Optimal rad setup: pressure

[bigben2k]

Originally posted by gmat
... just like some ppl hate coding in assembly...

What? Who hates coding in Assembler?

I'm reviewing everyone's answers, and I'll post something later. Today is not a good day

[NoSoupForYou]

No, no, no, no, no!

Increasing flow does not increase heat dissipation! It can't. At equilibrium, the rad always dissipates the same amount of heat - the amount added by the CPU.

What increased flow does is lower the delta-T between the coolant and the rad required for the heat flow. This decreases the overall coolant temp. It does not increase dissipation.

At equilibrium, the coolant temperature will drop by exactly the same amount in the rad as it increases in the block. Else we wouldn't be at equilibrium.

The best way to get the coolant temperature as close to ambient as possible is to increase dissipation to the air - either by increasing the fin area or moving more air. Sadly (since it adds to noise) this is the area most likely to show improvement in a WC setup, and the one most often ignored.

[Sirpent]

Originally posted by NoSoupForYou
Increasing the water flow rate through the rad does one thing, and one thing only. It increases the heat transfer coefficient between the coolant and the walls.

I think that when the flow is laminary, water flows much slower near the walls of the radiator than in the center of the channel. The effect on cooling is similar to adding a layer of thermal insulator (water does not conduct heat well without convection), which is bad. Increasing the flow rate is good because it can break the laminary flow pattern.

Is this what you meant, NoSoup? Or you had in mind something different?

Originally posted by NoSoupForYou
gmat - LMTD is a little over the top here - the temp change in the fluid is small enough that approximating it as constant won't introduce significant error.

Constant = arithmetic mean? Should be OK if the flow is not too low.

[Volenti]

The only problem I see in trying to draw links between CPU water cooling and car engine water cooling, is the desired result of each, both use overkill style setups, but a car's watercooling is designed to keep the engine within a specific temprature range controlled by a thermostat, where as a cpu's water cooling is designed to keep the temp as close to ambient as possible.

I propose an easy solution to this whole debate, throw away your rads and use evaporative cooling instead

[chazz469]

Originally posted by Cova
...Our rads run really inefficiently because we have a low delta temp between the water and ambient air. To get the rad to dissipate more heat (be more efficient) we need to increase that delta, but by doing so we also increase the temp of water passing through the CPU block, which is a bad thing...

Cova...what if one were to throw a couple pelts onto the rad, thus lowering the "relative ambient" of the radiator and increasing the deltaT. You obviously don't want to add heat to the coolant, as the point is to get it out of the coolant. So if you change the "relative ambient temperature, you increase the deltaT, the RAD works more efficiently, and I would think, cool the coolant below ambient since the "relative ambient" would be significantly lower that the "actual ambient".

[Dix Dogfight]

gone_fishin:

You state that higher flow makes a lesser temp diff in the wb but that's not why higher flow is better.

Exactly

Here is an example (all numbers are made up, but will hopefully demonstrate what I mean)

Lets say you have a setup with no rad, using tapwater and never use the same watermolecule again. At first we have a high flow 0.3 of lps and a CPU temp of 30C. Then we lower the flow to 0.03 and the temp rises to 40C.
The equation shows that the rise in temperature in outletwater will for a normal CPU=80W be:
80/(4180*0.3)=0.06C
80/(4180*0.03)=0.6C
So the outlet water will be 0.54 degrees hotter with the lower flow and the CPU will be 10C hotter.
The mere increase in watertemp alone can't produce the higher CPU temp.
So the lower CPU temp can't be a result of the lower temp diff between inlet and outlet as many people tend to think but rather a effect of the WB having better thermal transferproperties at a higer flow.


gmat:

1 - if you're here to flame ppl and dont understand humour go to [H] you dont belong here.

I never flamed anyone i simply asked you to stop assuming others not to understand all great things you have learned in school.
I learned a whole lot during my 8 years at the Royal Institute of Technology (Swedens version of MIT) so I know exactly what school's like and all the great thing you can learn there.
But telling everybody that I have the knowledge and don't share it isn't very productive.

