LIQUID SUBMERSION - a few helpful learned facts.

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.
4voc,
I'm not trying to undermine what your trying to accomplish. It's just a little more complicated. I was a Refrigeration specialist for over 10 yrs., both in the military and in the civilian arena. I have knowledge in convection, liquid, and Phase Change refrigeration systems. I also know both passive and active cooling, and try to keep up to date in Peltier systems (which ain't easy).

Liquid cooling is only viable/effective in large scale, fairly smooth areas. Believe it or not, most large malls use it. Thats what the fountain is for, it is the heat exchange to cool the water before it is pumped back the A/C system (though it is often supplemented with a secondary system). Only gases have the nessesarily low viscosity to properly cool a small irregularly shaped object, like a mobo.

Listen, I'm not trying to negate what you're attempting. I'm simply passing on information to help explain why it's not working as well as you thought it might. I've even tried to supply some additional info that you can consider and/or research at your leisure, if you decide to try again in the future. Heck, you could be the next genius to come up with a truly viable, efficent solution to cooling and retire rich and famous. It's not entirely out of the realm of possibility.. 😀
 
Rripperr, no worries man. I agree with your comments. I want to either/help people from wasting time doing something that will waste them time and be frustrating... or find possible solutions as I can't seem to find one yet that is feasable. I get frustrated sometimes on these types of post 'cause you get alot of Two Cents which are a. completely wrong and often damaging to the person if they follow that advice. b. just repeat of something that has already been said.
You are right that submersion is used all the time in industry and note that the liquid I used is a common transformer coolant (one of the replacements used for PCBs!!!)
You are also correct that if a PC board and it's Processor and other Chips only had large smooth surfaces, this would be a very easy project and conversion. The problem that I can't overcome at this time is that... and the problem that I want others to understand before they waste their time and money on is.... The board and it's chips aren't purely generating heat on their surfaces but also behind and around them in the tiny spaces and that liquids will insulate these areas not cool them even when you go overboard with recirculation and cooling. Keep in mind that if you cool and of the liquids we have discussed, say to zero F, they will become more viscous and at least on a micron-local level move much less if any from these spaces. Thereby you'll have even less heat transfer to these areas and a greater temperature spread across your chip and I would theorize that you could even crack your chips from the temperature changes. IE, but ice in a hot glass or a hot glass in the freezer, Or even spray your CPU Water block with canned air (if it has the typical plastic cover) while it's running or just been turned off.
Note to the wary: if you do use a chilled coolant for your water blocks, make sure your pump starts and stops after the pc powers up and down and preferably use completely metal blocks with no covers. You will shatter the plastics.

rcs2749: heatsinks where used, and tried without. The heatsinks work better in the submersion as they extend out the heat into the liquids and create more contact area. But, overall I never had a problem keeping the surfaces of the chips cool (you can tell by using a temp.gun and/or thermocouple probes) My company and shop has alot of cool test equipment so I could use the temperature probes to see where the heat was in most cases and adjust accordingly)
 
It seems to me the matter that is bothering you the most is the micro-"deadzone" pockets where the fluid itself is not circulating, yet these pockets are typically not the primary surface for disipating heat, although they do require some disipation.

My thought is, even without any movement in the fluid, the sheer volume and contact areas should be able to the required transfer heat much like a solid, and overcome its lower conductivity. Even if you were to have deadzones between pins or circuit boards and the chips themselves, the chips should be able to conduct (via their electrical contacts or otherwise) the excess heat through the area that is normally exposed to air, and far better when submerged.

I still think there is another problem at hand, perhaps its minutely shorting out.

Edit: Oh, and btw, the person I spoke about before, he is using synthetic motor oil in his P3-800(linux samba/router).
 
A truly unusual cooling method that would be viable (if somewhat difficult) would be to hermitcally seal the case, and pump 'moist' cool air through it. Moist air, around 40-60% humidity, can actually absorb a fantastic amount of heat fairly quickly. The only problem (besides accelerated oxidation in ferrous and copper components) is the evaporator usually condenses the water out of the air, which means you have to find some way to add it back in before it enters the case again (with out adding heat). It's something I've been thinking about off and on for the last 3 years.

Unfortunately, I haven't been able to arrive at a convenient solution. It woud probably require something just sightly smaller than a window unit A/C. Like maybe one of those small dorm 'fridges. Hmmm.. Duct the air in and out. Air goes in the top (near the 'freezer' section), take the air out near the bottom. Keep a 'pan' of water in to raise humidity levels.. Hmmm.. I'll have to think about this.
 
Rripper, a much easier method would be to take a lead from heavy electronics back in the late 70's/early 80s in submersibles(radar/sonar, etc) where they would make precision aluminum blocks that would mount between the component and their chasis. The inside of the chasis was ribbed for more contact area with the air inside. The chasis would then disipate all the heat.

Its kinda like the Zalman TNN systems, only using solid blocks to connect the hotspots to the heat disipating chasis, and they were sealed off.

That would be similar to your sealed off case, and without the moisture problems. I do not know what kind of "air" they had inside, but I'ld imagine it was just normal filtered air.
 
Rripperr: The misting idea is a good idea in that the evaporation of the water actually pulls heat out of air. This would be like a swamp cooler or misting tower which you should be familiar with. Bad news is. Water would never be a good option and using light coolants would shortly cause the same problems I had because it would condense and drip down behind the boards.

Praeses: I've been trying a bit of better block designing for heat dissipitation including different gas or water flows through them and arrangements to pull heat from the hot spots, I'm just a bit limited on time because I'm supposed to be running a company not playing. :) I have a few other ideas that I'll try soon so I'll keep you posted. Oh, and those finned Al blocks are called Heatsinks... which are alll over the boards already. :)
Praeses #2: Pulling heat through circuits, to then dissipitate is heating the cooled spots. As electrically everything was better then normal with this board (p5n32sli boards, first editions, where notoriously voltage flawed) and I could see the heat spots behind the chips. I think that primary or not, liquid behind the chips doesn't fly. My surface temps of the chips where steady at just above ambient, but the chip cores where ramping like a skateboarder on chrystal. Therefore, again, I'm pretty confident that these tiny spaces are the problem. In fact, another bit of evidence that I just thought of this moment is a phenomenon I've seen alot of , peoples OC'd boards with adequate surface cooling. I've noticed alot of boards get slight heat warping behind the processor socket when OC'd even when the chip itself is cooled with a highend waterblock. Normally these spots would not be a big deal, but when OC'ing they do become a big deal. A processor has alot of tiny molecule sized circuits in it, and if you have to transfer all the heat from one side to the other to keep it cool...your adding to the problem not fixing it as I think I've shown. Which really kinda sucks because, the liquid submersion fixes this problem for most of the board and it's components, just not some of the most critical spots where it makes it worse.
 
I just pulled a p3 2000mhz chip off an old board. Night and day. Its 4 times the size with 1/8 the cooling surface for the chip. The spaces are much larger and the pins are located mostly far away from the part of the chip that generates the most heat. Additionally, it's using (guestimate 'cause it's not worth looking up) about 1/10 the KW (without OC'n)in a space that's 4times larger. AND. The pins extend THROUGH the motherboard. They don't on most the newer boards, the processor pins are in a seperate socket, which is then connected through seperate pins into the board. Night and Day.
 
Do you think you would have had the same problem if you had tried with a socket 478 based board ?

or about the same question, do you think the pinless/holeless socket 775 exacerbates the problem ?
 
of course you ask that after I've left my office where I have a couple socket 478s on Dell boards...

So I pulled a few picks up off yahoo and it looks like they show it being in front of the board not integrated into the board. So, I did see a few ways it is setup in the pics but some look somewhat similar albeit a little less crowded then the 775 I used. I would assume though that you may have the same problem, BUT that it would be less, Keep in mind. The PENTIUM D Chips seem to run ALOT hotter and USE ALOT more energy then the chips I have in my P4-3000mhz socket 478 Dells.

So, just making an educated guess. Same problem would happen, but as it is using less electricity in the same amount of space with less crowding and less heat generation, it may be still a problem but not nearly as bad. I don't think these couple generation old Dell units can unlock and overclock the Processors easily, but I'm not sure. And honestly, they're good enough for work use so It's not worth trying it for me right now.


FYI, IF there are any technically savy readers with free time and trustworthy (you would have to undergo a security and drug test, and you are located in the Los Angeles Area I would consider letting you repeat/modify/update/fix/make better this project with my materials and document it for publishing to an online site or two.
 
You know, I checked out some pics too, and there may not be an easy solution to cooling underneath the 478 and newer CPUs. The seal and tolerances required for them to operate with the newer connection brackets nessesitates that the zone directly under the chips will be a "dead zone". Ventilating properly would require drilling holes THROUGH the mobo (without hitting any 'traces or xsistors)under the CPUs. Even then, depending on how much PCB you removed, it still might not be enough. :evil:
 
Well, you could try turning the case into one large submersible pump. You could use a siphon at the top, sealed with silicone, along with a vaccuum apparatus to suck the fluid out, which could then go to some sort of radiator, then back into the case. The "pump" would effectively be cooling itself, and the fluid moving inside the case would help dissipate any hot spots.

Just a thought.
 
hmm, you don't seem to read very well. there was a reason why I only tought college chem for one semester... at least i can fire my employees now.

Dude, read at least a few of the posts before you make a suggestion... .... a suggestion to do something was done from the start and discussed in multiple posts.

rripperr, I think, if you could do it successfully, it possibly could help... if someone wants to play with my setup rigs, they're more then welcome, I've got another idea for a new setup with some new hi-tech materials that I've got samples coming in on. This would be a bit easier then complete submersion and if it works would be marketable. I'll be trying it out within a few weeks on my core2/intel975x setup that I'll have tomorrow.
 
Hi,

I think your are asking too much from the current chips designs, they tend to dissipate the heat trough the top metal surface of the heat spreader and not from the bottom, which is thermicaliy insulated, the hot spots revealed by your infrared probe are heat build up inside materials with low thermical conductivity and not on theirs surfaces such as the inside the PCB, chip socket, etc.
The only way to avoid that is to change the chips design and there are some ideas such as:
http://www.zurich.ibm.com/news/06/cooling.html
 
Nice Link Mad guru. Thanks, I did find another few "future" materials used by a couple of my aerospace customers that may help me address the back of the chip a bit better. As far as the front goes though, this is a pretty article. I'll take a stab of making a slightly less intense version version of this.

And yes, I think I am asking too much out of the current chip/socket/mb design. :) but, i solve technical problems for a living so... :) doesn't mean I won't keep trying ... as long as i'm interested. or someones paying me...
 
jesus, just forget about submerging your computer, and just water cool it.
I could never imagine doing that, its just to messy, with little or no gain over a normal water cooling setup.
 
you called?? :twisted:

Of course, and let's keep bush as president, stick with a petroleum based economy, develop our parks into stripmalls, ......

hmm. didn't you learn to oc your p805 from some innovative person smart enough to look into it's mfg quirks, see possibility and test out it's oc abilities?

jesus has nothing to do with innovation... go shop at the mall and stay off the tech forums......
 
In regards to using fluorinert, 3M will try to dissuade you from buying fluorinert (they were the original manufacturer). They do produce other perfluorinate hydrocarbons (excuse the spelling) which have much lower gel temperatures and would therefore wouldn't thicken up till much lower temps. These compounds are also much cheaper than fluorinert, with all of the good properties preserved.
 
ooooohhhhhhh.......... sh*t
So tell me what ARE your "GOALS" in this project?
Tell me your GOALS and I tell you if they have any real world ADVANTAGE

Never mind I guess, this project was just for some weekend fun
 
I'm new to these forums here, as I was referred from other forums to find the thread author and his submersion warnings.

I had a question: since the whole case that will be submerged is not very large, does it make sense to:

a.) fill with some appropriate liquid (I was leaning towards something that I don't have to replace/refill every few months, i.e. not oils).
b.) fit fans that direct flow in some direction
- have some directional motion that circles the cool liquid around the case, and have the warmed liquid return to the original spot (** like a loop in water cooling).
c.) have some sort of cooling system, like 500W Peltiers, to cool the liquid.

That way, your liquid keeps circulating, and the probability that molecules remain in one single location is reduced greatly.

Does this work? Or is the official conclusion (Shout to the world) that submersion is not meant for computers?
 
The biggest difficulty with liquid submesrion is not that it doesn't work, it's that current hardware isn't geared towards it: flowing the liquid about is, as we could see, not always efficient, due to some liquid being sometimes trapped in a hot place anyway, and reaching boiling point. Were motherboards and such created for submesrion, with well sealed interstices and smooth surface, then submersion would indeed be successful; but right now, not only does it require exotic products to work on a long term, it requires very delicate hardware handling - and some boards are better at it than others.
 
Hi,

I think your are asking too much from the current chips designs, they tend to dissipate the heat trough the top metal surface of the heat spreader and not from the bottom, which is thermicaliy insulated, the hot spots revealed by your infrared probe are heat build up inside materials with low thermical conductivity and not on theirs surfaces such as the inside the PCB, chip socket, etc.
The only way to avoid that is to change the chips design and there are some ideas such as:
http://www.zurich.ibm.com/news/06/cooling.html
Click to expand...

Just a thought to trying to cool and get coolant movement under the bottom side of the cpu. It would be interesting if the cpu's "prongs" were like 2or3 inches long and it was suspended above the socket. . Then coolant or whatever your using could move under it all around it and even between the prongs

This was just a thought and I have no expertise in any aspect of what your trying to accomplish.
 
Howdy,

Fascinating post. Just throwing stuff out, but could possibly adding something that vibrates the MB shake loose the pockets of stuck oil and help to move it along. Or some high rate of bubbles, something to collide with the particles. OK, I know your talking the width of a hair and that air bubbles are bad, but how cool would it be to have those deep sea divers and treasure chests that blow bubbles in your computer cases!

I just write code, so I have no idea if any of this makes any sense. Yall are on a level I’m not even going to imagine to be at.

Keep up the posts…very cool stuff indeed.

Doody
 
If your rich you can do this and have it work nicely, get a few gallons of 3Ms phase changing coolant Perfluorohexane (prohibitively expensive for those without at least $20,000 burning a hole in your pocket). Modify your motherboard so the capacitors are covered with water tight shrink wrap if they are electrolytic.
Create an air tight case that has a place to suck air out of and a place at the bottom to pump the coolant back in and I/O ports and put your computer and the Perfluorohexane into that sealing it. attach a pump and condenser and cooling track to that and start up your computer.
It will probably pull quite a bit of energy and make sure to test the condenser and cooling system for leaks first as that stuff is slightly toxic. Overclock your new Tesla Super Micro that cost you a grand total of $15000 to buy and set up and about $400 a month to run and program a new game for it and play like a king. Dont forget to buy the fiber optics throughput and have your super frag fest with a ping of 0-2.
better yet use it for grafix rendering and rent it out or use it as your grafic design platform. or SL programing and scripting platform.
[strike] ???[/strike]
PROFIT!!!!
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom