For all you watercooling nuts out there - hard data

[r0ckstarbob]

you're right about aluminum being a little higher. my bad.

as long as you're consistantly putting energy (heat) into something, you are engaging in a steady state heat transfer model. idle or load have no relavance except to determine how much energy is being delt with. in both cases you're still perpetually putting energy into a substance. thats why its called steady-state heat transfer. there is always heat going into it whether in idle or at load.

ya know, i've just spent the last 9 months in school studying thermodynamics where i've had an opportunity to tweak and verify my results, have recently had my data and conclusions reviewed by professors here at the University of Washington (and have been accepted into the mechanical engineering program based in large part upon the weight of my work in this arena), am continuing my degree in Industrial Design at the Seattle Institute of Art, and my findings are available to the public for scrutiny. and in each instance, these findings have passed muster. it's not perfect and i never claimed it would be, but it's damn close - and a damn sight closer than anything you've ever done.



this thread has seen over 16,000 hits, it's data archived, and has been published multiple times on multiple sites.




calm down




you're ranting like a lunatic and i'm tired of having to explain the basic fundamentals of thermodynamics and heat transfer to you while you continue to try to forward arguements that, while may not entirely be without merit in some cases, are self defeating in your zealousness to be entirely right when you are not.




bye bye sap ol buddy. come back when you get a grip.

 

[safemode]

since we have lunatics debating i figured i'd have to say something.


A heatsink's function is to move heat from point a to point b and to make point B as big as possible. Because air sucks at cooling. Air has a sucky specic heat and has a sucky heat conductivity.


Air can only take in a certain amount of energy at any given time (constant as per temp) and a high specifc heat material made heatsink is going to be at the exact same temp as the low specific heat heatsink (assuming it's not lower than the substance it's transferring heat into).

This is because that middle-man substance (heatsink) cannot transfer it's heat that it absorbs, be it small or large specific heat, until it's temp differential reaches a temperature that's high enough to efficiently transfer it's heat to our ending substance (air even for watercooling guys unless they are plugged into tap and have constant new flow).

Everything in between the source and the end substance only have heat conductivity to care about. High heat conductivity means the temp differential doesn't have to be that high to begin transferring heat efficiently.

So the answer is, no, heatsink material's specific heat doesn't matter for a damn. Air's does because it's what we're dumping heat into and we get rid of that and completely new air comes along. We dont care what the air transfers heat into, it's not our problem.

Everything including air has heat conductivity to care about. We want everything to have a high heat conductivity to make sure we have as low of a temp differential needed to transfer heat as possible.

I dont really see what the big debate is. Your heatsink isn't going to work if it doesn't get hot enough to transfer heat to whatever it's supposed to transfer heat into and if that's not air then that's not gonna work until it get's hot enough to transfer heat. So it doesn't matter if you have a new compound that has the same conductivity as copper but can absorb 10W/g or only what copper can absorb before it increases 1 C, it's still going to end up being the same temp because air is still the same and it will only begin absorbing heat when you reach the necessary temp (which would be the same if all environmental conditions were the same). The exact same thing is true for water cooling and any other kind of cooling that eventually relies on air to absorb heat. Air sucks, water cooling is just an efficient way to increase the surface area the heat is spread across (and spread it evenly) so more air comes into contact with it at any given moment at least with inline systems).

It's _all_ about conductivity...the higher the lower the temp differential has to be and thus the cooler the cpu can be.

Is there really an argument over this? or did someone lose their medication?

 

[r0ckstarbob]

basically thats it. this whole arguement has degraded into some really obscure semantics that 99% of watercoolers/overclockers couldn't give a damn about... myself included. what it boils down to is you want high conductivity in your liquid because adjusting pump speed is far easier to do than trying to mess with radiator efficiency.

thanks for the lunacy check safemode.

 

[myv65]

I can only shake my head when folks get this worked up over what really amounts to a trivial matter.

To safemode,

Thanks for injecting a little sanity back into this thread. It was sorely needed.

To sappo,

You mentioned boxing earlier. I'd kindly ask both you and Scott to step to your respective corners for as long as it takes to cool off. You've both said some stuff in error even though both have been correct almost exclusively. Biggest problem I've got with your statements is you flit back and forth between discussing heat sinks and fluids while talking about the relative benefits of conductivity, specific heat, etc. They are different animals and have different factors that dominate their performance.

To Scott (r0ckstarbob),

Darn near everything you wrote is technically correct, but you put too much importance on "thermal differential".

To all,

No pun intended, but what it really boils down to is the basic equation defining convection, namely that q = h * A * delta-T. "q" is heat transfer, "A" is surface area, and "delta-T" is mean temperature difference between the liquid and surface. "h" is the ill-understood convection coefficient. Both "h" and "delta-T" must be applied over each teeny-tiny area as they vary according to the localized temperatures. Call 'em continuous functions if you're a mathematical sort.

"h" can not be analytically determined. In a nutshell, it's a function of how efficiently heat is being transferred between the fluid and the surface. More specifically, it's a function of localized velocity, surface shape, velocity profile (dependent on many factors including geometry and viscosity), and temperature-dependent fluid properties (conduction coefficient, viscosity, specific heat, etc.). Of all this stuff, we have little control over any of it.

We get to choose a block, pump, (insert rest of your system here), fluid, CPU, and to a degree the CPU power. For a given setup, we kind of live with what we have. Of all these things that go into determining convection, viscosity rules. Here is why.

Viscosity is a large determinant of boundary layer thickness. For a given fluid velocity and surface geometry the boundary layer thickness correlates to the viscosity. Higher viscosity means a higher boundary layer. Thick boundary layers mean that very little mixing occurs between the fluid "just passing through" and the fluid that's basically stuck to the solid's surface. Thick boundary layers place a premium on a fluid's conductivity and have no place in water cooling.

Low viscosity means low boundary layer thickness. Low boundary layer thickness means less separation between the fluid "just passing through" and the stuff basically stuck to the solid's surface. This reduces the dependence on the fluid's conduction coefficient. This is the first reason why low viscosity is critical.

The second reason is flow rate. Once you choose your system components, the flow rate you'll get depends on the fluid viscosity. Forget about varying pump speed. Most folks run AC pumps that run at a speed synchronous to the line voltage. Yup, there are ways around it, but most won't bother. Second, provided the power input by the pump doesn't become dominant (more than ~1/2 of the CPU power), more flow will always improve the CPU temperature.

High viscosity results in decreased flow rate. Decreased flow means decreased velocity, hence decreased convection coefficient.

This is the double-whammy of high viscosity.

You might as well toss both conduction and specific heat out the window in relation to fluids. With the right viscosity, you can always generate enough flow (with our typical fluids) to overcome low conductivity and/or low specific heat.

In heat sinks, quite the opposite is the case. Obviously viscosity has no bearing on a heat sink. Both conductivity and specific heat do, however.

A brief explanation of "steady-state": Steady-state means "does not vary with time". CPUs do not follow this definition, but it still applies. CPU power will vary with time, but in the worst-case it runs at 100% load all the time. If the cooling system can't handle this, it is not a viable cooling option.

In liquid cooling, some people think it can't be steady-state because the fluid is continually heating up and cooling. Nah, it's a matter of perception. An engineer analyzing the situation would define a "control surface" around the interested area. Say it's the block. The control surface would surround the block, cutting through the inlet and exit. Water crossing the inlet would always be the same temperature in steady-state. Water exiting the block would always be the same (and warmer than inlet) temperature in steady-state. The CPU would put a constant power level into the block.

OK, back to conventional heat sinks. Specific heat is grossly misunderstood by the computing public. All specific heat does is determine how quickly something changes temperature when you leave a steady-state condition. You have high specific heat? Temperatures won't change quickly. Vice versa is also true. You want an analogy?

Think of specific heat as car velocity and heat sink mass as car size. A bigger car can't negotiate a slalom course as quickly as a smaller car. You have more heat sink mass, it won't change temperature as quickly as a low-mass heat sink. You're going really fast? You won't be able to turn as quickly. You have high specific heat? You won't be able to change temperatures as quickly.

And much like cars, how quickly you weave through a slalom doesn't mean squat about how fast you can go in a straight line. This comes down to power and aerodynamics. In heat sinks, this comes down to conductivity.

Well, I've said my peace and hope it helps someone along the way. My thanks if you've managed to read this far without falling asleep.

 

[r0ckstarbob]

right-O... i'm off.



** hops off to happy hour at the Lava Lounge **

 

[sappo]

first of all to rockstarbob, i think i came on a little strong, and i said some things that i should not have. i'm sorry.. Also, thanks for listening. I didnt think i was getting through.

now to myv65, you have me sold on the whole convection thing.. That makes perfect sense me.

