Slashdot Mirror
CPU Cooling Insanity
moonboy
writes "I saw this over at Ars Technica.
This dude submerged his entire motherboard in mineral
oil. As if that weren't enough, he then and got a
5,000 BTU (window?) unit and circulates the oil through the
coils to keep it all cool." Don't expect Gateway to be
offering these any time soon... I suspect it will a bit
more than just void your warranty. It'll probably make
motherboard engineers come to your home under cover of
darkness carrying loaded shotguns :)
Why would it be a problem if the power supply or the hard disk fell in.
If you take for granted that the oil conducts as much as air, you can submerse anything in it you want.
The hard disk is sealed anyway, so no oil could get in there.
And as far as the other parts go, everything that can be exposed to air should be able to be exposed to oil.
I saw another post of someone worrying about a thin layer of oil insulating the jumpers.
Well, for the same part you could worry about a thin layer of air insulating jumpers in a regular setup.
This is really a brilliant idea, since the oil conducts heat much better then air. Maybe one day we'll all have liquid submerged pc's.
Actually, pc-makers would like this, because you'd have to buy the liquid from them, and they could make it impossible to get hold of the liquid unless you're oem, therefore forcing you to return your pc for any upgrade. Finally a way to control us completely, even better then intel's not-so-secret pIII tag.
The electronics on some deep sea submarines are encased in mineral oil. Though immersing electronics in oil might sound strange, it is not an uncommon practice. Here are some research papers on the topic:
http://tdei.sju.edu/tdei/index/eiidx.html
--------------
A Dylan hacker
Raleigh, NC -- When Joe Celery upgraded his home computer system, he ended up with a machine that was hotter than he bargained for.
Joe Celery was a member of an elite group of computer enthusiasts that call themselves the "3l1t3 0v3rCl0cK3rz Gu1Ld". The group's focus is a technique called "overclocking", which is a technical term for running a system faster than it is rated. Some overclockers do it to save money, but many, like Joe Celery, overclock for fame.
"I started out with a [Intel Corp. Celeron(tm)] 450A [300A processor overclocked to 450MHz," Mr. Celery stated in an exclusive interview from his hospital ward. "Sure, it ran Quake II real fast, but that wasn't enough. So then I tried watercooling."
Using approximately $500 worth of supplies, including copper tubing and a small air conditioner, he managed to get his computer to run as fast as one with a Pentium II chip that cost almost $300 more than his Celeron(tm) processor. It would process millions of instructions per second for as long as an hour before crashing.
He continued, "it was an incredible success. But I just wasn't satisfied. I wanted more." His next exploit involved a stack of Peltier cooling elements, flat devices that electrically transfer heat from one surface to another. The Peltier elements allowed him to crank his watercooling setup another 10%, to match the performance of a Pentium II chip costing $500 more than a stock Celeron. It used only $300 worth of Peltier elements, as well as the original watercooling apparatus.
Celery was markedly silent on how, exactly, he ended up in the hospital. After half an hour of prodding, he finally admitted what his latest creation was: "I tried to cool my system with hydrogen gas. It worked, until my hard drive spun up."
The resultant explosion caused approximately $15,000 worth of damage to Celery's neighbors' homes, notwithstanding the destruction of his own home. Analysts estimate the amount of damage was greater than the damage possible if a Pentium III Xeon chip costing $1000 more than Joe Celery's Celeron(tm) chip was used.
Celery left us with this final comment: "my next computer is going to be a Macintosh."
Speaking of liquid nitrogen making things brittle and shatter, once I was in a chemistry lab late one night when an evil cockroach happend to skitter across the floor. Well, we scooped the bugger up and let him join the fun in our flask of liquid nitrogen.
Yup, he was instantly converted into a deep sleep. Then, we tossed the baby and his bathwater onto the floor in those pretty balls of steam as liquid nitrogen is famous for. The cockroach landed in two peices. He eventually woke up and couldn't find his feet!
Has anyone ever tried to run an electronic circuit in temperatures that cold? I suspect the doped regions of transistors would behave differently and have different gain characterstics. I'm not sure a computer would compute.
At work we have two types of transformers: open core and oil filled. The oil filled are much smaller and can handle voltages a magnitude higher up to 150,000 volts at a few megawatts 100% of the time due to the circulation and insulating properties of the oil and large heat fins. The open air type transformers a much larger and require natural convection and only see 14,400 volts. The only advantage of the open air might be the large magnetic core to dampen voltage fluctuations.
Needless to say, the oil filled transformers are sealed to prevent contamination of the oil and prevent oxidation and cumbustion. If they are ever overheated, the oil tends to break down over time, lose its dielectric properties, and eventually short. Some oil filled transformers have large fans on the heatsinks to keep the oil at reasonable temperatures.
I'm not sure what blend transformer oil is for our applications as we have a contractor repair our damage, but you can get it in 55 gallon drums. I'm sure any other oil, including mineral oil would be just fine in this application (provided moisture does not contaminate the oil over time.)
