Chemical Forums
Chemistry Forums for Students => Undergraduate General Chemistry Forum => Topic started by: Spaminator on October 07, 2008, 12:36:58 AM
-
First, let me thank you guys for taking the time. Second, I am in the IT field and, like most of my colleagues, am a huge computer nerd, which leads us to my rather strange questions.
I have been kicking around the idea of completely submerging a computer in mineral oil. Sounds crazy right? Well it's been done before and it actually works quite well! What hasn't been done (to my knowledge) is a simultaneous water cooled/mineral oil submerged/intercooled mineral oil rig. Now I want to be able to run this machine for and extended period of time and I have no idea of what the long term effects could be. Then I thought to myself "I'll ask some chemists!".
Here's the question: Will mineral oil react in any negative way (ie: corode, cause the oil to break down, etc) with the following materials:
a) PVC (Polyvinyl Chloride)
b) Acrylic glass, AKA Plexiglass (methyl methacrylate). Also similar to this is Lexan.
c) Aluminum
d) Copper
e) Silicone (sealant)
f) Steel
g) Nickel
h) Plastics in general
i) Silver
I know mineral oil has a negative effect on some adhesives, which is why I'm concearned about the silicone. The water blocks I want to use on the CPU/GPU etc will be copper and copper tends to oxidize when exposed to certain things. If copper is a no-go I can use aluminum, but that tends to oxidize easily if exposed to certain things as well. Aluminum is also what caps the dies of the processor itself! Steel and nickel will also most likely be making an appearance in this build. I need to know about the silver becuase I will be using a TIM (Thermal interface material) that contains silver (Arctic silver 5).
I have been trying to find out if these materials would be safe in mineral oil over a long period of time for a while now, but have been unable to find out. I don't expect you guys to do any exhaustive research. All I need is a thumbs up/down for each material, even if you might only know about one or two of them!
Sorry for the lengthy post.
Thanks again for your time!
-
Um, I think you should be alright. Mineral oil is very, very unreactive. The only thing I can think of is that it would maybe act as a solvent for silicone? Would need someone to confirm that - not sure if you can even solvate polymers like that; don't have a background in polymers at all.
-
I'm just going to second everything nj said. Sounds almost exactly like what I was gonna say.
-
A couple of ideas though -
1) Mineral oil doesn't have a spectacular specific heat - there's a good chance your entire aquarium will start to get hot as the container itself is an insulator. I'd suggest a cylindrical tank, with a fan positioned in it such that it circulates the oil around the tank. As a top cover, you could cut some mesh grating so that it would prevent dust from getting in. You could then position a fan external to the tank, facing so that it circulates air over the top of the oil.
2) The fan is going to have a lot of pressure on it due to running in oil when it's meant to run in air. I'd suggest maybe bending the blades so that their cut through the oil is a less sharp angle.
-
Well nj you are right with the way mineral oil handles heat, but that is one of the reasons why I want to use it. Since there will be minimal "hot spots" in the tank, I can circulate the oil how ever I want and still cool it effectively.
Without cooling the oil at all it is likely the oil temp would be somewhere in the 80*C range, which is really high for any computer component and will most likely kill/shorten the life span of anything in the tank.
My cooling solution is to first water cool all of the major heat producers: CPU chipset GPU...even RAM. This will give me a nice head start and probably save a good 20-30* (oil temp) right off the bat! It is possible to just leave the heatsink/fans on in the oil and it seems to have little effect on the lifespan of the fan. There is a company that did this about a year ago and just left the fans on and they are still running just fine.
Then I will pump the oil through a radiator with fans on it. This will let me get the oil temp as close to ambient as possible.
-
Your oil is going to go rancid and stink like high hell. I actually did this when I was ~16 for overclocking a Pentium 133MHZ based on an article I read on the internet. This was years and years ago.
It worked fine, other then becoming nasty smelling.
-
HAHA! Thanks for the tip! I haven't heard of anyone else complaining about smell. I know some people have tried this with veggie oil, and that will go rancid after a while.
-
Yeah, I am remembering now I used either olive oil or not vegetable. Mineral oil should not go as rancid as fast. It will not go rancid per say, just get very dirty.
-
Yeah, that's what I figured. Any organic oils will break down over time (especially with heat) and eventually smell bad. More concearning is they lose the ability to transfer heat, and their electrically insulative properties. It could also coagulate and clog any pumps/hoses/radiators/filters, etc that it may come into contact with. It will also turn a yellow or brown in color, which will just look bad.
Any other thoughts on my original materials list?
-
Silica/Silicon Oil would also be a good candidate. I know it has a higher working temperature, you might look for data comparing the heat capacity and thermal conductivity of it and mineral oils.
Also, look at the viscosity as well. The lower the viscosity the more fluid like it will be, and will make it easier to stir/pump for cooling purposes and such.
-
I read about some folks who used silicone oil as well. It also works. I'm not sure about wether or not it's better than mineral oil but I do know that it is cost prohibitive. About double the price per gallon.
I did find an MSDS from one manufacturer on their mineral oil:
Reactivity Data
==================================================================
Stability: YES Cond To Avoid (Stability): NONE SPECIFIED BY MANUFACTURER. Materials To Avoid: DANGEROUS
REACTIONS W/STRONG OXIDIZERS Hazardous Decomp Products:INCOMPLETE BURNING CAN PRODUCE CARBON
MONOXIDE &/OR CARBON DIOXIDE & OTHER HARMFUL PRODUCTS Hazardous Poly Occur: NO
Conditions To Avoid (Poly): NOT RELEVANT
==================================================================
I don't know if this helps.
-
It means you're not going to die, as long as you don't do something nonsensical like add Mn2O7 to your oil.
-
Not sure how some of the plastics will stand up to longer term submersion in the oil??
Could you drop dry ice into the oil to cool it right down well below zero? - it won't freeze, but will get a little more viscous, but mainly it will get very cold and make good contact with all of your components and should cool them effectively.
-
Many plastics will swell, the more flexible the plastic, the more it will swell. This is because the tangle of molecules that makes up a plastic has void spaces inside where the oil molecules can slip. This process can take months. Silicones are great absorbers of solvents. A computer also has adhesives that hold things together. The effect on these would be unknown.
For a coolant, consider polyethylene glycols. They are still not very good conductors of heat, but they're better than mineral oil. Also, they come in a number of viscosities and melting points. The thinner the oil, the better convection you'll have. They will still react with polymers. Almost any liquid will cause polymers to swell and weaken.
Dangerous Bill
-
Hmm. I'll look into that. Thanks!
-
This question comes up quite often, and there is really only one answer. I stayed out of this discussion, because I don't want to seem discouraging, but before you all wear yourselves out ... Some liquids will corrode or oxidize the metals used, some will attack the resins used to make boards and chips. When you strike out all of them, you're left with one fluid -- Fluorinert made by 3M. It is expensive, and even when Cray supercomputers need to use it, sophisticated engineering is needed, and for reasons I'm sure are well outlined in some research papers authored by Cray engineers, still not ideal for their application. Mineral oil just won't work. If it did, why would Cray have had 3M develop Fluorinert?
http://en.wikipedia.org/wiki/Fluorinert
-
Arkcon, this site seems to suggest that mineral oil is very possible.
http://www.pugetsystems.com/submerged.php
Just because a high-end computer company like Cray paid 3M to come up with a material for their supercomputers doesn't mean that for hobbyists on small scales it doesn't work. I'm sure it has downsides, but like I said, for a hobbyist, I'm sure it works.
-
Ya Puget systems is one of the companies I've been basing alot of my information on. They seem to have a good handle on how to do it. I'm just more concearned about corrosion because I am going to simultaneously water cool my system.
-
Arkcon, this site seems to suggest that mineral oil is very possible.
Wow, according to that site, one year without any attack on the resins or sealant, just some adhesive labels coming off. That is impressive. Gonna have to bookmark that for the next time it comes up on /. Pugetsystems insists it's not a hoax, but it runs counter to everything I've read before. Do it Spaminator:, macman: will buy you a new computer if you fry it.*
*Disclaimer: I do not represent macman:
-
Hahahaha, I wish I had the funds to support spaminators endeavor, but I just bought a new computer myself (after my old one was splashed with a cup full of coke) :)