Also the boundry layer is something i have never thought of.... BUT (and far be it from me to start a huge debate).. isnt that the same thing as the surface tension? In which case the boundry layer would not be directly related to the viscosity. Because soap takes away the surface tension even though soap adds viscosity to the water.

Another problem i have with your post on viscosity (and correct me if i'm wrong here) is that virtually every liquid pump you can get (for a reasonable price) pumps water. Therefore, it is designed to pump liquid that is about as viscus as water. If the liquid is too viscus, it would obviously be hard on the pump, BUT I also dont think it would be good for the pump if the liquid had an extremely low viscosity either.

Example: A car's RPM's. You dont want the engine to be putting out too many RPM's (like a low-viscosity liquid) because it will be inefficient. You also dont want a car's RPM's to be too slow (as it would be with a highly viscus liquid).

 

[safemode]

the only reason you dont want higher rpms with the same power going into getting those rpms as the engine was designed to is because of thermal breakdown. High rpms cause more friction than the engine is designed to handle and more friction causes extreme heat and lubrication breaks down and you get a chunk of useless metal in your car. This is true of pumps and fans and everything else with moving parts. This is why you don't want extremely low viscous liquids..

 

[safemode]

Also i dont see the logic in this constant heat term. It doesn't make sense with itself. You're not putting a constant magnitude of heat into it and it causes much more confusion that it's useful.

Make up some other term to describe heat transfer material and a term to describe heat dumping material.

Heat dumping material would be air in almost all cases i can think of. This is a material that for all purposes is infinite in quantity and never re-used.

Heat transfer material is a reused and usually fixed amount of material that is used to move heat from the heat source to the heat dumping material and thus out of your environment.

Have fun with that.

 

[myv65]

Viscosity and surface tension are not the same thing. In fact, as Scott said in his write-up there are actually two measures of viscosity. They differ only in that one is divided by the density of the fluid.

Boundary layer thickness is pretty much a function of geometry, velocity, and viscosity. I'll quote a brief passage from my fluids book "Introduction to Fluid Mechanics" by Robert Fox and Alan McDonald.

In the boundary layer both viscous and inertia forces are important. Consequently, it is not surprising that the Reynolds number (which represents the ratio of inertia to viscous forces) is significant in characterizing boundary-layer flows."

Surface tension essentially defines how well a fluid sticks to itself. It's viscosity that determines how easy it is to pump.

What's really going on in the boundary layer is shear. The fluid at the surface does not move. The fluid "just passing through" moves with relatively high velocity. The transition between zero and the average velocity is the boundary layer. As a cheesy test, you can drag a stick slowly through some 50 weight oil that has dust, sparkles, whatever floating on top. You'll see movement in the dust well away from the stick. Do the same thing with water and you won't see as much motion. Just remember to go slow or you'll screw up the test.

As with engines, there are pumps that rely on the viscosity of the fluid getting pumped for things like seal lubrication. In the extreme cooling conditions that Scott is after, viscosity will always be higher than water at room temperature. Viscosity that is too low is not generally a concern for us.

 

[sappo]

Thats sounding alot like cohesion. Isnt surface tension a small film on the liquid and cohesion what makes the liquid hold itself togather??

 

[myv65]

Thats sounding alot like cohesion. Isnt surface tension a small film on the liquid and cohesion what makes the liquid hold itself togather??

That's what I get for writing late at night.

I should have said that surface tension is how well a fluid sticks to other stuff (adhesion), rather than itself (cohesion).

Anyway, fluids with widely varying viscosity have remarkably similar surface tension. In SI units, it's measured in mN/m, and most liquids fall in the range of 20-40 at 20°C. Water, though with lower viscosity than most, has higher surface tension. It comes in around 73 mN/m at 20°C. This data is all from the same text I referenced previously and pertains to contact of the liquid against air.

For anyone wanting more information on the topic, the text I referenced is a good introductory book on fluids. I'm not sure what revision they're up to now as I had the class, uh, a few years ago. One really nice thing is they provide reference lists to a lot of other texts as appropriate, particularly to offer a greater discussion of a specific topic.

 

[riprock]

Just an FYI. I watched this thread back late last year to determine what best to use in my new water cooling system:

XP1800+ @ 2000+ (1.6Ghz)
Danger Den Maze 2
Eheim pump
87' Chevette Radiator w/ 130CFM & homemade shroud
Homemade Resevoir
Silicone Tubing

I've been using windshield washer fluid (that the manufacturer in Tulsa told me was 70water/30methanol) for almost one year.

