For all you watercooling nuts out there - hard data

[r0ckstarbob]

hey y'all, thought i'd throw this up here. figured you guys would like it and could use it. asta.

yeah i'm new here to the Pro Forums. long time regular of the AMDmB.com and Overclocker.com forums, also over at the [H]ard|forums blah blah blah whatever... i am not my post count. good info is good info, ultra newbie or not. hope you don't take my seniority (or lack there-of) into consideration when you read this.

at any rate, i thought you guys would find this useful. have posted it on the other fourms i'm on and figured you kids would like the info too. heres whats up.

i've been tabulating data on coolants and liquids used in computer water cooling for the past 3 weeks and these are my results. This is more or less cut and pasted word for word from the other forums.

please keep in mind my focus for this research was to find a coolant that would work at -30F for my project The Core Project (new site) (old site here). for those of you considering putting antifreeze in your system, going the sub-Z route, or just want to put your coolant below freezing (32F), this is something you'll probably be interested in.

as for using antifreeze in your system, well, the numbers speak for themselves. Basically... DON'T.

big props go out to Karsta (on the Overclocker.com forums) and myv65 (Dave Smith on the AMDmB forums- the author of that AMAZING article on Thermodynamics in regards to comptuer cooling). both of whom i've been corresponding with and have been absolutely fantastic in providing me with answers and overall knowlege when i just plain ol got stuck somewhere, trying to understand all this. these are some REALLY knowlegeable and helpful guys, and not only in regards to coolants. they've been absolutely great. i can't express enough gratitude. big props also go out to Tenax too (also on the AMDmB forums). All three of them have been of invaluable aid. Thanks guys.

i can explain anything that you might have questions on here. and if i can't, i know a couple of guys that can.

here ya go. it's all on a GIF image and only 32k so you can easily save the image for your own personal reference if you like.



backup server



if you see something that you know for a fact is amiss on the charts, please feel free to point it out.

i hope this is helpful.

RSB

PS yes i'm aware that the chemical makeup of normal Air isn't O2... i just cheated there. Glycol refers to Ethylene Glycol Antifreeze using numbers provided by Prestone themselves for their most potent antifreeze; and Isopropyl actually refers to Isopropyl Alcohol or 2-Propanol (aka Rubbing Alcohol).

if you have questions, you might check the other links below. theres a damn good chance that someone in one of the three forums have asked it already. failing that just drop yer question here and i'll field it to the best of my ability.

http://www.hardforum.com/showthread....hreadid=251405

http://forums.overclockers.ws/vb/sho...threadid=39039

http://www.amdmb.com/vb/showthread.p...threadid=64169


enjoy.

[r0ckstarbob]

graphed data comparisons


Freezing Point: (lower the better)




Thermal Capacity: The Thermal Capacity -aka specific heat- figures are used only in reference to the Thermal Conductivity and is used to find the Thermal Differential.
The Thermal Capacity is how much Thermal Energy these substances can absorb/store.




Thermal Conductivity: The Thermal Conductivity rates how quickly these substances can absorb and transmit thermal energy and generally has the most direct bearing on cooling for our purposes.
(The higher the better)

[r0ckstarbob]

graphed data comparisons (cont.)

Thermal Differential: The Thermal Differential is the overall rating of the substance based upon the Thermal Capacity and the Thermal Conductivity of a substance. When it's all said and done, this is the chart/number that has to be looked at when putting the whole schebang together. This is the sum of a substances total effectiveness as a cooling medium (minus viscosity that is)
(the higher the better)

(higher the better)




This is the Dynamic Viscosity ratings of these liquids. Dynamic Viscosity is the measure of resistance of liquid in a tube or capillary. It's measured by pouring liquid in one end of a tube and measuring how quickly it reaches the other end. It is represented in Poise. The industrial standard uses Centipoise (cP). 1 Centipoise = .01 Poise.

the lower the number = the thinner the liquid.
the thinner the liquid = the better the viscosity (for our purposes that is)
the lower the viscosity = the faster we are able to pump our coolant.
the faster we are able to pump our coolant = the more efficient our cooling mechanism.

low numbers goooood. high numbers baaad.


