Pages:
Author

Topic: Mineral oil submersion? (Read 8364 times)

donator
Activity: 1218
Merit: 1079
Gerald Davis
March 11, 2012, 10:33:12 AM
#23
And I hope that will work better (4m pump head is a fair amount) because the others need 24V and thats no so easy to generate at those amps. 12V can be powered by a PSU. If its still not enough, Ill may just add more.

If that doesn't work out you may want to look into Iwaki MD series pumps.  Monster head and flow.  Normally used in commercial aquariums or industrial process water pumping.   I am almost certainly using an Iwaki for my watercooled server rack.    I am just not sure on MD-40 series or MD-55 series.

I wish I had a pond like you do (or stream would be even better Smiley ).  Having a giant water sink as the terminal source would make things much easier.
donator
Activity: 1218
Merit: 1079
Gerald Davis
March 11, 2012, 10:25:26 AM
#22


Sorry, I just have to add on this that refined mineral oil submersion is being used to cool datacenter servers.

It is a workable solution.  If mineral oil was potentially damaging over long term to so many possible board components then they probably wouldn't be submerging rackmount servers in mineral oil baths.

http://www.datacenterknowledge.com/archives/2011/04/12/green-revolutions-immersion-cooling-in-action/

"A four-rack installation of the Green Revolution liquid coolings solution, which submerges servers in a coolant similar to mineral oil."

As I indicated above there are lots of plastic safe, ion blocking, dielectric fluids.  They tend to be very expensive.  As in $50 to $200+ per gallon.   It isn't that it CAN'T be done it is that for mining GPU are 90% of the heat.  It is simply cheaper to put a waterblock on the GPU.

Still a cool link thanks for sharing it.  There are energy cost savings to immersion cooling or liquid cooling.  AC is expensive.   Removing 10KW of electrical load will require another 3KW (or so) in AC power.   If you can get the heat into a liquid you can remove 3KW easily with a large heat exchanger and fan (for the company in the link they are using evoporation coolers due to larger scale).

It just becomes INCREASED CAPITAL COST vs INCREASED COOLING COST.  There is a payback period involved.

hero member
Activity: 518
Merit: 500
March 11, 2012, 07:15:15 AM
#21
Im building an oil/water cooled setup.  The idea is simple: invert the motherboard, place only GPUs in the oil bath, pump the oil through a cooler, and in my case, the cooler will be placed in the filter of a large pond.

I have been doing experiments with an old frying pan and aquarium pump and an oil cooler. Lessons learned so far:
- mineral oil here is difficult to find at reasonable prices. I am going to use vegetable oil and replace it every year if needed.
-  As expected, the aquarium pump (even though a very powerful one at 1600L/H) was utterly inadequate. I gave almost no flow when the oil was cool, only at ~60C+ did the oil become fluid enough to generate some flow, but still way too little. I ordered this:
http://www.ebay.co.uk/itm/New-12V-DC-CPU-Cooling-CAR-Brushless-Water-Pump-DC50C-1240-High-performance-/120828500685

And I hope that will work better (4m pump head is a fair amount) because the others need 24V and thats no so easy to generate at those amps. 12V can be powered by a PSU. If its still not enough, Ill may just add more.

- air cooled, my small radiator was completely unable to cope with the 1750W frying pan. That too was expected. Submerged in cool water it seems like it should work, although I have yet to generate high enough oil flow to be sure.

- reinforced garden hoses seem to work fine, even at 60+C they didnt become weak. They are cheap (I have to run them a few meter distance to the pond). Dont use clear tubing if you opt for vegetable oil, sunlight is the prime cause for oxidization (=oil going rancid), far more than humidity or air.


Ill post some pics later when the pump arrives and I begin testing with an actual rig.
sr. member
Activity: 252
Merit: 250
Inactive
March 11, 2012, 06:40:39 AM
#20


Sorry, I just have to add on this that refined mineral oil submersion is being used to cool datacenter servers.

It is a workable solution.  If mineral oil was potentially damaging over long term to so many possible board components then they probably wouldn't be submerging rackmount servers in mineral oil baths.

http://www.datacenterknowledge.com/archives/2011/04/12/green-revolutions-immersion-cooling-in-action/


sr. member
Activity: 402
Merit: 250
March 11, 2012, 05:59:31 AM
#19
Also the heat transfer rate sucks for computers which cannot exceed too much the ambient.
If you can have large difference on ambient oil works.
legendary
Activity: 1386
Merit: 1004
March 10, 2012, 10:16:45 PM
#18
Oil cooling is good for all metal items.  It is bad when you add in some plastic items.  Some plastic items will degrade in oil.  It may take a month or a few months but something will go wrong.  It is good for show in computing but not for anything long term unless it designed from the ground up for that purpose.  I know of no computer parts 'rated' for oil.   There are a lot of plastic items in a modern computer. 
full member
Activity: 131
Merit: 100
March 10, 2012, 09:04:48 PM
#17
Oil is not a good cooling solution even with radiators.
If you RTFA you'd notice they are running at near cracking point - the temperature where CPU simply cracks internally due to heat!!