Humor in my world isn't statments like:
I know a whole lot of stuff about thermodynamics but i'm not going to post them here because the are to hard to understand anyway. Just belive my word. Higer flow is better.
Statements like that are totally useless.


NoSoupForYou

I think if you are going to have an attitude, learn to clarify a bit better.

On the subject of attitude:
Making a post clear and easy to read/understad produces a very long post and many ppl tend to ignore those. So in keeping a post short many ppl will read it but at the price of everybody not understanding all. So making the first post short but provocative (using words like BS and crap) usually produces a high number of replys and then try and elaborate after that usually works.
Some ppl on the other hand (not referring to anyone) simply sees the words BS and crap and therfore feel offended and don't take their time to read and understand what the poster want's to say.
Which is a problem with this type of posting approach.

Everyone=all ppl posting here:
One thing to remember is that not everyone here has english as their default language. So please try and read post's twice because we that don't know exactly what word to use, to make the message come forward the way we intended, might seem to have an attitude but perhaps only lacks the finer tune of the english language.

Well I'll leave you guys alone for the moment (4 weeks of vacation starting in 5h. So I'll just check back in in august to see if there has been any progress.

cheers

[gmat]

Originally posted by Dix Dogfight

So the lower CPU temp can't be a result of the lower temp diff between inlet and outlet as many people tend to think but rather a effect of the WB having better thermal transferproperties at a higer flow.

The total heat transfer through the block is the continous sum of heat transfers in finite elements constituting the block.
No matter how you put it, there's a close relationship between inlet / outlet temp diff and total heat coefficient.
But what we said earlier (and repeated it) is we dont care about inlet / outlet temp diff ! From our point of view (watercooling systems) it's only a side-effect.
And wasnt the thread topic about *pressure* ? (on that i'd want to hear fluid dynamics experts)

I never flamed anyone [... another flame follows ...]

Calling ppl "morons" and trolling that way aint very nice, dude.

Humor in my world isn't statments like:
I know a whole lot of stuff about thermodynamics but i'm not going to post them here because the are to hard to understand anyway. Just belive my word. Higer flow is better.
Statements like that are totally useless.

So you never flame anyone ? ah.
So far:
* i've posted all necessary thermodynamics equations (so far only one is of interest i admit)
* i've posted links where they are explained thouroughly
* did you read the thread ? I doubt it. Read before flaming... (where did i say "I know a whole lot of stuff about thermodynamics but i'm not going to post them here" ???)

Making a post clear and easy to read/understad produces a very long post and many ppl tend to ignore those. So in keeping a post short many ppl will read it but at the price of everybody not understanding all. So making the first post short but provocative (using words like BS and crap) usually produces a high number of replys and then try and elaborate after that usually works.

Doesnt that sound like a contradiction with your previous statement ? hmmm.
Anyways on pro/forums ppl DO READ long posts. Until now the level of these forums have been quite high.

I first stated that "thermal laws" simply say that higher flow is better. I never sh*tted on anyone or called anyone a "moron", neither i made a statement that i had "superior" knowledge or such BS.
Then ppl answered (politely and friendly) and i made more thourough statements (friendly, and politely again).
Then i posted some links to these formulaes to support my argument. (did you notice that ?)

I just try to be helpful and bring my small knowledge to these forums since i like the ppl here. And while i'm at work i cant take the time to dig into thermal equations, so i try to stay brief.

Note: i enjoy flamewars as well, so come on, i've got an integral flameproof suit.

One thing to remember is that not everyone here has english as their default language. So please try and read post's twice because we that don't know exactly what word to use, to make the message come forward the way we intended, might seem to have an attitude but perhaps only lacks the finer tune of the english language.

That is a wise statement, and a base of netiquette. For your information calling ppl "morons" wont help you getting friendly answers...

[Dix Dogfight]

NOTHING THAT FOLLOWS IS ANY KIND OF FLAMIN ON ANYONE OR ANYTHING. AND I HOPE THAT IT CLARIFIES SOME ISSUES.
AND WITH THIS I END MY POSTIN IN THIS THREAD (because of earlier stated resons) AND HOPE IT WILL BEER SOME KIND OF FRUIT TO BENEFIT US ALL.

gmat:
Here is what I wrote on the moron subject

And please stop asuming all others in theese foums to be total morons. There are actually alot of intelligent people here

.
I never stated you to be a moron If you read it again I hope you will see that. If you still take it personally I can only apologise.

we dont care about inlet / outlet temp diff ! From our point of view (watercooling systems) it's only a side-effect.

