>REAL< liquid cooling

[maelstrom]

Being more of a geek than a speed freak, I decided I should do something a little different with my computer.

I decided to sink the whole thing into a chilled mineral oil bath.

So i've built the case (1/4" acrylic and aluminum channel) and filled it up with 8 gallons of mineral oil. I've got a DC pump that will cycle the stuff at around 400gph using just the PSU, and I've got a peltier mini-fridge with a liquid resivoir in the loop to cool off the oil before tossing it back into the mix. And the fridge runs off the PSU also.

Of course, now I need to do some planning.

Any ideas of the best way to interface a liquid bath with the CPU/GPU/etc? I'm going to get one of those Zalman fan-lookin' heatsinks for the main processor (just cause it looks damn good), and there's a solid copper sink I saw for the GFX card that should work great. After that, I have no idea where i'll go.

I'm thinking I need some ducting or something to make the flow go across the CPU, GPU, Northbridge and memory, plus whatever else I want to throw in. I don't really see that happening right now, but I figure the circulation should be good enough to keep everything from burning up. (even though 400gph is less than 1cfm)

Has anyone done this to know if i need ducts/etc, or if I can just get away with IN and OUT?

I'll try and post some pics, and you can read more on my trials and tribulations here (http://www.dullsville.com/liquid.php).

[Farabomb]

:beerm8s: Welcome maelstrom :beerm8s:

Give a look here (http://www.liquidninjas.com/reviews.php?op=showcontent&id=15) for some of our members forays into goo cooling.

I can't offer much more than that but glad to have you posting .:)

[maelstrom]

It's actually one of the goo links that brought me here!

dicki (http://www.liquidninjas.com/bbs/member.php?s=&action=getinfo&userid=12) was using transformer fluid for his/her media, and mentioned something about the insulating ability as being a plus. (BTW, happy nearly-birthday dicki)

In any kind of thermal transfer fluid, shouldn't insulation be a negative aspect? I'm still trying to get the technical terms, thermal capacity is the amount it can hold, I guess thermal conductivity would be the property i'm getting at.

Is a fluid with high thermal conductivity more effective than a fluid with the same thermal capacity but lower conductivity? (My physics are WAY behind on this stuff, but this (http://hyperphysics.phy-astr.gsu.edu/hbase/heacon.html) sort of helps me explain.)

[maelstrom]

okay, saw the pic... guess it's "his media".

sorry dude

[evilmatt]

hi there,

I think minty_altoid came up with most of the physics bit.

I don't remeber mentioning anything about insulation. The main features that you need as I understood it were specific heat capacity and thermal conductivity.

My understanding of it was that the specific heat capacity tells you the rise in temperature of the goo for a given heat input.

I think of it as deg c per watt in goo but I don't know how acurate that analogy is.

the conduction would be the rate at which it takes heat from one place to another.

so if you have a high conductivity it will shift the heat fast from hot to cold. If the specific heat capacity is high it will take more heat energy to raise its temperature.

so in answer to your question I recon that all other things being equal the higher thermal conductivity would be a good thing as it would equalise hot to cold quicker.

The comparison of the midel to mineral oil was that it had a slightly lower conductivity but higher specific heat capacity such that it would conduct slower but require more heat to cause it to rise in temperature. There was also some issue with sediment with mineral oil but I can remeber what it was.

I think thats more or less the gist of the physics but I'll ask minty to give it a check when he comes back form holiday.

I think the major problems we had was the pump not being up to the job and the cooling loop not taking enough out of the goo on each run.

We used a zalman because we figured it'd give good contact with the goo.
I recon pretty much anything would do the job so long as you get the goo flowing round it ok.

Dicki has suggested doing a direct die cooling method which probably would work given sufficient flow if we had some suitably expendable hardware we might try it ;)

we had the outflow of out cooled loop pluged directly into our zalman so the coldest stuff was running though the cpu then put the inflow at the edge to try and get a flow going across the major components.

Anyway best of luck with your project

hope my ramblings have helped ;)

Matt

p.s. dicki as a woman oooo thats a scary thought, I can't even bring myself to picture it it'd give me nightmares :D

[dicki]

thanks maelstrom :)

well lets see...

the goo cooling "incident" was a great idea but we didn't take it anywhere near far enough. it looks like your working on the next step which is to add active cooling to it. the only thing i'll say is i'm not sure that mini fridge will be man enough for the job but give it a go.

as evilmatt points out we had issues with the pump not really being strong enough to shift the goo especially when the goo was at room temp when its quite thick this resulted in a low rate of flow through the radiator and then back into the tub and over the cpu...

i think any heatsink (without fans obviously!) which has quite an open design would be good. remember that you will need to move the liquid around a lot otherwise the hot liquid will collect around the heatsinks and limit the cooling ability (i think this is what i meant by insulating?)

anymore questions?

[maelstrom]

I think I was confused by the description of MIDEL, it's described as an "insulating fluid" and I thought that was thermal.

From what I can tell, the mineral oil that is compared to MIDEL on the spec sheet (http://www.midel.com/pdf/ComparativeData.pdf) would really be more ideal in a computer cooling system. It is more thermally conductive (spreads heat faster), even though it has a lower specific heat. I still haven't figured out which is more important, but my view is conductivity is just because that's the entire point of a heatsink. I'm not generating enough heat that the liquid isn't going to be able to absorb it, so I just want to "share the love" as it were.

The biggest selling point to me is that it has about 1/3 the viscosity, so the Eheim could probably throw it around almost like it was water. Please note that mineral oil comes in MANY different formulations, it can be found in viscosities from around 9 or 10 centistokes up through several hundred cst.

The sediment problem evilmatt saw might have been water (condensation), it tends to collect at the bottom of the case. This is especially annoying because I used a water-based fluorescent marker to dye the oil, but it settles out into little drops at the bottom after only a couple hours. I'm going to be trying some strata down there, will probably start with paper towels or some kind of absorbent cloth, but I'd like to eventually put some silica crystals down there to suck up all the water.

And no, the pelt fridge is definitely not man enough. As I mentioned somewhere, it takes overnight to chill eight cans of your favorite carbonated beverages to "cold enough" from room temperature. It's mostly a proof-of-concept sort of thing, like I have to squish my resivoir bottle to make it fit inside the fridge.

I still have my real, compressor-based mini-fridge from my dorm days, and since I could do nothing but freeze everything in it, I think it might be seeing the business end of a large drill sometime soon.

Does anyone know of a super-tiny submersible pump? I'm actually thinking of having one main pump for the radiator part, and one or more smaller ones to direct goo across the main components (even though that defeats my objective of reducing my EIGHT fans down to one pump...).

[ralf_c]

two things that crossed my mind. for a tuff job like that i think a heavy duty flood pump will do. and on a related note, moving the goo around will also need something with some gonads. maybe a Milwaukee heavy torque power drill.

[Synthohol]

heres .02 out of the blue!!
there is an electric waterpump for a 350 chevy block that runs on a car battery, consider a automotive radiator with electric fan, add the two together with a well kind of like an undergravel filter in a fishtank, it will suck the goo (eew) go through the fan cooled rad and pumped back to the tank probably aimed at the cpu and such.
there should be enough cooling to dunk the psu as well in the mix!!
just an idea:)

[Kabooka]

What ever type of pump you get to, it needs to have closer tolerances than the aquarium pumps we use, not impeller driven but maybe a vane pump... or a gear pump...

Just a thought.

[maelstrom]

As far as pumps go, I have a 12VDC 450GPH pump from pumpworld.net (apparently discontinued?). It might not be enough horsepower to cycle super-fast, but it's good enough. Remember the Ehiem-waterbucket picture from the goo incident? That's what this looks like with the light mineral oil.

If I had to replace my current pump, the 1670 (http://www.pumpworld.net/dcpumps.htm#J195012VDC) would probably be more than my fittings could stand, but in a good way. I might poke around in the junkyard around the corner, see if there's a suitable 12V waterpump attached to a halfway-decent radiator anywhere.

I'm still sticking with my original question though, does anyone have a good way to interface the goo with the parts what need to be chilled? I like the concept of the "undergravel filter", that would actually solve a couple of problems (condensate collection and better circulation than i've got right now), but i'm still worried to juice this sucker up and let 'er rip because I don't have any way to make sure the goo gets where it's needed. I've got a vertically-oriented mobo, ports, slots, etc. all stick out on the top (see here (http://www.abit-usa.com/images/products/kg7r-large.jpg) for my exact mobo and orientation). Everything is on the right, and gfx card is in the AGP slot, so that may make routing the flow a little easier. Basic idea right now would be to have the pump/radiator inlet on the left, and the output on the top-right, pointing toward the left across the CPU. I might need to put a barrier of some sort to the left of the gfx card from the top down to the bottom of the memory so that everything flows from the top-right all the way down to the bottom before exiting to the radiator on the left (in order, across CPU, GPU, NBridge, mem).

Going on a short holiday and moving offices this weekend, but i'll probably get something dipped before next Sunday night, that'll show me if i'm really going to have these problems.

(these posts seem to be really long, hopefully not boring anyone! :D )

[Kabooka]

Seems like one problem is that you have to control the "current" of the liquid to hit the areas of concern.... why not have the board suspend just under the surface, with whatever heat sink attached exposed, and then cascade the fluid directly on what you want to cool, you'd get the benefit of overall cooling on the board itself, and spot cooling where you need it ...

[maelstrom]

:doh: now you tell me! :)

I think I went with this design originally because I saw one of the Lian-Li cases that had a mobo drawer, and it seemed like it would be great to take that and use it as the base for this.

I'll probably fire it up as-is, then if it does/doesn't burn out i'll adjust from there.

[prokaryote]

A little off topic, but

Have you thought also about industrial oil pumps or even bilge pumps. May consider a water cooler block without the top (or bottom plate on) and place over the die. Have the coolant go into the block and have a short tube on the exit dumping into the top layers of mineral oil. Oil intake near the surface to capture the warmer oil. Put an inline oil filter in the loop and replace filter with a desicant?

Graingers is a great place to get odds and ends. Used it all of the time when working for the dept of Ag. Here (http://www.grainger.com/Grainger/wwg/homepage.jsp) is a link to their home page. Search on oil pumps or some such.

we now return you to your regularily scheduled programming

[maelstrom]

Grainger definitely rocks, I actually know someone who left my company to work for them, lucky SOB. My brother pointed me toward McMaster-Carr (http://www.mcmaster.com) for the same kind of stuff, but mostly both of those are out of my price range for the kind of pump i'd want to get.

I like the idea of the inline dessicant/filter thing, that'll have to be part of my 2nd generation model. I still need to find the stuff, I just haven't made it out to Home Depot yet.

This may sound like a strange question, but is it possible that i've built a Van De Graaff generator? I only mention it because there seems to be some static between the acrylic case and the plastic resivoir, and this has the same function as a VDG: pass one non-conductor across another one repeatedly at high velocity...

[minty_altoid]

Just to clear one point up, Midel is electrically insulating, not thermally insulating. It's actually a very good heat conductor...

The difference win conductivity between midel and mineral oil is not much really... but the lower viscousity of the mineral oil is a plus. Our current plans involving goo don't involve pumping it at all, just using it as a heat transfer medium and condensation preventer.

We did a bit more on the goo project yesterday... I'm sure Dicki will post a teaser pic once he has his pc that runs the camera back together (dicki - probably the res without mobo). Until we get the images and write the review my lips are sealed ;) but I will say we did see '-27C' on a meter.....

minty