The other way to measure viscosity is called the Kinematic Viscosity. It is measured by dropping a sphere into a bucket of liquid and measuring how quickly it reaches the bottom. It effectively measures the amount of drag upon an object passing through it and is measured in Stokes. The industrial standard is Centistokes (cSt). I haven't provided those figures because the Dynamic Viscosity is the important one for our purposes. But as with the Dynamic Viscosity, the lower the rating the better.

the Freezing Point, the Thermal Differential, and the Viscosity are the crutial figures to look at when making your decision.

Just a note here: the term "Glycol" in all these figures is short for Ethylene Glycol Antifreeze - Prestone to be exact - using numbers from their most potent antifreeze that they supplied me upon request via email.

The term "Isopropyl" in all these figures refers to Isopropyl Alcohol (rubbing alcohol aka 2-Propanol).

and Air is labled 02 in the chemical makeup. thats incorrect. it's just normal breathable air.

Keep in mind that water is the best stuff around for heat transfer. The closer we can get our liquid to the efficiency of water, the better. Water is the standard in all these figures and the standard in which we are trying to meet or exceed (when possible) when choosing an effecient coolant.

and as always, my knowlege is far from perfect so if you see something that i missed, please feel free to backchannel me at language@speakeasy.net and we'll get it all fixed up.

[r0ckstarbob]

below is the graphed data giving direct comparisons between the thermal properties of Antifreeze and Methanol at different ratios to water.

I should note here however that nowhere did i actually find these figures, these are figures that i extrapolated based upon the data i have compiled. it should be relatively accurate, or at least close enough to more or less predict efficiency within 5 or 10% of true.

What we're looking at here are two very comparable substances. The differences between them are often times negligible, with Methanol generally acting a little bit better then Antifreeze. Where the differences REALLY become evident is in the Viscosity. The Viscosity of Methanol is what really gives it a HUGE lead over Antifreeze from the word go; and the colder it gets, the more apparent that fact becomes. Antifreeze is most definately a great product for cars which operate at 200+ degrees where viscosity doesn't become much of an issue in that kind of heat - the heat thins it out. Methanol would be terrible in a car because of it's extremely low evaporation point. Alternately it's also why Antifreeze is going to be terrible in your computer. Viscosity becomes a BIG issue in computer cooling, especially the colder you go. With computer cooling, the vapor point doesn't often become much of an issue and we don't have much to worry about there (it doesn't take much to reduce that worry to "nil" with a decent radiator in fact). but that Viscosity, man thats the ticket. Water thins out the Antifreeze. Methanol thins out the Water... and the thinner our coolant, the faster we can pump it. the faster we can pump it, the better heat transfer we can effect. Thats how it works kiddies. The fact that it's so potent and has such a low freezing point means that we don't have to use as much of it either which keeps our water to substance ratio considerabley higher with Methanol as opposed to Antifreeze.

Antifreeze should be called Antiboil. thats where it works best and where the focus was on when they designed this stuff initally.

Antifreeze is good for HOT environments
Methanol is good for COLD environments

try to swap them out and you're going to get some really inefficient performance.

anyhow, here ya go...

Graphed data comparisons between Antifreeze and Methanol at various ratios.

Freezing Point: (Lower the better) IMPORTANT CHART




Dynamic Viscosity: (Lower the better) IMPORTANT CHART
low numbers goooood. high numbers baaad.




Thermal Differential: (Higher the better) IMPORTANT CHART




Thermal Conductivity: (Higher the better)




Thermal Capacity: (reference Thermal Differential chart)




the Freezing Point, the Thermal Differential, and the Viscosity are the crutial figures to look at when making your decision.

I'm sorry that these are all on a vertical plane. i tried to make them as easy to read as i could, but with so much information, i had to do them all vertical otherwise they'd run way off into the borders.