So we are in a agreement then.
My first post was an attempt to explain (for anyone interested) why lesser flow doesn't produce a much higher tempdiff and that it is not the major contibutor to the lesser performance that a lower flow does.

Until now the level of these forums have been quite high.

Now do i take this personally or do you mean that they still are. Or you might mean a whole other thing.
I can choose read it as a flame or not. Do you se what i mean when I say that small differenses in language = how I choose to write. What words I use can make a world of difference in how the recipient interprets it.
If the recipient onyle reds through it fast and don't take into account that we are from different parts of the world then these kind of misshaps will continue to happen.

I wrote:

I never flamed anyone i simply asked you to stop assuming others not to understand all great things you have learned in school..........

gmat wrote:

I never flamed anyone [... another flame follows ...]

I can only assume that my message didn't get through properly again and was interpreted as a flame.

[NoSoupForYou]

Wow, I can't believe we are still debating this.

It is clear - inlet to outlet temp differential is a function of the heat input and flow rate only. That's it.

No matter how you put it, there's a close relationship between inlet / outlet temp diff and total heat coefficient.

Nope. Well, not really anyway . There is correlation, but no causality.

If you have a higher flow rate, you will have a lower inlet/outlet delta-T. That's true. And if you have a higher flow rate, you will have a higher heat transfer coefficient. Also true. But to say that a lower inlet/outlet delta-T causes a higher heat transfer coefficient is technically incorrect. It is the higher flow rate that causes both the lower delta-T and the higher heat transfer coefficient.

And Dix - yes, maybe a couple of opening flames generates more interest in the thread. But it's bad interest. Look at the last few posts - little information, mostly flames and explanations of why a flame wasn't really a flame.

It's a sad state of affairs when a post more than 10 lines long gets ignored. But at least those that might benefit may read it.
Please, someone read my earlier post, and if you have a question ask it. But it seems like the concept has fallen on deaf ears.

[Quickmcj]

No....I have read it all

And it is very interresting

[Quickmcj]

Gmat has showen us some links and math that I dont get, but it is very good stuf for ppl that understands it.

I though......who does know the answer to all this questions etc. Well it surely must be the carmanufactors....because if they create an very efficiant heatcore, they will only have to make it "small" and thereby they will have less expensives. And that is there interest. So these guys must "know it all" when it comes to heat transfering. Well then it sould also be the best heatcore's around because....the know what they are doing. Therefor......buy a auto heatcore made of cobber and with the right measures to fit your case and you would be pleased. Or is every thing I said wrong ?

[gmat]

NoSoup: consider my silence as a total agreement.

Quickmcj: You got it right. Actually the very 'heater cores' you can buy from D-Tek and various watercooling shops are auto parts or slightly modded auto parts.
Look at threads started by hmale, there's one where he shows his new heater core. It's designed to cool down race car engines up to 600hp. It will be used to cool down a PC, he he he.

Besides, bb2k started this thread to talk about *pressure* issues in rads. Do we want low or high pressure ?
From my (modest) knowledge:
- we want pressure to work for us, not against us.
- there *are* pressure differentials throughout our water lines, and not negligible, even with our tiny pumps.
- the very few serious tests i've seen tend to prove that low pressure in rads (as opposed as high pressure in WB) is a good thing.
- i've yet to talk to a fluid dynamics expert (i know a few, ill see them this WE) about this, because this problem is clearly over my current knowledge (heck i'm a computer scientist, FD courses are a few years back...)
- this problem does *not* involve flow or temperature. We're talking about pressure here.

[Quickmcj]

I have seen the tread from Hmale......didnt bother to write it in here

[Cova]

Originally posted by chazz469


Cova...what if one were to throw a couple pelts onto the rad, thus lowering the "relative ambient" of the radiator and increasing the deltaT. You obviously don't want to add heat to the coolant, as the point is to get it out of the coolant. So if you change the "relative ambient temperature, you increase the deltaT, the RAD works more efficiently, and I would think, cool the coolant below ambient since the "relative ambient" would be significantly lower that the "actual ambient".

As I mentioned before..., this really is unrelated to this thread (which is about flow but should be about pressure). But I'll reply quick to this here.

1. You will have a hard time mounting pelts onto a rad without restricting air-flow through the rad (assuming you want the pelts on one of the water channels)

2. You still need to cool the hot side of the pelts.

If I'm bored later today I'll start a new thread and explain my entire theory and the possible solutions I've come up with so far.

[Sirpent]

Originally posted by gmat

Besides, bb2k started this thread to talk about *pressure* issues in rads. Do we want low or high pressure ?
From my (modest) knowledge:
- we want pressure to work for us, not against us.
- there *are* pressure differentials throughout our water lines, and not negligible, even with our tiny pumps.
- the very few serious tests i've seen tend to prove that low pressure in rads (as opposed as high pressure in WB) is a good thing.
- i've yet to talk to a fluid dynamics expert (i know a few, ill see them this WE) about this, because this problem is clearly over my current knowledge (heck i'm a computer scientist, FD courses are a few years back...)
- this problem does *not* involve flow or temperature. We're talking about pressure here.

Gmat, what are those serious tests? I don't see how pressure (and *not* flow) can be important. Suppose you attach an extra pump to the bleed line, and this pump is trying to pump air in or out changing the pressure inside your system. Do you think this can have an effect on cooling? Why?

[gmat]

Actually i'll talk to an expert tomorrow and we'll discuss about that precisely.
For what i understood so far, there's a relation with Bernouilli equations, Euler equations, and flow acceleration through local low pressure zones.

Sirpent I dont think going so far is going to be of help, though i'll try to consider this as well. All i wanted to consider was our common WC problem, aka pump->rad->block or pump->block->rad (neglecting a res).

More on that tomorrow.

(edit) btw serious tests were a rad round-up on overclockers.com (ask BillA for this, if my memories are correct).

[myv65]

First of all, my apologies. Much of what follows is sure to upset some folks that have posted in this thread. I intend no disrespect and have no desire to upset anyone, but I find it hard to read the mix of fact and fiction found here. For the most part, people stating claims have been correct. Unfortunately, there are interspersed errors and misinterpretations mixed in with the rest.

gmat,

You've got a pretty good grip on most of this, but some of your statements make me wonder. You have obviously studied the topic of heat transfer, and most of what you've said is true. Some of it, though, is nonsense. To your credit, you're only referencing someone else's flawed results.

Theres a point in the circuit where pressure switches from positive to negative. (not: thats *always* the case even with our poor centrifugal pumps) you'll want that point to be *before* your rad and *after* your waterblock(s).
For more info on that there's a very good article at overclockers.com

Someone posted about this on AMDMB a couple of weeks ago. My reply: "Absolute garbage".

Heat transfer from a surface to a liquid or gas depends on many factors including surface shape/roughness, delta-T, and myriad fluid properties (most of which are temperature dependent). For incompressible liquids, pressure has no bearing on heat transfer. Sure, if you wanna get technical, water is compressible, but its bulk modulus is ~200,000 psi. The pressures in our systems pale in comparison.

It's also an error to state that pressure always changes from positive to negative. This too depends on a couple of things. First, if your system has an open air reservoir, then the only regions of below-atmospheric pressure reside in the pump casing. Second, if your system is truly sealed, the suction line may be at less than atmospheric pressure. Usually it will be, but if your system was filled with relatively cool water and heated, the expansion of the water will create a static pressure higher than atmospheric. Whether or not you get a vacuum at the suction then depends on elasticity of the tubing and the overall temperature rise of the fluid as well as flow rate and velocity in the lines.

Anyway, getting back to the overclockers thing, I'm specifically referring to a claim that relocating the block and radiator lowered this guy's temperatures on the order of 3°C. This is absolute crap. A system dumping 100 watts into the fluid (75 W CPU and 25 W pump) needs only about 32 gph to keep the peak water differential to 1°C. Claiming a 3°C change only tells me that the guy doesn't know what he's doing.

Sorry if I sound like I am going off on you, but just want folks to fully understand that pressure has no bearing on heat transfer coefficient of an incompressible fluid (aside from determing flow rates).

NoSoupForYou,

You said something about delta-Ts and radiators that didn't sit right. Perhaps I've misinterpreted your meaning.

What increased flow does is lower the delta-T between the coolant and the rad required for the heat flow. This decreases the overall coolant temp. It does not increase dissipation.

Not really. You are forgetting the other side of the equation, er, radiator. The heat transfer from the fluid to the tubing in the radiator is generally not the limiting factor in radiator performance. What is usually limiting is the air flow across the radiator. A higher fluid flow rate will reduce the (already really, really low) temperature differential between fluid and tubing walls. This is normally smaller than the delta-T between air and tubing, especially at the fluid inlet to the radiator.

This next bit also applies to the argument of "parallel" radiators or flow rate through radiators. It should be completely obvious that a radiator will deliver its "best" fluid outlet temperature if the fluid stays in for a very, very long time. Given enough time, the fluid will eventually reach the ambient air temperature, for all practical purposes. On the opposite extreme, very high fluid velocity may leave insufficient time for the heat to escape. This should also be obvious. Imagine a radiator with a fluid flow path one inch long. No way you could get the water to cool off because you'd have insufficient time and area to get the job done.

Everyone isn't using huge radiators simply because:

1) There's only so much room.
2) You can get all the performance you need with a reasonable sized radiator. Beyond a certain point, you have rapidly diminishing returns.

To all,

There is a disconnect for some understanding the relationship of pressure, flow rate, flow velocity, and heat transfer. Pressure along with flow resistance determine flow rate. Flow rate divided by cross sectional area determine flow velocity. Flow velocity, and not pressure or flow rate, has the largest impact on heat transfer coefficient.

What a good block does is transform upstream pressure into high velocity. It does this by reducing flow area. The pressure drop that occurs over the block is much like the spray nozzle we put on our garden hoses. It you run a hose with no nozzle, the water comes out with good flow rate, but not impressive speed. A spray nozzle drops the overall flow rate (like a water block), but results in much higher velocity in the flow that remains. This is the key to getting good heat transfer in the block, where all energy must pass through a very small area.

Velocity is also the key in the radiator, but it's the air velocity that is key. Increasing air velocity allows decreasing radiator size and dwell time.

Again, sorry if any feelings got hurt and sorry for such a long post, but hopefully it helps clarify some things.

[chazz469]

Wow, dude. You just dropped science!!! How is it that you know all of this stuff?

[myv65]

Originally posted by chazz469
Wow, dude. You just dropped science!!! How is it that you know all of this stuff?

LOL, I'm going out on a limb and guessing that yours is not a sarcastic reply. If it is, so be it. If it isn't, I'll answer truthfully. UW-Madison BSME and many years in industry working on hydraulic and heat transfer systems. Biggest heat transfer unit I ever worked on was ~3500 kW consisting of a couple of fuel oil burners heating oil to ~550°F for a paper mill. Biggest chilled water system probably on the order of 30 tons (archaic engineering units, most home CA systems are ~1-1/2 tons). IIRC, 12000 BTUs = 1 ton, so I guess about 360,000 BTU chiller.

Lately I've taken to writing miscellaneous heat transfer guides for AMDMB trying to share the knowledge, translate the engineering geek-speak, and dispel myths. Haven't really got the time I'd like to devote to it, but having fun anyway.

[gone_fishin]

myv65, I've enjoyed your post very much and it is very well thought out. I think perhaps your oversimplified description leaves some doubt.

Quote, "It should be completely obvious that a radiator will deliver its "best" fluid outlet temperature if the fluid stays in for a very, very long time. Given enough time, the fluid will eventually reach the ambient air temperature, for all practical purposes. On the opposite extreme, very high fluid velocity may leave insufficient time for the heat to escape. This should also be obvious. Imagine a radiator with a fluid flow path one inch long. No way you could get the water to cool off because you'd have insufficient time and area to get the job done" end quote

If the heat delivered to the rad is in very small quantities then the time needed to dissipate is also small in the rad. The amount of heat absorbed per volume of water before it reaches the rad is smaller in high flow but repeated many times over compared to low flow in the same time frame. Is my understanding in this flawed?

[gmat]

Originally posted by myv65
if your system is truly sealed, the suction line may be at less than atmospheric pressure. Usually it will be, but if your system was filled with relatively cool water and heated, the expansion of the water will create a static pressure higher than atmospheric. Whether or not you get a vacuum at the suction then depends on elasticity of the tubing and the overall temperature rise of the fluid as well as flow rate and velocity in the lines.

Thanx ! That was exactly the kind of answer i wanted to read.
Actually my system is:
* closed loop, no res, no airtrap
* sealed
and tubes / pipes are rigid.
Given :
* the quite high flow restriction of a waterblock and our small tubes,
* the way our centrifugal pumps work (they are not volumetric, they only work by applying a pressure), the pressure differential in a working pump between inlet and outlet should always be around it's head pressure,
* the quite low temperature differential accors the circuit (1°C or less). Is 1°C enough to produce a noticeable static pressure ?
i've been wondering what was the *real* pressure distribution around a closed loop of water. Maybe CFD would answer that..
You seem to say that in every case the pressure is positive around the whole circuit, which i doubt by experience. (say, i only "doubt", ie i may be quite wrong on this).
What i'm considering here is the dynamic effects.
In a turning flow; the radial kinetic forces tend to create a pressure on the outer side, and an underpressure in the underside (Bernouilli, if my memories serve me right). This effect is mostly noticeable in rads ans waterblocks, under CFD or other flow analysis tool scrutiny. The question is, whats the difference when "ambient" or "static" pressure of the flow at this point is "high" or "low" ?

Notes:
* I doubt that with rigid tubing the pressure "suddenly" drops at pump inlet. Correct me here if i'm wrong.
* i've got a rather powerful pump (rated at 27W), but i doubt it dumps all its 27W in water. Actually it's quite hot to the touch, and since it has fins i suspect it dumps a good part of its heat in the air. Besides the engine (hot part) is totally insulated from the water chamber, since it's a mag drive. The percentage of heat dumped by the pump into water has been a hot topic but until now i've seen no convincing answer.

gone_fishing: you got it right.

[myv65]

Originally posted by gone_fishin
If the heat delivered to the rad is in very small quantities then the time needed to dissipate is also small in the rad. The amount of heat absorbed per volume of water before it reaches the rad is smaller in high flow but repeated many times over compared to low flow in the same time frame. Is my understanding in this flawed?

I'll tell ya, I'm not sure if I should reply here or to another thread running around here that says basically the same thing. You are correct in all aspects, but you don't really state a conclusion. Barring that, I can't say if you are right or wrong.

I posted that last bit late at night and didn't really put down quite all I wanted. Well that and the fact it was already a really long post.

Just about everyone seems to get the fact that energy input to a cooling system is almost constant regardless of flow. It is not quite constant because processors don't put out truly constant heat and pump energy will change ever so slightly as system temperature (and fluid viscosity) change. So if we ignore these minor changes, we can say that flow rate multiplied by delta-T equals energy transferred. Not stating anything new here.

Somewhere else in another thread, JimS said something like what you said above. Namely, the percentage of time water spends in the radiator is independent of flow rate. His reasoning thus went along the lines of at flow = 1X, the water is in the radiator 5% of the time. At flow=2X, the water is still in the radiator 5% of the time. Cooling should therefore be roughly equivalent.

The fact is, energy transferred must be equivalent, which means that the delta-T at 1X is two times higher than the delta-T at 2X. This says absolutely nothing, however, about the actual fluid temperatures.

Also, consider a person that uses a reservoir. Given the same tubing and flows, maybe their fluid only stays in the radiator for 2% of the time. Assuming the same flows, does this mean that their radiator somehow performs poorly in comparison? Absolutely not. A radiator has no "brain" to know that is has the water for 5% versus 2% of the time. It knows only flowrate (fluid AND air) and temperatures.

