Advanced search

Message boards : Graphics cards (GPUs) : Nvidia and liquid submersion cooling

Author Message
Palamedes
Send message
Joined: 19 Mar 11
Posts: 30
Credit: 109,550,770
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 34017 - Posted: 25 Nov 2013 | 3:38:41 UTC

So I'm goofing off in my man cave playing with creating a submersion computer like the puget systems aquarium computer ( http://www.pugetsystems.com/mineral-oil-pc.php ) using my own parts that are lying around..etc.

It works but the nvidia 285 card I had sitting on my shelf worked like a champ, until it hit the oil at which point it died.

The rest of the computer works fine, but bios boot gives one long 2 short beeps which means dead video card.

Does nvidia have something that detects submersion and prevents it? Why would the mineral oil kill it?

Thoughts?

Dagorath
Send message
Joined: 16 Mar 11
Posts: 509
Credit: 179,005,236
RAC: 0
Level
Ile
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 34019 - Posted: 25 Nov 2013 | 11:09:41 UTC - in response to Message 34017.

In their description of their V2 system, they (Puget Systems) say they had a GTX 285 working fine in oil. Of course you may have a different brand.

Due to the viscosity of the oil the fan won't spin at anywhere near normal speed. Perhaps your GTX 285 monitors the fan RPM via the fan tachometer and if it finds the fan RPM isn't correct for the amperage it's delivering to the fan then it assumes the fan isn't functioning correctly and refuses to startup?

Maybe Puget Systems' 285 doesn't check if the fan RPM agrees with the amperage? Have you tried cleaning the oil out to see if it runs at all or is it fried for good?

____________
BOINC <<--- credit whores, pedants, alien hunters

Betting Slip
Send message
Joined: 5 Jan 09
Posts: 670
Credit: 2,498,095,550
RAC: 0
Level
Phe
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 34020 - Posted: 25 Nov 2013 | 12:18:42 UTC - in response to Message 34017.
Last modified: 25 Nov 2013 | 12:21:27 UTC

So I'm goofing off in my man cave playing with creating a submersion computer like the puget systems aquarium computer ( http://www.pugetsystems.com/mineral-oil-pc.php ) using my own parts that are lying around..etc.

It works but the nvidia 285 card I had sitting on my shelf worked like a champ, until it hit the oil at which point it died.

The rest of the computer works fine, but bios boot gives one long 2 short beeps which means dead video card.

Does nvidia have something that detects submersion and prevents it? Why would the mineral oil kill it?

Thoughts?


Can't help with the problem but just wanted to say it looks "cool".

They could add a bigger tank with a partition which would contain water and have a couple of warm water fish in as well.

Profile Damaraland
Send message
Joined: 7 Nov 09
Posts: 152
Credit: 16,181,924
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwat
Message 34025 - Posted: 25 Nov 2013 | 21:14:36 UTC - in response to Message 34017.
Last modified: 25 Nov 2013 | 21:25:17 UTC

Thoughts?

Here are some (just thoughts):
- Mineral oil is very good heat tramsmissor. It's used in most of Power Transformers, so this part ok.
- One problem I see is that you need a spetial pump becouse oil density is hight. And you should move the oil. It you put a radiator behind with fans no real big gain??
- If you put GPUs you have to consider a lot of heat I am not sure who well heat will be dissipated throught the glass area, probably you wont know until you test it.
- Another problem I see too is that the heatsinks are not designed to work this way. I would suggest to put passive heatsinks as big as you can. Even in the GPU.
- Consider that it's going to be very heavy. That shouldn't be a problem, I had seen 1.000 l (kg) aquarium, but watch out if you have a leak...
- It might smell a bit as fried oil.
As I said just some thought I saw your link but didn't read carefully all.
Good luck.

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 34030 - Posted: 26 Nov 2013 | 20:35:18 UTC - in response to Message 34025.

Fans move air. To move liquid what's needed is a propeller or a peristaltic pump. Unfortunately such pumps require a bit of power (maybe 50W). Better heatsinks would be useful, as said. I think the design of the tank and positioning of the GPU's would be key. Lets just say the deep fat fryer design is thermodynamically challenged! Something more like a pluming system, where heat is piped around and cooled/warmed.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Profile Damaraland
Send message
Joined: 7 Nov 09
Posts: 152
Credit: 16,181,924
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwat
Message 34092 - Posted: 1 Dec 2013 | 16:21:23 UTC - in response to Message 34030.
Last modified: 1 Dec 2013 | 16:49:24 UTC

It seems some companies are working on this to cool servers. I think this articles are interesting:
http://green.blogs.nytimes.com/2012/09/06/cooling-a-computer-server-with-mineral-oil/?_r=0
http://www.grcooling.com/

I look throught the web and there are a few videos on youtube of computers with submersion cooling. I read and it seems mineral oil don't smell. Somepeople are using Jonhson's Baby oil (90% mineral oil).
http://www.youtube.com/watch?v=Eub39NaC4rc
http://www.youtube.com/watch?v=AiqKycb7uh4

Reading more the original article you posted the "V3 DIY Kit Launch" pulls 820W out, so I don't see any reason why these system won't work. I don't know how much noise will the 9 fans on the rear do.

Profile Damaraland
Send message
Joined: 7 Nov 09
Posts: 152
Credit: 16,181,924
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwat
Message 34093 - Posted: 1 Dec 2013 | 17:41:39 UTC
Last modified: 1 Dec 2013 | 18:14:44 UTC

I found this too. It's a liquid called Novec. You don't need heatsinks.

http://en.wikipedia.org/wiki/Novec_1230
http://www.youtube.com/watch?v=CIbnl3Pj15w
http://www.youtube.com/watch?v=Z_X_hgtlJpA
http://www.youtube.com/watch?v=vw9px3ittDg
http://www.youtube.com/watch?v=NymeDU96pac&list=TLKR_in1Sisci7_2px05luN0i-sTaXia7G

This liquid has 36ºC boiling point
http://multimedia.3m.com/mws/mediawebserver?mwsId=tttttvW9lEgUmy7VPZA_qZ0t2XW62EW9iXut2Xut2tttttt--

And here some interesting discussion on this liquid:
http://www.overclock.net/t/1209583/3m-novec-7000-group
____________
HOW TO - Full installation Ubuntu 11.10

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 34125 - Posted: 4 Dec 2013 | 20:38:13 UTC - in response to Message 34093.

If you're boiling the liquid you'll need a sealed system to recover the vapour. Otherwise it will be a short run. And I'd be at concerned about a high concentration of this stuff in the air I'm breathing. I don't know anything specific about this Novec, though.

@Palamedes: does your "mineral oil" conduct electricity? Of course it shouldn't, but if it does you found the error source. Put multimeter tips in the oil and measure resistance depending on distance. You should get infinite reading until the tips are very close, otherwise you've got some conducting liquid.

Pure water is not conductive either, but any solved ions will make it conduct. These can come from anywhere in the ambient, even dust. Maybe some contamination happened to your oil as well? It won't dissolve salts the way water does, but electricity really likes finding the "best" pathway ;)

MrS
____________
Scanning for our furry friends since Jan 2002

Profile Damaraland
Send message
Joined: 7 Nov 09
Posts: 152
Credit: 16,181,924
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwat
Message 34136 - Posted: 6 Dec 2013 | 12:51:52 UTC - in response to Message 34125.
Last modified: 6 Dec 2013 | 12:52:37 UTC

@Palamedes. I missunderstood your question (matter of my "perfect" english). I've been reading about this stuff.
More thoughts:
- Maybe as ETA is saying there's dust in your oil causing micro shortcuts. The solution to this is filtering. I saw some desings on the internet. People filtering with cofee filters, but I would give a try a Aquiarium filter with passive materials (I could help you with this).
- I don't think you have a conductivity problem with the mineral oil becouse otherwhise you should have problems with the board too. And probably you would have a power loss throught ground causing your Residual-current device activate.
- If I have to give a guess I would say that oil is not cooling properly the card. Did you take out the case and make sure the oil was moving properlly?
- If you put bubbles, there could be the cause of the problem. Normal air pumps puts condended water into the oil. I would put a dried air pump or a aquarium pump to move the oil as much as possible.

@ETA

If you're boiling the liquid you'll need a sealed system to recover the vapour. Otherwise it will be a short run. And I'd be at concerned about a high concentration of this stuff in the air I'm breathing. I don't know anything specific about this Novec, though.

Yes you need to have a sealed system. The evaporating loss is very low. The liquid is very expensive.

