone interested in helping me with my design using it.

To highlite the pro's and cons as I see them:
CON'S:
There are different formulations, and choosing one without a low boiling point would be important.
Making a sealed enclosure may be difficult.
It's very expensive, about $220 dollars a gallon

PRO'S:
I would cool not only the CPU, GPU and CHIPSET (something that a well designed water cooling kit would do for about $400) but it would cool down the RAM, GRAPHICS RAM, PCI chips, the voltage regulation circuitry on the MOBO (something that most overclockers overlook) and everything else that would be heated by overclocking. Something that Vaporchill units (at $900) and watercooling can't do.

In spite of the $220 a gallon price tag, it's as cheap or cheaper than any other way of cooling that I've seen, and potentially more effective.

It would allow for auxilliary cooling of the fluid to bring temperatures down even more, (if I ever decide to go to that extreme) without having to deal with condensation issues.

Once the "case" is made, as long as the section that would pass the cables is made as replacable, this would be a case that could be used over and over again for future computers.

I don't do "upgrades" to computers, so "easy access" to the mobo is not essential to me. If there is something out there (such as a new VID card) that compells me enough to buy it, I simply surround it with a new computer - hence, no "upgrades".

Before anyone mentions Liquid Nitrogen, or Dry Ice cooling, I'm not interested in that extreme. I merely want to be able to cool EVERYTHING that needs cooling, effectively and safely. The most extreme idea I have at this point would be to cycle the fluid through a radiator installed in a freezer unit, but until then, just cooling the fluid in the radiator with fans would be sufficient.

What I've seen from those that have tried HFE's is that they attempt to go to the extremes. Fun experiment, but I'm looking at this as a long term solution. What I need to know is simple:
Is there ANY reason (aside from cost, aesthetics, ease of access, evaporation, boiling points, etc.) that HFE's canNOT be used this way?
Would the components melt away? Are there adhesives in the capacitors that would dissolve? Would the chips themselves UNlaminate? Does anyone KNOW of a good reason this can't be used as I've described its use above?

Can anyone think of any PRO's or CON's that I've missed, and should be considering?
Does anyone out there have any ideas about how to build the case, or any hints and tidbits about doing a project such as this?

My plan at this point is to use the MSI 875 board, Intel 3.2e, some OCZ 4400 Dual channel DDR and overclock to 4gig. Video card will likely be the ATI 800 (even though the 6800 is better in open GL) Subject to change, of course :D
Thanks in advance for any help.
 
There are only two slight problems I can think of.
1. You may need to make your own blocks depending on your design.
2. The pumps seals may not be compatable with the fluid.

Best design I can think of would be to use a AC unit as a fluid chiller.
Have your working fluid pass through the cold side rad and then on to your pump.
Have only the inlet side of your blocks attached to the pump via a manifold block if you can get a pump with the right head/ gpm.
Large passive heatsinks on vrm's, south bridge, psu?.

I would love to build something like that but being single handed has it limitations.

Luck........ :D
 
Just a foreword to begin. I've posted this issue on 4 other forums, and this is the FIRST place I've posted where it's apparent that people actually READ my post, UNDERSTOOD what I've said, and have posted replies that relate to my questions. I APPRECIATE IT!

Tigerbiten, to respond to your concerns:
"1.You may need to make your own blocks ....."
I have to decide on a final approach to the circulation issues of the fluid. I have a couple of methods in mind. One ( the most likely at this point) is to use either stainless or copper tubing to direct the freshly cooled HFE over the most critical components by aiming the flow right at them. That would mean making a manifold for the dispersion. You're correct about the seals in the manifold/block and pump. Some seals that would be fine with water and most other liquids would actually be permeable to the HFE's much lower viscosity, and leak out.
"2. The pumps seals may not be compatible with the fluid"
Part of that is answered above, in careful selection of the types of seals I use. As for the pump itself - it would be a magnetic drive type of pump. In other words there would be no shaft going from the impeller to the motor. Magnetic linkage would turn the impeller.

IF I go to the extent of cooling beyond ambiant temperatures, then a heat pump of some sort would be likely.

Your comment about passive heat sinks was right on the money with what I was planning. Put the heat sinks on, for greater surface area and better thermal draw from the chips, and aim the tubes right at the sinks. The flow would also help the currents to flow UNDER the chips to cool them there as well, another plus of the very low viscosity of the HFE's.

As I mentioned in my original post, the idea is to make this case with a replacable cable/thru section so that I could use the case over and over again on future builds, making a new cable/thru panel only when needed. However, as I've bragged about the advantages of the low viscosity of the fluid, something occured to me. Would the fluid actually wick up the cables (due to capillary action)? If so, I'd have to figure out a way to defeat that as well.

Bottom line SO far is that NO one can come up with a valid reason NOT to do this....yet. So it's still a go until someone puts a bump in front of me that I cannot get around.
SOMEBODY STOP ME BEFORE I DO THIS! Or bless me and get the heck outta the way :)

Cheers!
 
for cooling, maybe you could forget pumps and all that, and just have a giant heatsink apparatus that is sort of a lid to the whole thing. And half of the heatsink would be submerged, and the top half would be cooled by a fan. Don't know if that would cool it well enough though. I have no idea :p
 
a few points, generally when plumbing for flourinert, your advised to use Stainless Steel
and you need "canned" pumps, the viscosity is so low it will spring a leak and bypass rotors that while they are fine for water, are simply not up to the task of flourinert

all connections are going to be like that and substantially increase the cost
download the applicable pdfs
3M™ Fluorinert™ Electronic Liquids

Selection Guide
and fead the FAQ Flourochemicals in Heat Transfer Applications

Ive seriously investigated this, and its likely 3 times as costly to implement than your currently thinking, I cant seem to find the PIC I hoped to post of the Cray that used to be up a NCAR which was cooled with Flourinert, but will when I find it, a more complicated and beautiful web of Stainless Steel plumbing Ive never seen.

Both Fluorinert and Novec HFE fluids have very low surface tensions and viscosities. This combined with very low contact angles these fluids have on most surfaces, means that they will leak very easily through the smallest of passages. Seals which would easily hold back water at 10+ atmospheres may leak generously with fluorchemical fluid. Through shaft seals have been used successfully with fluorochemicals, magenetically coupled or canned pumps are generally preferable. These pumps are much less apt to leak

and while its compatible with say tygon tubing, again the seals are a huge issue
ideally welded Stainless Steel is recommended for long term maintenence free use.

of course this would apply to any container, like your case, where all the welds would need to be contiuous
and all connection passthroughs for wires should be above the fluid level


Id say there isnt alot of point to this exercise to simply cool to ambient
and would highly recommend including a heat exchanger directly in the case to a phase change (or better a cascade phase change) heat exchanger it being far easier to plumb and pump refrigerant
in which case and you will need to remotely locate the CMOS battery, it cant deal with cold

but it cuts your plumbing and pumping down to nill, barring possibly some way to directing the colder fluid to the right areas
which might actually be done with a few paddle wheels :p

Good Luck ;)
 
With the Flourinert, acrylic cases would work, assuming tight seams. It's apparent from your post that you HAVE researched much of this, and it's great to see your reply.
I found the links you posted a few days ago, and learned also that HFE and FLOURINERT are actually 2 different compounds, but I think you know that already.
At any rate, Flourinert is the way to go, and sealing is most important.
The pumps are easy enough to get, and you're correct about those also.
My main concerns in the beginning were:
1. Would it affect the silicon wafers
2. Would it affect the heat transfer compound (as I DO intend to put heat sinks on the CPU, GPU and CHIPSET, and perhaps on the graphics and system ram)
3. Would it affect, dissolve or be absorbed by ANY component found on a motherboard/pci-agp card
4. Would the fluid have a tendency to wick (capillary action) up the cables, and then evaporate.

At this point all of my concerns but one (Number 4) have been answered.

IF I can find a way to prevent the wicking, or find proof that it wouldn't wick at all, then the first phase of this project is a go.

Either acrylic, or stainless (not sure yet which)
Plumbed internally to aim jets of FLOURINERT directly at the most critical chips, and secondary jets to keep the fluid moving.

IF phase one is successful (No runaway fluid losses) then phase 2 is to find a way to cool the system beyond ambient - and I like your idea of immersing the heat exchange within the case.

The sticking point right now is still the cables and fluid loss. Any ideas there?
Again, thanks for your response. Oh, I've seen some pictures of the Cray, without the detail of the plumbing, and it was beauriful!
 
well consider these points

you should be able to seal any wick action in most cables
Im imagining a worse case senerio like an 80 wire round, and compared to the cost of the rest of the project, IDC connectors, a crimp, and tested sealant and a rework of a set of high quality cables if needed wouldnt be that much more

if you do go subambient, youd want a different flavor (obviously)
again if sub ambient, case insulation will make a huge difference
from closed and open cell to the latest and greatest in vaccume panels or enclosures seen in high efficency freezers and refrigerators
http://www.glacierbay.com/ultra-r.asp


my considered application wasnt an immersion setup and while I havent abandoned it, its definately postponed till there is considerably more money
(I want to cool a whole rackmount worth of components, and would be pumping the flourinert as the coolant)
 
Very nice ideas, all of you, but there is one thing I'd do differently: Why use heatsinks on the components to be cooled? If you chose, say, a fluid with a boiling point of 40°C, it would vaporize as soon as it ouched the chip, taking away thermal energy. The bubbles would rise, new fluid would come in Contact with the chip, it would boil... you know what I mean. This would eliminate the need for pumps, too.

