Sign in with
Sign up | Sign in
Your question

Maintainable Sub-Zero Cooling: Issues and Solutions (please comment)

Last response: in Components
Share
January 28, 2012 4:38:21 PM

Sub-Zero Cooling: Issues and Solutions (please comment)

Background:

I am not a gamer but rather a financial portfolio manager who performs investment quantitative analysis. I regularly write and execute code that will consume 100% of a core’s capacity for days on end. I am limited by the number of operations I can process due to the fact that adding a single additional operation can result in an exponential increase in processing time (taking days to weeks).

As such, I have recently spent a fair amount of time looking into extreme over-clocking and system cooling. Over-clocking is relatively straight forward, but excessive system cooling has a number of limitations, particularly in the realm of sub-zero cooling. Having an extensive IT back-ground I know that any system needs to be not only effective, but also supportable and maintainable.

LN2 cooling is anything but supportable and maintainable. Water cooling is both supportable and maintainable, but any temperature below the dew-point results in condensation. Insulating against condensation is neither easy nor fool-proof, particularly if you are cooling multiple system components (i.e.; cpu, memory, bridges, video cards, drives, etc.). Mineral-oil submersion cooling is marginally supportable and maintainable, but its major drawback is that mineral-oil has only 23% (I believe) of the heat conductivity of water. However, mineral-oil is a water-dispersant. A motherboard submerged in mineral oil will be well protected from condensation.

A Possible Solution??

I would like to propose a water-cooled system in which the water (and the necessary additives) are cooled well below the dew-point via the use of a heat-transfer unit submerged in a chilled substance which has its temperature maintained by a refrigeration unit (construction of your own choice; there are hundreds of options. Simple Freon systems can create temperatures of -36C).

To combat condensation, the water-cooled motherboard would then be submerged in mineral oil. Any number of steps could be taken to prevent heat-exchange from the mineral-oil to the cooled water including simple insulation and/or cooling of the mineral-oil. However, knowing that mineral oil is not a good conductor of heat we do not have to worry excessively about the loss of coldness into the mineral-oil.

With the correct additives added to the water, this cooling system would allow us to pump sub-zero coolant to the system components without concern of condensation.

Yes, I could always build a Beowulf cluster and write multi-treaded code from the ground up. I do however think this would be more difficult, costly and time consuming.

So, before I set-out to build such a box, I would love to hear what you-all think would be the key obstacles to success. What am I missing, what am I forgetting, and what am I failing to understand?

Thanks so much! Sorry if this has been previously suggested :) 

Trip
January 30, 2012 6:47:08 AM

dives said:
Sub-Zero Cooling: Issues and Solutions (please comment)

Background:

I am not a gamer but rather a financial portfolio manager who performs investment quantitative analysis. I regularly write and execute code that will consume 100% of a core’s capacity for days on end. I am limited by the number of operations I can process due to the fact that adding a single additional operation can result in an exponential increase in processing time (taking days to weeks).

As such, I have recently spent a fair amount of time looking into extreme over-clocking and system cooling. Over-clocking is relatively straight forward, but excessive system cooling has a number of limitations, particularly in the realm of sub-zero cooling. Having an extensive IT back-ground I know that any system needs to be not only effective, but also supportable and maintainable.

LN2 cooling is anything but supportable and maintainable. Water cooling is both supportable and maintainable, but any temperature below the dew-point results in condensation. Insulating against condensation is neither easy nor fool-proof, particularly if you are cooling multiple system components (i.e.; cpu, memory, bridges, video cards, drives, etc.). Mineral-oil submersion cooling is marginally supportable and maintainable, but its major drawback is that mineral-oil has only 40% of the heat conductivity of water. However, mineral-oil is a water-dispersant. A motherboard submerged in mineral oil will be well protected from condensation.

A Possible Solution??

I would like to propose a water-cooled system in which the water (and the necessary additives) are cooled well below the dew-point via the use of a heat-transfer unit submerged in a chilled substance which has its temperature maintained by a refrigeration until (construction of your own choice; there are hundreds of options. Simple Freon systems can create temperatures of -36C).

To combat condensation, the water-cooled motherboard would then be submerged in mineral oil. Any number of steps could be taken to prevent heat-exchange from the mineral-oil to the cooled water including simple insulation and/or cooling of the mineral-oil. However, knowing that mineral oil is not a good conductor of heat we do not have to worry excessively about the loss of coldness into the mineral-oil.

With the correct additives added to the water, this cooling system would allow us to pump sub-zero coolant to the system components without concern of condensation.

Yes, I could always build a Beowulf cluster and write multi-treaded code from the ground up. I do however think this would be more difficult, costly and time consuming.

So, before I set-out to build such a box, I would love to hear what you-all think would be the key obstacles to success. What am I missing, what am I forgetting, and what am I failing to understand?

Thanks so much! Sorry if this has been previously suggested :) 

Trip


I'd think a hermetically sealed container with a powerful dessicant and the air replaced with N2,(not liquid, just to displace water vapor) might be easier/less messy. How far below 0 do you want to go? I can get 4.9Ghz stable on my i7 3930k folding@home with an h100 and ambient room temps of 16C. cpu never tops 35C.

Another option if multicore is possible, but not a cluster, are the new Interlagos AMD chips. You can put 64 cores in one computer for around $3000. I am actually planning on getting a quad G34 board in Feb, then adding 16 core cpus as my needs increase.
m
0
l
January 30, 2012 12:26:00 PM

Tului,

Thanks for the reply. That's a great idea. Any idea how to make the holes where the cables enter and exit air-tight? I would imagine that there would also have to be a couple valves by which to vacuum out the air and then introduce the N2.

I do have a 16 cpu box co-located at the Chicago Mercantile Exchange. The issue with multi-treading right now is that I instantiate an Excel 2003 object (which is MUCH faster then Excel 2010) as the back-bone of my calculation engine. It is single threaded. So, I need to focus on getting a single core running at the fastest speed possible, or start writing a LOT of multi-threaded code (I think building a box would be easier).

Cheers! - Trip
m
0
l
Related resources
January 30, 2012 11:48:41 PM

This is an incredibly interesting topic...please let us know how this goes...very curious on your findings.
m
0
l
February 1, 2012 2:07:54 AM

Seal the wires with silicone or draft-stop like material. Personally I'd opt for a lexan box inside another lexan box with the inner box wrapped in thinsulate. I'd probably leave the tops lightly sealed and/or on hinges to allow access. You could use a schrader valve or even a PVC valve I suppose. If you got a high quality dessicant, it'd pull the moisture out of the air pretty quick regardless.

Another option is putting a "mini-split" air conditioner in a small room or large closet and using 1U, 2U 3U or 4U server EEB or EATX racks mounts to house the computer. They have excellent ducting inside to direct the flow to the CPUs as well as redundant power supplies. I think Supermicro's cheaper 4U's with 24 hot swap SAS/SATA drive backplane starts around $3000. Put a 4P Interlagos system in there and add CPUs as you need the power.

Could you not migrate your application to Access or better yet SQL or mysql? I don't know the workload you're running but I imagine a real database would scale better for feeding a hungry application all the data it would need.
m
0
l
February 1, 2012 2:14:25 AM

Wow. What a topic. Anyways what are you going to use as a case?
m
0
l
February 8, 2012 3:14:21 PM

Thanks for all the thoughts guys! As for what I'm going to use as a case depends upon what method I use for mitigating condensation (haven't decided yet). But I'll be sure to keep everyone posted once I start the build.

I do have one other question for the group. In that I plan on using sub freezing temperature in the water cooling components, what types of "anti-freeze" can I use that will limit corrosion, is comparable with soft PVC tubing, and won't reduce the thermal capacity of the water too much? I know vegetable glycerin is a potential additive. Any other thoughts?
m
0
l
February 8, 2012 5:17:11 PM

This topic has been moved from the section Overclocking to section CPU & Components by Jpishgar
m
0
l
February 11, 2012 2:39:31 PM

Hello Jpishgar,

I was just wondering you have decided to move this tread to "CPU & Components" when the entire contents is about maintainable sub-zero cooling for extreme overclocking?

All the best,

Trip

jpishgar said:
This topic has been moved from the section Overclocking to section CPU & Components by Jpishgar

m
0
l
!