Dagorath
Send message
Joined: 16 Mar 11
Posts: 509
Credit: 179,005,236
RAC: 0
Level
Ile
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 34274 - Posted: 13 Dec 2013 | 11:34:16 UTC - in response to Message 34136.

In a different thread Damaraland wrote:

Another thing, if anyone can answer about mineral cooling I'll apreciate it.I'm thinking in doing some hardware with Arduino to make a quite coplex cooling system with temperature control (just for fun and I'm bit sick of the fan's noise).


Arduino is very cool but you could do what you want at far lower cost with an 8051 or 6809 microprocessor. You could even use your PC to control all that. It's easy if it has a parallel port. Even if you don't see a parallel port on the back of the case the mobo might have header pins for a parallel port, even newer mobos often have them. RS-232 serial ports haven't totally disappeared either. They're not as easy to use as a parallel port but they're easy enough. USB ports are not that difficult to use either and the connectors are cheaper, smaller and less prone to damage than parallel or serial connectors. And USB is ubiquitous.

Cooling is one of my favorite topics :) I was thinking that it might be possible to create an almost totally silent very efficient cooling system with just 3 components:

  1. mineral oil
  2. a container, perhaps an aquarium, or it could be something opaque and cheaper than a transparent aquarium
  3. a large U shaped aluminum or copper heatsink
  4. 2 stir propellers driven by 12 volt gear-motors



Think of a container filled with mineral oil and an aluminum heatsink in the shape of an upside down U. One leg of the U is immersed in the oil while the other leg is outside the container. It is easy to find 6 mm (.25") thick aluminum plate and create a U shape. The width of the plate would be slightly less than the length of the container. To keep it simple and easy to fabricate, it doesn't have fins. Instead of fins it relies on large mass and the large surface area of the flat plate.

Gear-motors are a combination of a small electric motor and gear reduction box in one unit. Robot hobbyists use them and I think they are not too very expensive. They are available with various gear reduction ratios and horsepower. The stir propellers and gear-motors would be immersed in the oil to circulate it around the tank, past the CPU, GPUs and hot side of the heatsink.

Does that description give you an idea of what I am thinking about?

I wonder if a heatsink that large would conduct enough heat to keep the oil cool. Of course it depends how much heat is being generated but let's say it's an i5 Haswell plus 2 X GTX 780 system. Does anybody know some simple formulae for calculating the heat carrying capacity, in watts, of a plate of common grade aluminum with known length, width and thickness?

Here is another idea, I call it "the calorie scoop"...

It consists of a container full of mineral oil and a circle of aluminum or copper, call it "the wheel". The wheel rotates slowly, maybe 2 or 3 RPM. Half of the wheel is immersed in the mineral oil while the other half is exposed to the air. The half in the oil picks up heat as it rotates through the oil while the half exposed to the air sheds its heat to the air. You could say it scoops calories and dumps them into the air. There could be a soft wiper blade just above the oil to wipe it off the aluminum to facilitate flow of heat from aluminum to air. Or you could weld fins to the wheel to increase surface area or just bolt fins on but welded fins conduct heat faster. The fins would stir and circulate the oil.

While I'm at it a third idea...

All mineral oil bath cooling systems I've heard about use a glass or plexiglass fish aquarium for the tank, probably because used ones are inexpensive and you can see inside the tank. The problem is that neither glass nor plexiglass conduct heat well so the large and potentially useful surface area of the tank is wasted. Then one must add potentially noisy and possibly expensive pumps, radiators, etc. to remove the heat. If the tank were made of aluminum or copper it alone might conduct enough heat. If it were made of corrugated plate it would conduct much more heat. A large, slow, quiet, ceiling fan would make it even better. The beauty of it is the only moving part is the fan but they run for many years and they're very inexpensive.

____________
BOINC <<--- credit whores, pedants, alien hunters

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 34300 - Posted: 14 Dec 2013 | 13:16:54 UTC - in response to Message 34274.

Regarding your 1st idea: relying on u-shaped metal won't be very effective in transferring the heat. A heat pipe should fare much better. I'm assuming your reservoir contains water? If your oil immersion tank was made from metal, you could contact this directly with the water and have a massive surface for much better contact than trying to build a "heat bridge" between the two containers.

The 2nd idea: I don't think it's a good idea to break the systems sealing. And it wouldn't be all that necessary, I think. Exchange the top plate on the oil for a metal plate and it will already pick up quite some heat. Put fins on top and you've got enough surface area for that ceiling fan (or slow spinning, big PC fans - probably more expensive but higher quality and may consume less power). Also add some bottom fins to the metal plate and you've already got some powerful cooler.

As you see from these comments I'm all for that metal case. You can extend this concept of "fins in, fins out" to all sides. And you wouldn't need some water reservoir nearby, the oil itself is probably already enough (compare that volume to how little is inside water cooling loops).

To judge overall cooling capability one can approximate that the heat generated from the PC components is distributed evenly inside the oil. In steady state the cooling system has to dissipate just as much heat as is being produced. So one would have to consider the air cooling as the limiting heat exchange. Given some geometry and air flow there will be a oil temperature at which the two powers are equal. Real chip temperatures will be somewhat higher, due to temperature gradients being present (the heat is not evenly distributed instantaneously), but I think this would be a sound first approximation. Some data and material properties would be needed, though.

MrS
____________
Scanning for our furry friends since Jan 2002

Profile Damaraland
Send message
Joined: 7 Nov 09
Posts: 152
Credit: 16,181,924
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwat
Message 34307 - Posted: 14 Dec 2013 | 15:09:29 UTC - in response to Message 34300.

More I think, more I see this challenguing...
Maybe the "mistake" is to talk about this here and not in a specific cooling forum, there's a lot of people that have tried this, but don't seems there's a lot of sucess.
Anyway, I think copper is not a good option because of corrosion with mineral oil. That could make the mineral conductive with time.
Building a box with all aluminium but the front with cristal won't be difficult. The silicon used to build aquarium is very strong and heat effective.
I think any system like this should have a filter. An aquarium filter should do to clean big particules from dust.
I saw this interesting idea too, it's from a aluminium radiator. There you can see the size needed depending on water temperature and size:

I will have a look a RS232 temperature sensor to control the fans. That might be good idea too to fan control.

TJ
Send message
Joined: 26 Jun 09
Posts: 815
Credit: 1,470,385,294
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 34316 - Posted: 14 Dec 2013 | 21:04:08 UTC - in response to Message 34307.

If copper (Cu) reacts with the mineral oil over time, then aluminum (Al) will do more easily as copper is a more noble metal then aluminum. The less noble a metal the more easily it will react. However there are additives for the oil to prevent this, but the oil need to be refreshed every certain period, depending on the grade of the oil.
____________
Greetings from TJ

Dagorath
Send message
Joined: 16 Mar 11
Posts: 509
Credit: 179,005,236
RAC: 0
Level
Ile
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 34319 - Posted: 15 Dec 2013 | 7:49:03 UTC - in response to Message 34300.

Regarding your 1st idea: relying on u-shaped metal won't be very effective in transferring the heat. A heat pipe should fare much better. I'm assuming your reservoir contains water? If your oil immersion tank was made from metal, you could contact this directly with the water and have a massive surface for much better contact than trying to build a "heat bridge" between the two containers.


The first idea uses only 1 container filled with mineral oil and the mobo would be immersed in that oil. One leg of the U would be in the oil, the other in air. Obviously one would need to support the cold side (air side if you prefer) of the inverted U with brackets or whatever but that's small details. If the tank were on a table one could make the cold side long enough to extent down to the floor. It would be more like an inverted letter J than an inverted U. An average work table is roughly 750 mm tall. Add 400 mm for the height of the tank and you have about 1300 mm length on the cold side if it sits on the floor. That arrangement would be very stable but a pair of brackets to brace the cold side to the table would make it even better. The question is... how much plate does one need on the cold side to suck the heat out of the oil fast enough? Assume 850 watts at the source (ie. in the oil), how much plate, at what dimensions (l X w X thickness) would be required in ambient 25C air? And how much additional capacity would be gained with a fan blowing 25C air over the cold side? There's gotta be formulae to give a pretty close estimate.

The same idea would be even easier and cheaper to implement with just a flat plate. Put the table the tank sits on beside a wall and allow the cold side to stick up above the tank. Brace it to the wall or ceiling for stability.

The 2nd idea: I don't think it's a good idea to break the systems sealing. And it wouldn't be all that necessary, I think. Exchange the top plate on the oil for a metal plate and it will already pick up quite some heat. Put fins on top and you've got enough surface area for that ceiling fan (or slow spinning, big PC fans - probably more expensive but higher quality and may consume less power). Also add some bottom fins to the metal plate and you've already got some powerful cooler.


I think you misunderstood this one. The rotating wheel is oriented vertical with the shaft it rotates on horizontal. Any point A on the circumference enters the oil on the left, rotates through the oil where it picks up heat then emerges from the oil on the right where it begins to shed heat to the air. Point A rotates up to the top then back down into the oil on the left side tom pick up another "scoopful of calories". I think it could be very effective since a calorie (if you think of it as a particle rather than energy) doesn't have to travel through the height of the hot side of the heatsink to get to the the cold side. A calorie has to travel only half the thickness of the plate at most where it meets a calorie moving in from the other side of the plate.

As you see from these comments I'm all for that metal case. You can extend this concept of "fins in, fins out" to all sides. And you wouldn't need some water reservoir nearby, the oil itself is probably already enough (compare that volume to how little is inside water cooling loops).


Neither of the 2 previous concepts include a water reservoir. I like this 3rd idea too unfortunately it's probably the most expensive one to build if you build out of flat plate then attach fins later. However, it could be built out of corrugated plate which has far more surface area per linear unit of measure than flat plate. The 4 sides would be of corrugated with corrugations running vertically. The easiest way to make the bottom would be with flat plate but if one were willing to do some difficult cutting the bottom could be corrugated too. Any way you build it this concept it requires welding. It's easy to weld with the proper equipment and technique but the equipment is not common.

To judge overall cooling capability one can approximate that the heat generated from the PC components is distributed evenly inside the oil. In steady state the cooling system has to dissipate just as much heat as is being produced. So one would have to consider the air cooling as the limiting heat exchange. Given some geometry and air flow there will be a oil temperature at which the two powers are equal.


Exactly. It follows from the Law of Conservation of Energy so we can use simple energy methods. Then the equation begins as:

power generated by computer = power shed by heatsink

The left side is approximately the sum of the TPDs of the major heat producing components plus 10% to cover the others. Right side is some function of heatsink material thermal conductivity and dimensions expressed in terms of T (temperature), solve for T to get the equilibrium temperature for input dimensions, or, pick the target operating T, sub it into the equation and solve for dimensions. Rig a test and see how close the real numbers compare with the theoretical.

I wonder if someone's already whipped up a heatsink configurator-calculator wizard?

Real chip temperatures will be somewhat higher, due to temperature gradients being present (the heat is not evenly distributed instantaneously), but I think this would be a sound first approximation. Some data and material properties would be needed, though.


Very sound. OK, off to find data and material properties or a wizard.

____________
BOINC <<--- credit whores, pedants, alien hunters

Dagorath
Send message
Joined: 16 Mar 11
Posts: 509
Credit: 179,005,236
RAC: 0
Level
Ile
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 34321 - Posted: 15 Dec 2013 | 8:45:36 UTC - in response to Message 34307.

More I think, more I see this challenguing...
Maybe the "mistake" is to talk about this here and not in a specific cooling forum, there's a lot of people that have tried this, but don't seems there's a lot of sucess.


Puget systems mentioned by Palamedes in the OP seems to have had great success over several years. I find it odd that his GPU quit immediately after submerging it in oil but since he hasn't answered any questions to help clarify the situation (we can't read his mind and his explanation is ambiguous) there isn't much one can conclude. Who else has tried oil immersion and failed? Every attempt I've ever read about aside from Palamede's attempt has been successful. My posts are simply attempts to explore ways to make it cheaper as the one thing I've noticed is that people who are successful are using some very expensive components.

Anyway, I think copper is not a good option because of corrosion with mineral oil. That could make the mineral conductive with time.


Agreed.

Building a box with all aluminium but the front with cristal won't be difficult. The silicon used to build aquarium is very strong and heat effective.


Build whatever you wish but I would not assume a simple aluminum box will dissipate enough heat. That's what ETA and I are discussing... a way to approximate what temperature a system will run at before spending the time and money to build it.

I think any system like this should have a filter. An aquarium filter should do to clean big particules from dust.


I agree but that's a minor detail that can be added on after one has proven the thermodynamics of the system works.


