Intel Uses Mineral Oil to Cool Servers, Finds Success

[tuanmai]

Intel tests submerging servers into oil to cut power and cooling costs.

Intel Uses Mineral Oil to Cool Servers, Finds Success : Read more

 

[CaedenV]

lol, changing the oil every 10 years is a negative? One would think that would last much longer than the servers that sit in the goop.

... wish my car could last without an oil change for 10 years

 

[LukeCWM]

Are the servers continuously submerged? Or just dunked when they need to be cooled? If the former, this sounds pretty cool.

However, if submerged, you lose the ability to place a fan over the hottest components. Hopefully a passive heatsink over the CPU is enough once submerged.

 

[spentshells]

must look into this.

 

[Robert Pankiw]

[citation][nom]LukeCWM[/nom]Are the servers continuously submerged? Or just dunked when they need to be cooled? If the former, this sounds pretty cool.However, if submerged, you lose the ability to place a fan over the hottest components. Hopefully a passive heatsink over the CPU is enough once submerged.[/citation]

You would need to leave it submerged, because once it is removed from the oil, it heats up (quickly). Servers (and all computers) need continuous cooling, not just once every hour, or every time a process is run (processes are always running).

 

[LukeCWM]

[citation][nom]kawininjazx[/nom]That's awesome, I'm gonna go find something to submerge my PC in.[/citation]

I wonder what are the cooling properties of tapioca pudding. Second thought, cooling your PC with something delicious doesn't bode well for the long-term stock of your cooling solution. =]

 

[kawininjazx]

That's awesome, I'm gonna go find something to submerge my PC in.

 

[spentshells]

I know this is pretty popular technique in the oil and gas industry where they can not afford sparks

 

[SteelCity1981]

White Mineral Spirit does a great job cleaning off Mineral Oil.

 

[acadia11]

[citation][nom]LukeCWM[/nom]Are the servers continuously submerged? Or just dunked when they need to be cooled? If the former, this sounds pretty cool.However, if submerged, you lose the ability to place a fan over the hottest components. Hopefully a passive heatsink over the CPU is enough once submerged.[/citation]


Why would you want or even need to put a fan on any of the components you just submerged it in mineral oil? And the mineral oil is far more efficient thermal conductor than air any day of the week.

 

[Guest]

yes, they're submerged 24/7. interestingly enough, though, fans with strong motors can actually run in the oil (as long as they've been designed for it) to keep it circulating, since it's non-conductive.

 

[Zingam]

Yeah, new business! Oil change for computers! Nice one Intel!!! Keep on!

 

[mavroxur]

I wonder if they've formulated an oil that doesn't cause boards to de-laminate after long periods of use. They made no mention of making boards or servers especially designed for this. Back in the day you saw lots of people dunking boards into oil and it worked great for a while, but the PCB's would ever so slightly swell up after years of submersion and become ruined eventually.

 

[TeraMedia]

Once upon a time, Cray sold equipment that had to be submerged in liquid N2. What you bought was just as much a refrigerator as a piece of computer equipment.

I can imagine a pre-constructed, modular-designed "pod" complete with server blades, SSD arrays, redundant power supplies and I/O connections all immersed in a sealed tank of mineral oil with built-in circulators that a company would lease for say 3-5 years. When done, the equipment would get reused and redeployed for a less-demanding task, or recycled for its metals and mineral oil.

 

[jtd871]

Puget Systems sells Aquarium PC kits for adventurous DIYers.

http://www.tomshardware.com/2006/01/09/strip_out_the_fans/
http://www.pugetsystems.com/submerged.php

They claim no problems even with submerged fans. They were cautious with mechanical HDDs, though.

 

[rubix_1011]

The article fails to recognize that you still have to cool the oil- the heat being absorbed by the oil from the servers has to be dissipated or removed, but this is not discussed at all. You cannot simply fill a bucket full of mineral oil, submerge a PC and call it a day. We have these kinds of discussions in the watercooling forum all the time and it simply isn't as simple as passive submersion.

 

[jtd871]

[citation][nom]rubix_1011[/nom]The article fails to recognize that you still have to cool the oil- the heat being absorbed by the oil from the servers has to be dissipated or removed, but this is not discussed at all. You cannot simply fill a bucket full of mineral oil, submerge a PC and call it a day. We have these kinds of discussions in the watercooling forum all the time and it simply isn't as simple as passive submersion.[/citation]

The article does, in fact, mention the CarnotJet cooling system in the 1st paragraph.

 

[alexmx]

[citation][nom]rubix_1011[/nom]The article fails to recognize that you still have to cool the oil- the heat being absorbed by the oil from the servers has to be dissipated or removed, but this is not discussed at all. You cannot simply fill a bucket full of mineral oil, submerge a PC and call it a day. We have these kinds of discussions in the watercooling forum all the time and it simply isn't as simple as passive submersion.[/citation]

As long as the reservoir is big, I think you can; working under the same premise as the passive coolers: big ass CU/AL fins that dissipate the heat. (plus a pump to move the oil)

In the end they might still need AC for cooling the servers but I think that it would be less.

 

[aoneone]

wow so I can go to walmart, but like 20 bottles of mineral oil and a glass fishtank and go to work with my pc desktop?! COOL!!

 

[Vatharian]

I've already tried oil-cooled PC and I've found that cleaning hardware afterwards is fairly easy if you have right hardware. I've used ultrasonic bath filled with isopropyl alcohol and ended with squeaky-clean hardware, ready to use with standard AC. Downside is cost of isopropyl and it needs to be changed often (to clean dozen of 1U server boards you'll need at least 30-50 liters), and ultrasonic bath needs to be in sound-isolated room or everyone and everything will go crazy in quite a radius. In controlled industrial environment it shouldn't be a problem.

 

[CaedenV]

[citation][nom]rubix_1011[/nom]The article fails to recognize that you still have to cool the oil- the heat being absorbed by the oil from the servers has to be dissipated or removed, but this is not discussed at all. You cannot simply fill a bucket full of mineral oil, submerge a PC and call it a day. We have these kinds of discussions in the watercooling forum all the time and it simply isn't as simple as passive submersion.[/citation]
Yes, you still need a cooling solution, but this is hardly a problem. The hardest part of cooling is getting extreme temperatures away from very small areas, which the mineral oil does quite nicely. Once that is done you can do the rest through heat-pipes, radiators, or even some sort of refrigerant, but because the heat is so easily dispersed from the most dense areas, you can use much lower impact cooling solutions to do the cooling, thus the cooling costs of ~2-3% compared to a normal cooling solution in a traditional server of 40-50%. The same amount of over-all cooling (heat dispersion) is required, but it is simply much easier to cool something with a surface area of several hundred square feet (like a radiator), compared to cooling something small like a U1 or U2 heatsink.

 

[CaedenV]

I have always wanted to do an oil cooled PC where it is a tank-within-a-tank system. Have an inner tank with the PC and oil, and then an outer tank full of tropical fish. Sadly, my understanding is that the inner tank would get so hot that it would kill all the fish, so I would need to find some other form of cooling solution to help the fishies out.

Also, I would have to have a PC worth sacrificing to the cause. I'm not about to do it with my i7... and I don't think I can convince my wife to let me try it with her computer... If only I had a gullible neighbor...

 

[JTWrenn]

I always wondered if you could use mineral oil in a standard water cooling rig to make it safer. Then spills would not be nearly as big an issue, just a bit of a mess. Seems like they must not be corrosive if you can dunk a server in it. I need to do some investimagating.

 

[mythostd]

[citation][nom]rubix_1011[/nom]The article fails to recognize that you still have to cool the oil- the heat being absorbed by the oil from the servers has to be dissipated or removed, but this is not discussed at all. You cannot simply fill a bucket full of mineral oil, submerge a PC and call it a day. We have these kinds of discussions in the watercooling forum all the time and it simply isn't as simple as passive submersion.[/citation]

I believe the point of the article had more to do with the energy savings, which was covered,

"the tested servers only needed another 2 to 3 percent of server power for cooling, down from the typical 50 or 60 percent overhead of standard servers. In comparison, the world's most efficient data centers from Google or Facebook still run with about 10 or 20 percent overhead."

Obviously you need to have the rest of the infrastructure to dissipate the heat from the oil, but you need that for the standard air conditioning used to cool server room.

 

[Onus]

Hardware service calls would be a messy nightmare. If this is another "component goes bad, so just toss the whole blade and get another," I want nothing to do with that kind of waste.