I could imagine a much "prettier" setup where the case is made of painted iron, closed, sealed, and painted. Then lower the freon pressure in the air conditioner to allow much lower temperatures when the gas expands inside the coils.
Then people might think this is cool and not be offended by the "scraps of styrofoam" and parts laying around everywhere. Looks like a prototype to me...
There are a couple of liquids which you might consider for your cooling to avoid the liquid problems. The one I would look strongly at is acetone.
Yes, the stuff smells bad, but it will remain as a liquid (and not get gooey or gelataneous) down much colder than -30 or -40 C... I use dry ice/acetone baths all the time, and it's still very much a liquid at -70C.
It's immiscible with water, so you won't have any water-contamination problems, and any water that hits the cold acetone will instantly freeze.
Although it has a relatively high vapor pressure at room temperature (i.e. it smells bad) the vapor pressure isn't nearly as bad at cold temperatures, and you won't smell it at all. And you won't smell it anyway if you have a cover on the case.
So anyway, you have my vote to use acetone as your cooling fluid.
>How about inserting the motherboard into a
>freezer with some kind of humidity control to
>eliminate condensation. That way you could
>overclock the entire system bus.
Essentially, the idea is to replace standard atmosphere with something that conducts heat as well or better and can be cooled more effectively,
AND is inert/non corrosive/non electrically conductive, right? So why not a pure nitrogen atmosphere? Of course, having a sealed unit would be a pain.
Pure antifreeze? Rubbing alcohol? (Things
that wouldn't freeze solid.) (Baby oil == mineral oil plus fragrance.) Of course, the problem with 'water'-cooled anything is the pain of the maintanance.
I've gotta disagree.
The dielectric constant is not important as a standalone figure. The capacitance is a function of the dielectric constant and the distance between plates. The air (or mineral oil) side of a PCB trace goes some distance before coupling. In fact, the only meaningful coupling increase might be to adjacent traces, creating increased crosstalk. Remember the other side of the trace is looking at about 0.06" of FR4 (dielectric constant of about 5) to the next layer, which should be ground below impedance-controlled traces. The incident-wave height should not change due to this. Changes to the FR4 material's thickness and/or dielectric constant would be much more siginficant.
BTW, this got me thinking about the system's bypass caps, so I checked. Tantalums will only drop in capacitance by about 8% at -40C(or F) and X7R ceramics will be about the same. Z5U ceramics will drop more, but not as much as if they were running at +70F. A quick search gave me no info on low-temp characteristics for electrolytics, but most are spec'd down to -40 or lower.
Yeah, I wondered about that too. Apparently the guy didn't understand conventional cooling methods, so decided to try something completely different.
Mineral oil? It has a very low electrical conductivity; otherwise, the board would short out. However, it also has a very low thermal conductivity, which means the components not directly in the garden pump path are probably net effect being heated.
There's also the pressure problem. What's the MTBF while being hit with 170 gallons/hour? MB components weren't designed, nor were they attached, with that kind of abuse in mind. My suspicion is that he'll get a critical failure inside a week of continuous use.
It will work better. I was a radar tech in the Navy, and the SPS-49 radar uses a water cooled klystron. Pure water has very low conducivity. We had no problems even though the klystron is powered at 40 KV. Corrision wouldn't cause any problems because water dosen't cause the problem, electrolyss caused by impurities is the cause of corrision. The drawback is keeping the water pure. You have to perform daily checks for water purity and have extra water on hand for changes. A far simpler way to keep any air cooled device working is to keep it clean. We used air filters and cleaned the filters once a week. The inside of the equipment was cleaned once a month. We also kept the equipment in air conditioned rooms, with the temp set below 70. You could get some dryer vent hose from the hardware store and duct air from the vent directally to the air intake on your PC.
Quemadmodum gladius neminem occidit, occidentis telum est
I can't believe that I had to slog all the way to the bottom before someone mentioned the most obvious caveat in doing this! Mineral oil, alcohol, acetone, what do they have in common? They're potent oxidizers! They'll eat anything organic. Wait until the oil (or whatever) eats through the styrofoam and dumps itself onto the carpet. It could be really fun with the more volitile stuff, let it drip down to where there's a pilot light, then wait for the vapor pressure to build...
Of course you could get to find out how long a PC can run after the PC boards delaminate and the plastic packages melt. After it does quit, you could sell it as high-tech art. I see the potential for making real profit on this!
Mr. Natural -- Cat Herder
... and though I hate to chill the enthusiasm here I have to point out that my handle is exactly what anyone doing this is likely to get: overshoot. Mineral oil has a dielectric constant of about 3 instead of the 1.0 for air, and that means that the surface traces on the MB will be both slower and lower impedance than they are designed to be.
Incident waves from the ICs will be smaller and may not make threshold, termination will be mismatched, there will be reflections from every change of layer, signals will take longer to get across the board, you end up with clock skew, and crosstalk will increase.