[r0ckstarbob]

and last but not least, a note on Surface Tention and The Barrier Layer...

i have this same post over at the [H]ardforums and Overclocker.com fourms (which incidentally also got stickied to the top of both of them - YAY) the post is even under the same name and is about word for word the same thing. incidentally this post is also over at the AMDmB forums too.

at any rate, it's been really cool to see the different responses and see the different questions come up. this one is from the [H]ardforums. someone asked about how surface tention and the boundry layer all came into play so thought i'd post it over here for general overall knowlege. i suspect someone woulda asked the question sooner or later anyhow, so perhaps i'm beating it to the punch.

enough babblin. here goes.

***

Originally posted by Amadon
This is some of the best data I have seen posted on the forums.. ever.
Thanks a million bro.
Although I'm not one to go in for sub ambient cooling, mainly because of the maintinence and that I'm paranoid, my following topic applies to both watercooling without a phase change system or peltier units.. and to the opposite.

The thing that I haven't seen addressed is surface tension.
I know very little about it really. But I assume that it is not caused by air "touching" the water. I DO know that if the surface tension is broken.. say.. with soap, that the water flows more freely over a surface.
Couldn't one conclude then that a lack of surface tension, at least in thin (as in depth) water creates a more favorable viscosity?


I always thought this was probably how water wetter helped.

Tell me if I'm wrong. or onto something.

Amadon

okay this is what i've gathered. this is in part from an extremely helpful friend with a PhD in chemical engineering (whom i've been in correspondence with about this) and is a summary of what i've learned about surface tention and the boundry layer having been put into my own words more or less.

theres surface tention and theres the barrier layer. whats the boundry layer you ask?

when a fluid moves across a non moving surface, there is a point where the fluid comes into contact with that surface. (obviously). we'll call the surface the Block.
when liquid moves over the block, the surface of the liquid reacts against the friction of the Block where they contact one another and the motion on the surface of the liquid slows down or stops at that contact point. this is called the boundry layer.

as we know, moving liquid quickly across a surface is what provides us the best heat transfer performance. when any or all of that liquid stops moving and sits there due to the friction it's encountering, it's suddenly doing a less efficient job. the liquid now composing the boundry layer is relying on normal convection to warm itself up enough so it rises off the block (and thus out of the barrier layer) and to be swept back into the waterflow. it's only when it warms up enough to leave the boundry layer due to normal convection that it's replaced with something cooler and the cooling cycle can continue.

creating a small amount of turbulence on the surface of the block will prevent the boundry layer from forming and thus increase the Heat Transfer potential of your apparatus. this can be achieved by soldering little nubs on the surface of your water-passages or simply by roughing it up with small perpendicular grooves in the surface of the metal (such as with a file or rasp). you can also run a small wire or spring (if the channel is big enough) down the length of your tubing if you're relying on your tubing to provide you with some of your cooling ability. (anyone not using copper piping for this last technique should forget it as rubber/nylon/vinyl/silicone tubing isn't heating or cooling your liquid one way or the other and theres really no point - on a slightly different note, this technique might help you guys with radiator problems looking to grab another couple of degrees of cooling power from it. i haven't tested it, but in theory it should be a decent idea).

basically you just want to find a way to interrupt the water that crosses across the surface of the block to prevent the boundry layer from forming.

summary: the boundary layer impedes heat transfer and it becomes thicker with increasing fluid viscosities.

However, the surface tension is a little different in that it reflects how much a fluid spreads out on a solid surface (i.e. low surface tension) vs beading up on a solid surface (i.e. high surface tension) if you wait long enough (equilibrium). The viscosity will affect how fast(rate)the fluid will spread out or bead up. The surface tension is the property which reflects the fluid's ability to wet a surface.

Even if a fluid has a low surface tension and can potentially wet a surface very well, it is not good for heat transfer if the fluid is too viscous and slooows its speed to wet or be replaced by itself with new cooler fluid
(surface renewal). Thus, even though you need the surface to be wetted (low surface tension) you also need it to be easily replaced (low viscosity).