But apart from that: A+ idea!
 
the whole objective is to keep the chip lower than 40C normally :p
I see what you say about the vaporization, but youd go through alot of fluorinert, containing the gas and reclaiming it would be pretty involved
(and youd need to control the amount of new fluid being displaced through vaporization in an open container based on the temperature of the fluid in the container, to low and convection only would occur not phase change, and since each chip is a different temperature, that wouldnt really work)

the reason for immersed heatsinks it to enlarge the heat transfer interface
and there might be much more effecient heatsink to "open" liquid convection than the typical to air convection models

personally Id use a waterblock and vaporization would only occur if the fluid temperature actually raised that high, but it wouldnt unless the pump failed but then I wasnt goin for an immersion model


what your sort of discribing is called phase change spray cooling and is better than either, also substantially more difficult, its better with water than fluorinert, but because its nearly impossible to seal the chip in the long term flourinert is used,
its akin to the sunnyD direct die cooling (Dr Surly & Spodes Abode experiments) and is described a bit here
the Marquise Project
employing diamond substrates and phase change spray cooling
SprayCooling.jpg
 
Meh. I stand by my heat-superconductive heatsink made entirely out of carbon-nanotubes idea. It would go to Alaska, and dissipate the heat there. Because it's a superconductor of heat.
 
yes, but the medium it transfering to isnt
and its interface with the chip isnt unless its bonded somehow
something like that will benefit from fractal geometry where the surface area to volume ratio can be really high, but youd still need to address the differential temperatures of your heatsink to the medium (fluid) and how that medium is being replaced (which is where phase change spray cooling would come in)

the tougher problem is the interface with the chip, there just isnt a way to bond a removable heatsink currently. The diamond substrate above is done in the manufacturing phase, and nanotube technology needs to be employed the same way (which is underway)

carbon in the right matrix is great stuff, but growing that matrix to a contigous and large enough interface isnt very easy
your heatsink would benefit greatly from phase change spray cooling if youd ever be able to connect it with a molecular bond to the point source of the heat ;)
 
not one too far off though, thats largely what they are trying to do ;)
 
To address your last paragraph,
If we had technology to manufacture carbon-nanotube products with the ease we can manufacture silicone ones today, it would be quite possible to build the entire thing out of carbon nanotubes. The semiconductive variety for the internal chip sircuitry, the superconductive variety for leads, the heat superconductive variety for the heatsink, and a resisting variety for a hard shell. Comes to think of it, you could build a, say aircraft hull completely out of nanotubes, complete with integrated high-effeciency electronics, extremely lightweight, and almost indestructible outer shell.
 
Id agree on the technology anaolgy
but Silicon fab hasnt actually scaled up to the size to do an airplane :eek: :p

that contiguous matrix issue again is the real nut to crack in either tech for larger scale
 
Thanks for the GlacierBay products. They will be VERY useful.
As for the cables, let's see if I can paint this picture.
Originally I was going to submerge the mobo entirely, with it laying on it's back, similar to a desktop (vs TOWER) arrangement.
Then it hit me,,, what if I put it vertically, as in a TOWER, but instead of the ports pointing out the back, they pointed out the top - AND that the FLUORINERT level did NOT reach the backplane, but rather it stopped just short... immersing everything on the motherboard BUT the connecting points of the cables?
Except for possible migration from evaporation, might that eliminate any wicking?
Brain's still in neutral on this, but something will click before long.
Still might come down to sealing the cables,, but that might be much more difficult than we think.
Another point, not only do the heat sinks provide more surface, but the metals are more efficient at pulling heat away from the chips than any liquid would be.
Remember, it's not the temperature of the HEATSINK that's important, but the temperature of the CHIP. Temperature probes can be very misleading when placed poorly.
 
One additional thing you'll have to worry about is wicking action of you motherboard. Most mb are some form of fiber reinforced, and the fluoroliquid will find it's way along the interface. Can't tell you without testing how badly it will swell the mb material itself without testing the specific fluid/resin combo, but what I'd worry about are local area of heat. With that low boiling a solvent, you might be delaminating the mb if the fluorocarbon reaches an area of high power dissipation and starts to form tiny gas bubbles in the material.
 
You bring up a good point. However there are several different formulations of Fluorinert.
Off the top of my head I can't recall the one I was thinking of using, but it's boiling point was around 80c, not the 50c of the #72 formula.
The reason that the #72 appeals to most is that it has a VERY LOW freezing point. Since I don't plan to go to "extremes" in cooling, I don't need that low of a freezing point. Again, assuming that phase one of this is successful, then phase 2 (cooling below ambient temps) will be the next step.
The most extreme active cooling I would desire would bring me down to only -30c, around what a phase change system can do, but this will do it on all circuits.
So the higher boiling point should alleviate any worries about radical expansion separating the laminations, and the freezing point of the formula that I'm looking at is still well below the temperatures that I'll attempt to reach.
 
Back
Top