Most of these effects won't cause immediate failure. Or even frequent failure. Maybe just enough to make the system run like it was on Losedoze.
Lacking <sarcasm> tags,
That guy has a good idea, but he seems to have overlooked that if he removes a card, a thin layer of insulating oil would form over the connections, and it would be near impossible to clean it off.
If you just wanted to overclock the CPU, you could mount a peltier on the CPU, put a thermistor (sp?) on another part of the CPU, and built a simple thermostat that keeps the CPU at just the right temperator. That way you don't have to worry about condensation if the CPU halts for some reason and the peltier supercools it.
How about inserting the motherboard into a freezer with some kind of humidity control to eliminate condensation. That way you could overclock the entire system
bus.
-- Paperwork is the embalming fluid of bureaucracy, maintaining an appearance of life where none exists.
Hello All,
:)
The second box will be completed on Tues I hope (if I stop ICQing everyone AND stop responding to all of my emails).
Dr. Ffreeze
http://www.accsdata.com/DrFfreeze
Hello,
I have softmenu. Abit BX6 Rev. 2. But I have removed cards and put them back in with no problems.
Dr. Ffreeze
Hello All,
I can't wait until I get to work on my site. Work and such. This project was for me and so was the web site. I was just tinkering. I talked about it and no one listened or said that I was crazy. Ok. Not a problem. BLAM. 20,000 in on day! Questions out the wazoo. I will answer all emails but it will take some time. I WILL be updating with some benchmarks (dugh). I will get very detailed if the desire to know is out there. I am still in TESTING. Box 2 Tuesday 5-31-99 should allow for all out AC operation with no worry of condensation. Box 2 will come complete with LID!
Dr. Ffreeze
Hello,
:)
Might I share a bit? "Didn't understand conventional cooling methods, so..." Not true. Not true at all.
True. Very True. I am still in testing. The web page was mostly for my tinkering, otherwise I would (and will) offer MUCH many explinations. The 170 GPH pump is temp. Box 2 will allow the coils to be submerged in the oil.
Pressure? Not when the oil gets cold. It's like Mapel Syrup on a cold Winter day. I knew that I needed TONES of capacity if it were to pump the oil when it got to extreme temps. I was wrong in that the pump (or any for that matter that would fit) is not strong enough. I will have to look at other ways to agitate the oil.
Possible suspicion, but faulse. 3 weeks and running.
Dr. Ffreeze
PS. Not trying to be rude.
Hello All,
Thin layer of oil has not caused a problem.
I wanted to overclock my CPU and Video card. Many also overlock the fact that the RAM and therefore the FSB (front side bus) will be able to be overclocked. The lock on Intel sucks though.
Dr. Ffreeze
Hello All,
:) You saw a spring in my step! Here I am STILL not updating my site or siliconing my box 2! (I had some good talks though)
:( In a "dry test" (no oil) I got down to -38.5 C, but frost stoped me yet again. I got to 10 C or 12 C (NOT negitive) when I noticed ice forming and powered down.
Some quick info.
Who an I?
Average Joe overclocker, a bit extreem, a dreamer, and a bad spellar.
Why did I do this?
Always wanted to cool it a bit more but hated condensation problems (still have em but not for long). I love to learn and tinker. The concept was simple.
Why do I not tell any good info of cpu or new speed?
Long story short. I have talked about this type of thing for years (even on Tom's list server). No one paid heed unless they said that I was crazy (true). I have my own web page (blah). Out of two years I have about 2,000 hits. Guess who's homepage my site is? Me. So most of them were me hitting my own site. I took some pictures and tried a Kodak digital service. I also was trying to learn some FrontPage98 on a server that supported FrontPage Server Extensions. Posted pics. Mailed pics to Voodoo Extreeme, HarOCP, and talked on Ace's Hardware. No bites except for an offer IF I gave an exclusive. No dice. I wanted any and all to learn, look and question. EverQuest and something else made me have to reformat my drive (forget). This caused me to loose my web site access (new). I started tinker on designing some concrete lined speakers and crossovers. I got home from work and BLAM. 4,600 hits!!!!!! Wow!!!
5 news sites! WOW! Email and tons of it. I started answereing them. I tried to update my site explaining the speed and such but I had problems (still do).Time for bed (I run a 225,000 sq. ft. super store at night). I got up and 11,500 hits!
What the HELL speed are you at?
Retail Celeron 333 MHz
Fan Heatsink 416 MHz
Liquid cooling 416 MHz (WHAT?!?)
The quick is that until I get the coils submerged in box 2 (that I need to start) I will not be able to run the AC non-stop.
What are my Gaols?
See my web site and kick my arse if I don't update it 6-1 or 6-2.
Eamils are all welcome. I will help anyone in similar interests.
I hope this clears some things up,
Dr. Ffreeze
new site
http://www.accsdata.com/drffreeze
old site
http://members.iquest.net/~opto