I saw this interesting idea too, it's from a aluminium radiator. There you can see the size needed depending on water temperature and size:


Radiators are abundant. I can go to the local hardware store and buy enough radiator to cool 30 computers or just 1 computer. I can go to my friend's shop and get all the free tube and fin radiator I want. Automotive parts stores carry probably 100 different radiator sizes or more. Tell them what size you want and they'll sell one to you. The thing is... how much radiator do you need to do the job you want to do? How big of a pump do you need? Are you going to slap a bunch of parts together and find it doesn't provide enough cooling? What will you do then? Or will your creation give you twice as much cooling as you need? If it does then perhaps you could have spent half as much money. It's called engineering. Engineer first, acquire parts later for the greatest chance of a successful outcome on the first attempt without spending too much money.

I will have a look a RS232 temperature sensor to control the fans. That might be good idea too to fan control.


Have a look at this PWM controller . It uses a PIC mcu (microcontroller) and external ADC. Probably costs about $8 for parts. It's very simple and easy to build and should do what it's designed to do reliably for many years. He's using it to cool a disk but it would work on just about any device including a GPU.

It doesn't interface to a PC but that makes it easier to build and program.

____________
BOINC <<--- credit whores, pedants, alien hunters

Dagorath
Send message
Joined: 16 Mar 11
Posts: 509
Credit: 179,005,236
RAC: 0
Level
Ile
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 34322 - Posted: 15 Dec 2013 | 8:58:55 UTC - in response to Message 34316.

If copper (Cu) reacts with the mineral oil over time, then aluminum (Al) will do more easily as copper is a more noble metal then aluminum. The less noble a metal the more easily it will react. However there are additives for the oil to prevent this, but the oil need to be refreshed every certain period, depending on the grade of the oil.


Yes, I thought I heard there are additives to prevent corrosion, thanks for confirming that. So what do you need to do to refresh the oil? Do you have to replace it or just dump in more additive? I read mineral oil is very expensive so I would not want to replace it very often.

There must be other types of oil that are non-conductive. Motor oil is just long non-polar hydrocarbon chains so I can't see it being very conductive. Ethylene glycol should not be conductive either, I would think. Does one really need clear fluid? Why? As long as it works I don't need to see my PC. It can be submerged in motor oil as long as it works and it's cheap. How about hydraulic oil?

____________
BOINC <<--- credit whores, pedants, alien hunters

Profile Damaraland
Send message
Joined: 7 Nov 09
Posts: 152
Credit: 16,181,924
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwat
Message 34326 - Posted: 15 Dec 2013 | 11:29:52 UTC - in response to Message 34316.
Last modified: 15 Dec 2013 | 11:38:48 UTC

If copper (Cu) reacts with the mineral oil over time, then aluminum (Al) will do more easily as copper is a more noble metal then aluminum. The less noble a metal the more easily it will react. However there are additives for the oil to prevent this, but the oil need to be refreshed every certain period, depending on the grade of the oil.


This is wrong. Aluminium has very, very little corrosion.

From Wikipedia: "Aluminium is remarkable for the metal's low density and for its ability to resist corrosion due to the phenomenon of passivation."

I find it hard to explain. I studied this a long time ago, but you don't have corrotion unless there's "Metal dusting" or "Galvanic corrosion" occurs when two different metals have physical or electrical contact with each other and are immersed in a common electrolyte.

Dagorath
Send message
Joined: 16 Mar 11
Posts: 509
Credit: 179,005,236
RAC: 0
Level
Ile
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 34327 - Posted: 15 Dec 2013 | 12:24:19 UTC - in response to Message 34326.
Last modified: 15 Dec 2013 | 12:24:57 UTC

Aluminum is very reactive. If exposed to air it very quickly oxidizes but the oxide forms a hard layer on the surface that oxygen cannot penetrate. Formation of that layer is an example of passivation. The layer of oxide is usually just a few nanometers thick but it prevents further oxidation. If the oxide is removed then oxidation begins again and continues until a new layer of oxide forms.

I am not sure what chemical reaction occurs when mineral oil contacts aluminum but apparently it does happen and apparently passivation does not occur so the reaction continues.

Iron is fairly reactive though not as reactive as aluminum. It oxidizes too but the oxide layer is porous therefore it does not passivate. Oxygen penetrates the oxide and reacts with non oxidized iron below the oxide. Also, the oxide is not hard and tough like aluminum oxide so it easily separates from the iron which of course decreases it's ability to passivate. Most alloys of iron and carbon, steel, do not passivate either. The notable exception is an alloy called Corten which does passivate. Stainless steel, an alloy of iron, carbon, and various amounts of nickel, chrome and columbium are extremely non-reactive at normal temperatures and pH level. Some grades of stainless steel are resistant to hot acid. Ferric chloride, however, will eat through stainless steel and nearly every kind of metal very quickly.
____________
BOINC <<--- credit whores, pedants, alien hunters

TJ
Send message
Joined: 26 Jun 09
Posts: 815
Credit: 1,470,385,294
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 34337 - Posted: 15 Dec 2013 | 23:46:03 UTC - in response to Message 34326.

There is a lot of information in Wikipedia but not all is true or the only truth.
I know what I studied at university about the oxidizing reactions of metals.
You can test it easily. Take, aluminum, copper, iron, non magnetic stainless steel, silver, yellow copper, what you can find (a golden (wedding) ring as it will not react). The pieces have to be the same size or the same mass. Put a strong (same amount every time) acid on all the metals and see the results. Do it outside, not in the wind and wear always safety goggles. Take the same metals and pour a strong base (bicarbonate of soda) over and see the reactions.

I am not saying that mineral oil will react that spectacular, but if it reacts with copper, it will do for sure with aluminum. There are also different processes to purify (melting)aluminum from bauxite. And so you have aluminum with different properties and strength.

And at last, keep in mind that most mineral and all crude oils are not healthy, they are hazardous and become smelly when heathen. Therefore is silicon oil used in laboratories to make oil baths (same like water baths) but can heated up to around 400­°C. I don't know the conductivity of silicon oil, never measured, but it is very expensive. Perhaps Wikipedia has other information, but it is you call. I know what I know.
____________
Greetings from TJ

Profile Damaraland
Send message
Joined: 7 Nov 09
Posts: 152
Credit: 16,181,924
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwat
Message 34339 - Posted: 16 Dec 2013 | 6:24:11 UTC - in response to Message 34337.
Last modified: 16 Dec 2013 | 7:20:04 UTC

There is a lot of information in Wikipedia but not all is true or the only truth.
I know what I studied at university about the oxidizing reactions of metals.
You can test it easily. Take, aluminum, copper, iron, non magnetic stainless steel, silver, yellow copper, what you can find (a golden (wedding) ring as it will not react). The pieces have to be the same size or the same mass. Put a strong (same amount every time) acid on all the metals and see the results. Do it outside, not in the wind and wear always safety goggles. Take the same metals and pour a strong base (bicarbonate of soda) over and see the reactions.

I am not saying that mineral oil will react that spectacular, but if it reacts with copper, it will do for sure with aluminum. There are also different processes to purify (melting)aluminum from bauxite. And so you have aluminum with different properties and strength.

And at last, keep in mind that most mineral and all crude oils are not healthy, they are hazardous and become smelly when heathen. Therefore is silicon oil used in laboratories to make oil baths (same like water baths) but can heated up to around 400­°C. I don't know the conductivity of silicon oil, never measured, but it is very expensive. Perhaps Wikipedia has other information, but it is you call. I know what I know.


You know what you know. The conditions you are talking about is without passivation efect. That's all explained in the wikipedia and in many chemistry books too. But this is not taking the aluminium to the sun, to the moon, puting it to 400° C, or puting it in sea water. In normal conditions aluminum don't oxidate. But is true as said before that needs oxigen to pasivate and maybe in an oil container there's no much oxigen. I'm not sure for that.
You are not the only one that went to university and studied chemistry.

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 36281 - Posted: 14 Apr 2014 | 19:38:36 UTC

Sorry for not responding for a long time. In the mean time I have done no further work / thinking regarding this topic.. but today stumbled upon this Matlab toolbox for thermal simulations. At first glance it seems capable of simulating the designs discussed here.

There's a trial version and a 6 months students version for anyone to try, given access to Matlab.

MrS
____________
Scanning for our furry friends since Jan 2002

Post to thread

Message boards : Graphics cards (GPUs) : Nvidia and liquid submersion cooling

//