That's why glycol sucks, incidently, compared to water or water/methanol as a transfer fluid in cooling applications. Glycol finds it specific use when
you need to increase the boiling point as well as use it as a heat transfer medium as in autos.

basically by breaking up the surface tention you're increasing the heat transfer potential of your liquid by making it less sticky to itself. you're essentially making the water more fragile and easier to break up which allows for better turbulence on the surface against the block.

use a combination of the two methods to break up the boundry layer on the block and decrease the surface tention of your liquid and you're rockin like dokken.

basically the boundry layer can be addressed by altering the surface of the block. the surface tention can be addressed by treating the liquid. and both techniques can and should be used in collaboration with one another to recieve the most performance out of your heat exchanger AND your coolant if at all possible.

hope that explains things a bit.

PS - a cheap and perfectly effective 33:66 ratio methanol/water mixture can be commonly found (without additives) in certain brands of windshield wiper fluid, you just have to look at the contents.

[ECUPirate]

thanks for the data, bob.

question: You've established that meth is better than glycol for our applications. Great. But how does it react w/ the pump and tubing, etc.? I don't know much in this area.... Will the methanol negatively affect the plastics, or wear down the pump... cause leaks.. whatever? What about its effectiveness at preventing galvanic corrosion?

[r0ckstarbob]

when mixed with water it's PH balance drops way through the floor, though even in it's pure state it's not too bad. there doesn't seem to be any problems with corrosion, does not adversely affect the pumps or the tubing noticeably, it is thinner than water so you will have to take a little extra precaution and be careful about leaks, make sure that your fittings are good and snug. beyond that, it's good to go. it does not impact the effects of galvanic corrosion one way or the other. you still can't use a copper block with an aluminum radiator.

[ECUPirate]

I don't. I was mainly concerned about the pump. I suspect methanol is a helluvalot cheaper than water wetter, also.
You mentioned windshield wiper fluid? Is that typically just meth & water? Can you buy straight-up meth? What do you use?

[EMC2]

Bob, would you prefer private messages?

[r0ckstarbob]

nope, not at all.

i mixed my methanol and water by hand and is what i'm using - it wasn't until later that someone in one of the other forums that i posted this on found out that windshield wiper fluid also contained a 33:66 rato methanol/water mix though he had to do a little looking at the contents.

[Jessfm]

RockstarBob, can i hav permission to copy your charts/graphs to Coolhardware ( host on our server). ?

[r0ckstarbob]

sure. are they not loading up quickly?

[Jessfm]

Nope m8, they are fine.
But I get asked about this stuff all the time from some Uk sites.
Our community site Coolhardware.co.uk will host them as a hard refference.
Forums are O.k, but I h8 trying to search through for old topics that might help others.

[r0ckstarbob]

just looked at the site - looks nice. yeah i've no problem with that at all. you can cut and paste the whole thing word for word if you like, or whatever. if you want a summarized bit you might find this helpful.

http://languagehammer.net/core/coolant.html

same info, though slightly better organized.

[MeltMan]

so what you are saying is use windshield wiper fluid to cool our computers?

[r0ckstarbob]

basically, yup.

i'm saying that antifreeze needs to stay away from our systems and that methanol works literally thousands of times better, and that yes, after after all my reasearch and unbeknownst to me at the time, windshield wiper fluid has been using roughly this same formula forever and a day. knowing that, just keep in mind to read the contents for additives that might adversely affect your computer. the best fluid out there for our purposes will only have distilled water and methanol. some of them add small soap agents, some of them make fluid out of watered down antifreeze (Ethylene Glycol) or Propylene Glycol. those are the ones to avoid.

[r0ckstarbob]

and yeah, i want my own avatar so this is post 11



tee hee

[EMC2]

Bob

Nope, not at all

Ok, here goes

Ok, first I'll agree with you regarding methanol and water being much better for the liquid side of the cooling equation of computers from a thermal standpoint.

Hmmmm... where to start regarding the rest... too much for one post

Note: all comments are geared towards the cooling systems at hand - low viscosity, low temperature, newtonian fluids. (the term 'low' being relative - temps under 80C and viscosities such that the liquids are considered "thin" compared to "thick" fluids like oil)

* and the winner is *


Viscocity


The kinematic viscosity is the important value Dynamic viscosity alone is of lesser importance.

First a little piece of info you may not know. Kinematic viscocity and Dynamic viscocity can be related to one another through a material's density. Here's the relationship:

Kinematic viscosity v (Greek letter nu)
Dynamic viscocity u (Greek letter mu)
Density p (Greek letter rho)

v = u/p


In the case of thermal designs using liquid, there are 3 important and inter-related issues:

The Reynold's number (an indicator of flow type - from fully laminar to extremely turbulent) for the fluid flow - this affects directly both the frictional losses and the film coefficient.

Frictional losses through the system - this affects the final mass flow rate (which subsequently also affects the velocity of the flow).

The Film Coefficient (also called the convection heat-transfer coefficient) - in this realm, a measure of the heat transfer between the fluid coolant and the solids it passes through (both block and rad in this case).


Here are the fluid characteristics which are used to calculate each of these 3 very important values:

Reynold's Number - this is a dimensionless number (no units of measure associated with it) that is a relationship between the inertial forces and the viscous forces in fluid flow. The characteric of the fluid used to calculate this number is the kinematic viscosity.

Frictional losses - the number of friction factor correlations in unreal, but - they all use the same flow properties, the differences are in how they handle the geometry of the channel the fluid flows within.

For laminar flows, it's pretty straightforward - it depends only on the Reynold's number. For turbulent flows, it depends upon both the Reynold's number and on the surface features and geometry of the channel the fluid flows through. No direct fluid properties are used - these are taken into account through the use of the Reynold's number.

These frictional losses are usually discussed in terms of head and are what determine what the flow rate through the system is for a given pump. The flow rate of course also determines the velocity of the flow at any point in the system.


Film coefficient - this is the 'ultimate' value that determines the heat flux from the solid (water block in this case) to the fluid flow and from the fluid flow to the solid (heat exchanger, radiator, heater core). The basic equation for the heat transfer is as follows:

Q/A = h * (Tw - Tf) where Q/A is heat flux per area, h is the film coefficient, Tw is the Wall temperature, and Tf is the Fluid temperature.

The film coefficient is calculated from the following characteristics of the fluid flow - the thermal conductivity of the fluid, the Reynold's number for the flow, and the Prandtl number (another dimensionless value that characterizes convection - it relates the hydrodynamic and thermal boundary layers, linking velocity and temperature and is equal to the kinematic viscocity divided by the thermal diffusivity).


Hopefully, you can now see why kinematic viscosity is the "important" viscosity as it were .

ps. Be careful around here about asking for Avatars and labels - just ask ECU Pirate, lol

[EMC2]

Originally posted by ECUPirate
You've established that meth is better than glycol for our applications.

Oh sheeet ECU, don't go giving your computer speed, alcohol, or prestone

[Jessfm]

Bob,
excelent stuff !
Ill get it up , cheers ( usual credits to you etc etc)

[ECUPirate]

lol... yeah.. just look under the forum "random nonsense" for my thread, "I got me a new avatar, yo", and "I don't wanna stink no mo.".. that should explain everything.

[r0ckstarbob]

Originally posted by EMC2
Bob

Ok, here goes

Ok, first I'll agree with you regarding methanol and water being much better for the liquid side of the cooling equation of computers from a thermal standpoint.

Hmmmm... where to start regarding the rest... too much for one post

Note: all comments are geared towards the cooling systems at hand - low viscosity, low temperature, newtonian fluids. (the term 'low' being relative - temps under 80C and viscosities such that the liquids are considered "thin" compared to "thick" fluids like oil)

* and the winner is *


Viscocity


The kinematic viscosity is the important value Dynamic viscosity alone is of lesser importance.

First a little piece of info you may not know. Kinematic viscocity and Dynamic viscocity can be related to one another through a material's density. Here's the relationship:

Kinematic viscosity v (Greek letter nu)
Dynamic viscocity u (Greek letter mu)
Density p (Greek letter rho)

v = u/p


In the case of thermal designs using liquid, there are 3 important and inter-related issues:

The Reynold's number (an indicator of flow type - from fully laminar to extremely turbulent) for the fluid flow - this affects directly both the frictional losses and the film coefficient.

Frictional losses through the system - this affects the final mass flow rate (which subsequently also affects the velocity of the flow).