Oil sucks as a cooling medium in this usage. (Infact, in almost any)



Yup, oil cooling totally sucks.  It's just every power company in the world has no clue what they are doing.  I bet you could become a billionaire by selling them water blocks.

http://en.wikipedia.org/wiki/Transformer_oil
sr. member
Activity: 402
Merit: 250
March 10, 2012, 08:55:01 PM
#16
Yup, oil cooling totally sucks.  It's just every power company in the world has no clue what they are doing.  I bet you could become a billionaire by selling them water blocks.

For this task it does.
Just because something works in another scenario does not mean it's a silver bullet for all scenarios.
member
Activity: 77
Merit: 11
Twitter:@watersNYC
March 10, 2012, 08:37:35 PM
#15
Thanks!

It sounds like oil submersion is not viable for efficient Bitcoin mining...

Does anyone disagree?
sr. member
Activity: 378
Merit: 250
Why is it so damn hot in here?
March 10, 2012, 08:31:55 PM
#14
Oil is not a good cooling solution even with radiators.
If you RTFA you'd notice they are running at near cracking point - the temperature where CPU simply cracks internally due to heat!!

Oil sucks as a cooling medium in this usage. (Infact, in almost any)



Yup, oil cooling totally sucks.  It's just every power company in the world has no clue what they are doing.  I bet you could become a billionaire by selling them water blocks.
donator
Activity: 1218
Merit: 1079
Gerald Davis
March 10, 2012, 07:49:32 PM
#13
So what about distilled water to keep this discussion in the spirit of the post? I know that its not a long term solution, but what are the pros and cons.



As in submerge the cards in distilled water?  You will kill the system.  Period.  Distilled water has a very low conductivity but it soaks up ions from contaminants like a sponge.  If anything is less than clean room clean it will increase the conductivity of the DW and kill the cards.

There are fluids which are non condcutive are resist the absorbtion of ions but they tend to be very expensive (more expensive than using solid waterblocks and cheap ole distilled water).
http://www.youtube.com/watch?v=1iz4o3W6IJM

Some early super computers have processing boards which were submersion cooled.
legendary
Activity: 2058
Merit: 1452
March 10, 2012, 07:25:42 PM
#12
So what about distilled water to keep this discussion in the spirit of the post? I know that its not a long term solution, but what are the pros and cons.
Salts will eventually leech in
sr. member
Activity: 402
Merit: 250
March 10, 2012, 07:22:44 PM
#11
So what about distilled water to keep this discussion in the spirit of the post? I know that its not a long term solution, but what are the pros and cons.

Fried system? Cheesy

But i guess if you use enough of electronics varnish .....
sr. member
Activity: 406
Merit: 250
QUIFAS EXCHANGE
March 10, 2012, 07:14:37 PM
#10
So what about distilled water to keep this discussion in the spirit of the post? I know that its not a long term solution, but what are the pros and cons.

sr. member
Activity: 402
Merit: 250
March 10, 2012, 06:06:29 PM
#9
Oil is not a good cooling solution even with radiators.
If you RTFA you'd notice they are running at near cracking point - the temperature where CPU simply cracks internally due to heat!!

Oil sucks as a cooling medium in this usage. (Infact, in almost any)
hero member
Activity: 672
Merit: 500
March 10, 2012, 05:55:04 PM
#8
Done oil before. Huge pain in the ass. Unless you like thai massage parlors.
I would stick with water, oil still needs to be circulated and it is very very very^12 very messy and inconvenient.
sr. member
Activity: 406
Merit: 250
QUIFAS EXCHANGE
March 10, 2012, 05:51:48 PM
#7
And to my note about oil creeping up the cables, if you made a bridge deck to house your pump and rad setup, just connect your cables there and continue the connections on the top.

hmmmm, thinking about all this now.
hero member
Activity: 504
Merit: 500
FPGA Mining LLC
March 10, 2012, 05:46:08 PM
#6
Mineral oil has quite poor thermal transfer capacity, google around about it and you'll notice it doesn't really work for computers at all.

If you see the load temperatures in that article it becomes imminently clear ...
Bubbles.... So basicly, they are moving the heat out with air

So with bubbles I should expect oil submersion cooling to be on par with my water cooling? I'm not looking to keep my cards ice cold - I just want them beneath 60c (so that I can o/c) on load.

I don't think those bubbles were really intended as serious cooling, more as a visual effect. If you want to dissipate a lot of heat 24/7 (e.g. for mining) you should add a radiator (like they did at some point as well).
member
Activity: 77
Merit: 11
Twitter:@watersNYC
March 10, 2012, 05:37:41 PM
#5
Mineral oil has quite poor thermal transfer capacity, google around about it and you'll notice it doesn't really work for computers at all.

If you see the load temperatures in that article it becomes imminently clear ...
Bubbles.... So basicly, they are moving the heat out with air

So with bubbles I should expect oil submersion cooling to be on par with my water cooling? I'm not looking to keep my cards ice cold - I just want them beneath 60c (so that I can o/c) on load.

I could probably re-purpose an old water cooling rig to radiate the oil, or should I just add a bunch of bubbles?
sr. member
Activity: 406
Merit: 250
QUIFAS EXCHANGE
March 10, 2012, 05:21:55 PM
#4
Also, cables have great capillaries properties. After about 3 weeks, you may notice oil coming out the other end of your cat5, ect.

Ive noticed a few 5870 blocks around on the market. Thats where I would go if your going to water cool. Also remember that watercooling a rig is different that a gaming machine. 
Pages:
Jump to: