liquids

[RangerJoe34]

what is the best non conductive liquid i can use in a watercooling setup?

[Thelemac]

Water. :)

It just has to be distilled.

[CrystalMethod]

If you really have the $$$ to throw around, 3M Florinert™. But, it'd be a kind of waste seeing as you're probably not going to submerge your rig in the stuff. It's the only safe non-conductive liquid I know of.

[r0ckstarbob]

roger that distilled water...

[cowanrg]

agreed. you dont need special 3M liquid... distilled water is non-conductive, and VERY good thermal properties for cooling.

[Crash893]

does alchool (isotopal)

i thought the water would become "stilled" after a few days in the setup.

i know pure water wont conduct but you have to keep it pure

[mnx]

isn't motor oil non-conductive?

mnx

[YMAN]

Originally posted by mnx
isn't motor oil non-conductive?

mnx

Even if it is not, You would need one overclocked pump! :D

------------------------------------

Id say your best bet is 3M Florinert...But extreamly expensive!!!

[r0ckstarbob]

isopropyl (2-propanol) isn't dielectric, i don't know about motor oil but mineral oil IS dielectric, even if it IS thick as hell. definately not good for traditional water-cooling.

distilled water IS dielectric but you're right, the minute that water gets contaminated or dirty, it just became a conductive liquid - and it takes very little to contaminate it.

i guess it depends on what manner you hope to use a dielectric liquid for. for submersion, mineral oil is definately the way to go. for water-cooling, its not good at all.

are you just trying to use a dielectric liquid as a precaution?

you may try doing a search on the web to find a compound or additive that, once added to water, will render water dielectric. that would actually be a nice chunk of information to have here if there was such a substance.

i'll try to do some diggin around in my spare time, but right now there isn't much of that to be had so probably won't have any results for quite awhile.

keep us posted.

RSB

[mnx]

I'm just looking to cool my tbird directly....

I used to use distilled water directly on my duron but the water corroded the surface mount components and their pads and so when I was messing around with it, a few of them got pulled off. (it still worked but would crash once every 24 hours or so and could never make it through a prime 95 torture test at default speed...)

I was thinking of using some really thin motor oil or baby oil, then i could cool the oil with either air or water, and keep my chip at room temp... I like the idea of direct die cooling over using a block for 2 reasons, you don't have to worry about mounting it (just silicone it together and forget about it) and it's way more efficient at transfering heat than any block espically on chips with such small cores...

mnx

[r0ckstarbob]

wow, this topic seems to be something of a fad lately - direct cooling of the CPU... like the idea just suddenly re-erupted spontaneously in various people throughout the country.

direct cooling of your CPU isn't going to work as well as using a waterblock/heatsink. theres a load of reasons for this as i'm sure you know, but most of them have to do with surface area.

whats easier to cool?
one square centimeter putting out 100 watts of heat per second
- or -
10 square centimeters putting out 10 watts of heat per second?

if the goal was to merely reduce the amount of resistance between the heating element and the cooling mechanism, waterblock manufacturers would be making the floors of their blocks as thin as they could. or even better, if it was true, then we wouldn't even use heatsinks in an aircooled system. air is the cooling medium. we'd just point a fan over top the CPU directly.

i hope this doesn't come off as rough, but Direct Cooling is just a flat-out BAD idea guys. theres NOTHING out there that can meet the demand a viscous heat transfer liquid would have to meet in order to make it a viable option in computer cooling... and theres no substance out there that still wouldn't further benefit from the use of a substance that increases the surface area by which the cooling mechanism and the heating element interface with one another producing a MUCH higher heat transfer coefficient.

there isn't even ONE fluid out there with the surface tention, thermal conductivity, or specific heat index to make direct cooling a CPU a viable option. not one.

trust me guys. chuck the idea of direct cooling. i can provide data if you like.

peazzout
scott

[mnx]

quote: "whats easier to cool?
one square centimeter putting out 100 watts of heat per second
- or -
10 square centimeters putting out 10 watts of heat per second? whats easier to cool?"

Your logic is flawed, yes in theory this is true but no heatsink dissipates those 100W perfectly even over it's entire surface... Plus your heatsink and processor are not one solid piece of metal, and therefore heat will not transfer between them as efficiently as water flowing over the core directly could...

quote:

"trust me guys. chuck the idea of direct cooling. i can provide data if you like."


How about I provide data from the direct die rig I actually had this summer for over 1 month straight (before I decided to pull it apart and accidently wreck my duron...)

Duron 700MHz @ 1.575V running d.net client for many many hours...

when cooling the water (fan / sponge method)

room temp 22C
water temp 24C
cpu temp 25C

A quick run of radiate reveals a c/w rating of .13... Sounds pretty good to me...

When I didn't cool the water at all (just let it cool itself in the bucket:

room temp 22C
water temp 29C
cpu temp 30C

A c/w rating of .26 still is pretty decent...

mnx

[r0ckstarbob]

this is the message i got from Dave Smith (myv65) when i PM'd him about this a couple of months ago...

am glad to see that you got good performance out of your system the way you did, but i still believe that your use of a copper block would have been of great bennefit to your setup. i mean, i'm glad you were able to MAKE it work, but theres easier and more efficient ways go go about it, and by in large, the use of a metal medium is going to be of the utmost importance.

my question:
ok, question. when we cool - are we cooling the metal and letting the metal cool the subject? or are we merely trying to reduce the amount of resistance between the subject and the cooling mechanism itself?this question has a direct impact on design specifications.


myv65 wrote on 09-17-2001 08:23 PM:
here's the gist of it.

Convection through a metal such as copper is much more efficient than convection from a solid to a liquid. The CPU puts out all its power through a surface area roughly 117mm^2 (the die). This is nowhere near enough surface area for a liquid to get a hold of that heat via convection.

What the heat sink metal does is take all that heat in through a surface area of 117mm^2 and conduct it throughout its volume. The heat flows through the heat sink from higher temperature regions to lower temperature regions. In a typical air-cooled heat sink, the ratio of surface area to metal volume is very high, ie lots of really thin fins or small diameter pins. These thin fins or small pins don't conduct heat very well, but provide copious surface area for convection to do its thing. Conductive efficiency is sacrificed in order to increase convective efficiency. The goal is to balance the lost conductive efficient with the gained convective efficiency.

If you want to visualize this qualitatively, graph a curve where y = 1/x * 1/(1-x) between the x values of 0.01 and 0.99. "x" represents the "split" between conductive and convective efficiency. When x is near zero, the conductive efficiency is very high, but convective efficiency is very low. Vice versa when x nears 1. "y" represents overall thermal resistance.

The reality is that the relationship is much more complicated, but that should give you the general idea.

... continued
myv65 wrote on 09-17-2001 08:31 PM:
With liquid cooling, the general situation is unchanged. What differs is how you split the relative efficiencies. Water convects roughly 20X better than air. This means you don't need to go to the same extremes sacrificing conduction through the metal to get good convection. You still require the metal to get the heat from that little die and spread it out some, however.

Even with your very cold coolant, you will need a block of some sort on top of the CPU. For this block, I would stick with copper; although, aluminum would also suffice. I'd tend to stay away from stainless steel. It's heat transfer is truly abysmal and it also has a higher coefficient of thermal expansion. While it's very unlikely, it is possible that the block could shrink and expand placing a shear force on the die that would chip a corner. You'll be assembling at rooom temperature. When you cool down, the block will shrink and try squeezing the corners of the die inwards. If you go back to room temperature, the block will expand and try to push the corners outwards. This expansion/contraction is very, very minor, but a lot more on stainless than other metals.

If you are making your own block for the CPU, you really need to understand its purpose and how it achieves it. I don't want to get into this too much unless you're heading down that road. If you can find a good ready-made block, it will be much easier than making your own.