The Film Coefficient (also called the convection heat-transfer coefficient) - in this realm, a measure of the heat transfer between the fluid coolant and the solids it passes through (both block and rad in this case).


Here are the fluid characteristics which are used to calculate each of these 3 very important values:

Reynold's Number - this is a dimensionless number (no units of measure associated with it) that is a relationship between the inertial forces and the viscous forces in fluid flow. The characteric of the fluid used to calculate this number is the kinematic viscosity.

Frictional losses - the number of friction factor correlations in unreal, but - they all use the same flow properties, the differences are in how they handle the geometry of the channel the fluid flows within.

For laminar flows, it's pretty straightforward - it depends only on the Reynold's number. For turbulent flows, it depends upon both the Reynold's number and on the surface features and geometry of the channel the fluid flows through. No direct fluid properties are used - these are taken into account through the use of the Reynold's number.

These frictional losses are usually discussed in terms of head and are what determine what the flow rate through the system is for a given pump. The flow rate of course also determines the velocity of the flow at any point in the system.


Film coefficient - this is the 'ultimate' value that determines the heat flux from the solid (water block in this case) to the fluid flow and from the fluid flow to the solid (heat exchanger, radiator, heater core). The basic equation for the heat transfer is as follows:

Q/A = h * (Tw - Tf) where Q/A is heat flux per area, h is the film coefficient, Tw is the Wall temperature, and Tf is the Fluid temperature.

The film coefficient is calculated from the following characteristics of the fluid flow - the thermal conductivity of the fluid, the Reynold's number for the flow, and the Prandtl number (another dimensionless value that characterizes convection - it relates the hydrodynamic and thermal boundary layers, linking velocity and temperature and is equal to the kinematic viscocity divided by the thermal diffusivity).


Hopefully, you can now see why kinematic viscosity is the "important" viscosity as it were .

ps. Be careful around here about asking for Avatars and labels - just ask ECU Pirate, lol

i had some help with this, several chemists, a PhD in chemical engineering, and a couple of mechanical engineers. all of whom pointed at the dynamic viscosity as being the measurement most relavant to what we're doing in computer cooling. this might be a technical issue, as the kinematic viscosity of methanol also vastly oustrips antifreeze (or ethylene glycol if you like) but to break the differences down between them into laymans terms as i understand them, dynamic viscosity measures the amount of resistance or drag a liquid moving through a tube or capillary deals with (or a hose such as the ones we're using in cooling computers) and kinematic viscosity measures the amount of drag or resistance a solid object encounters when moving through a liquid (such as dropping a ball in a bucket of liquid). airplanes, cars, buildings and submarines measure the kinematic viscosity because the object is moving through a liquid (air or water in this case). measuring the amount of resistance or drag of oil in a pipe however (for example) will use the figures found in the dynamic viscosity ratings.

i am purposely using laymans terms so everyone can understand this and the differences between them. am very familiar with reynolds numbers and laminar flow coefficients.

in the end, both apply in some regard i suppose, and i have no doubt your knowlege of this is quite comprehensive as you've demonstrated. but i'm going to stand by my figures and the means in which it's measured. i feel it's the best way to represent the differences and directly provide us with the most accurate figures for a comparison that is relavant to what we're doing. thanks though for the time and energy you spent putting into this though. very informative.

[r0ckstarbob]

Originally posted by Jessfm
Bob,
excelent stuff !
Ill get it up , cheers ( usual credits to you etc etc)

rock and roll mr man. drop a link when you get it put up. i'd like to see it. just for curiosoties sake.

[Jessfm]

Its up
http://www.coolhardware.co.uk/module...wcontent&id=20

Did not know what name to use for credit.
If you can email me credits information Jess@coolhardware.co.uk , ill get it added right away.

Cheers m8 , hopfuly this will lead ppl onto your other interesting work.

[Brad]

I've seen all this before at those 3 other forums, very cool work, the one thing I want to ask is even though methanol and water have a lower thermal conductivity and heat capacity, because when you add methanol the viscosity drops so much, would a methanol and water mix be better than straight water at ambient temps?