Under full load:
Room air temp 25c
Case air temp 26c
direct die thermal probe temp 33c

My system runs 24/7. When I'm not actually using it, it runs distributed.net. There's no scum in the lines or resevoir. No leaks and the pump is still quiet and working fine.

 

[Molybdym]

My god, this is the best data out there. All this information. . .all the numbers. . .AHHH[insert pop sound]. You, you wonderful overclockers you. We owe you a thanks.

 

[r0ckstarbob]

hey guys, just a heads up

theres been a rumor around that people are having a hard time pumping methanol actualy BECAUSE of it's lower viscosity. someone mentioned that it was a pump issue - in that a possible work around solution would be to NOT use a centrifigual pump like 90% of us use... something i hadn't even considered.

any truth to that do you think? are there others having a hard time pumping methanol due to the fundemental nature of a centrifugal pump being what it is? i've not had any problems but on the other hand i'm not using a centrifugal pump and haven't personally run into the problem...

thoughts on this would certainly be welcome.

 

[grover]

rockstarbob: I agree completely at your conclusions and that plain old water *should* be better than antifreeze, but in my own experience I actually found the reverse to be true. When I replaced my water system with 50/50 petsafe antifreeze (propylene glycol), pretty much the WORST thing I could do by your numbers, my CPU temperatures instantly decreased 1°C, indication that propylene glycol is cooling my system better than water and not 5-6° worse as ideal calculations predict. Any idea why this is true? At the very least it indicates there are variables that haven't been considered here.

Do you know of any experimental data showing the trends for a number of systems with different coolents? I'm particularly interested in water wetter- everything I've heard about it automotive-wise seems to indicate it's a waste of money.

 

[r0ckstarbob]

i dunno, theres lots of reasons. in the scheme of things, 1C doesn't really seem that much of an improvement. not to discount anything, but you can get plus/minus a couple of degrees just by adjusting flow rates and 'tweaking' your own specific system into set if that makes any sense. if i had to hypothosize (or guess) i guess i'd say that perhaps the antifreeze brought the flow rate down a bit more which in your case might have actually been better for your system? it's been proven that each watercooling system has a "sweet spot", know what i mean? maybe you got closer to the sweetspot. i dunno. so i guess one option is to shrug it off and say "fuggit, it works so why mess wid it?" which i'd prolly end up doing . the other option i guess i'd suggest straight off the bat would perhaps be to try to go back to 100% water and play with different flow rates for a bit and see if you can't get better results and find your sweet spot. just a guess. other than that i have no idea.

RSB

welcome to the forums

 

[grover]

Since then, I replaced the pump (couldnt' fit a GF4 in my case with my old resevoir, heh. Figured I might as well upgrade the pump while I'm at it!) and temps have since dropped another 2°C. So, couldn't be hitting the sweet spot from reduced flow causing the improved performance.

 

[r0ckstarbob]

well, i dunno then. there are innumerable factors to consider to these things - whole systems thinking is critical when trying to troubleshoot watercooling setups if only because there are so many variables. guess if it were me and i got better results with antifreeze and i knew for a fact that antifreeze was worse for my system than pure water, i suppose i'd take that as a red flag pointing to evidence that there was more performance to be squeezed out of my box in some manner, something that i was overlooking somehow or something that could use tweaking.

again, without knowing more about your system it's hard to say off hand. also again though, if you're happy with your performance, then i say chuck it. what is theory in comparison with real world results?

peaz

 

[r0ckstarbob]

Just an FYI. I watched this thread back late last year to determine what best to use in my new water cooling system:

XP1800+ @ 2000+ (1.6Ghz)
Danger Den Maze 2
Eheim pump
87' Chevette Radiator w/ 130CFM & homemade shroud
Homemade Resevoir
Silicone Tubing

I've been using windshield washer fluid (that the manufacturer in Tulsa told me was 70water/30methanol) for almost one year.

Under full load:
Room air temp 25c
Case air temp 26c
direct die thermal probe temp 33c

My system runs 24/7. When I'm not actually using it, it runs distributed.net. There's no scum in the lines or resevoir. No leaks and the pump is still quiet and working fine.

niiiice temps btw

 

[Rottys-R-Us]

what it boils down to is .

Nice work.

I like the what it "BOILS" down to, nice punn