There is no steadfast answer that applies to all situations regarding "what's the best flow". The general rule-of-thumb answer is "use as much flow as possible without dropping the residence time in the radiator too much". OK, so that doesn't really tell us anything. What does it mean?

It means that heat transfer in an air-cooled radiator by its very nature has its peak cooling at the entrance and its minimum cooling at the exit (as fluid and air temperature approach one another). When the residence time is too low or the air flow is too low or the surface area is too low, the delta-T between fluid and air must increase to get the same quantity of heat transferred. When residence time is high and air flow is high and surface area is high, the fluid will exit at practically the same temperature as the air. As you begin to decrease any of the three variables, the fluid-air delta-T will begin to increase. It will increase gradually at first and more dramatically as you continue dropping residence time/air flow/surface area.

So long as you don't "go over the hump" where the delta-T curve gets very steep, the flow through the radiator is not yet too high. Simple to state, hard to put an absolute number on it (and it's a different number for every setup).

[myv65]

Originally posted by gmat
the quite low temperature differential accors the circuit (1°C or less). Is 1°C enough to produce a noticeable static pressure ?

I don't have my texts with me, they're at work. Water reaches its peak density at 4°C. It is sort of parabolic going away from 4°C. 1°C won't change the density much, probably on the order of 0.01-0.1%. To know the pressure rise, I'd have to verify this number, know the tubing rigidity, and look up the *real* bulk modulus. You may be surprised how much static pressure such a small temperature change can generate if the system is truly rigid.

i've been wondering what was the *real* pressure distribution around a closed loop of water. Maybe CFD would answer that..
You seem to say that in every case the pressure is positive around the whole circuit, which i doubt by experience. (say, i only "doubt", ie i may be quite wrong on this).
What i'm considering here is the dynamic effects.
In a turning flow; the radial kinetic forces tend to create a pressure on the outer side, and an underpressure in the underside (Bernouilli, if my memories serve me right). This effect is mostly noticeable in rads ans waterblocks, under CFD or other flow analysis tool scrutiny. The question is, whats the difference when "ambient" or "static" pressure of the flow at this point is "high" or "low" ?

No, I don't say that pressure is positive around the whole loop. It may or may not be, depending on the system. I will guarantee you that it is positive in an open system provided the reservoir is at the high point. In an open system, it may be under a slight vacuum in the last leg if the reservoir is at a lower elevation.

In a closed system, "most" often it is at a slight vacuum leading to the pump suction. You can easily change this with a standpipe filled with water, though you may need a higher ceiling. If you place a T ahead of the pump suction and run a vertical (open at the top) line and fill it with water, the water height in that run will define the pressure at the suction (and eliminate any chance of having a vacuum). This is obviously not a practical way to run a system, but does serve to illustrate some points about pump systems.

I'm not quite sure what you're getting at with your comments about radial flows. Yes, there are secondary flows caused by bends and temperature changes among other things. Yes, they create localized variations in velocity. This manifests itself as localized variations in convection coefficient. Sure, you could model this with CFD, but your answers are only as good as your assumptions placed on the model.

More practically, you can run pressure taps at the inlet and outlet of each individual component if you want to know the overall restriction imposed by each part. Knowing the delicacies of secondary flows may be useful to a block designer, but I've seen little evidence that the major manufacturers really understand what they're doing from an engineering standpoint. Fact is they don't have to understand it well. I say this for two reasons. First, marketing drives sales because the majority of buyers buy stuff that "looks cool" or they heard about from a buddy. Second, there really isn't a dramatic difference in performance among the main blocks. So long as the block is sufficiently restrictive as to create a high velocity via a pressure drop, it'll do just fine. Where blocks will disappoint is if something else in the system constricts flow so much that the block velocity is too low.

To answer your final question about ambient or static pressure, it's irrelevent. Static pressure by its very definition is the same in all directions. This means that it can't affect flow rate (other than increasing flow area in soft tubing), so has no, repeat no effect on heat transfer. Relative pressures determine flow rate and are all that really matters.

Notes:
* I doubt that with rigid tubing the pressure "suddenly" drops at pump inlet. Correct me here if i'm wrong.
* i've got a rather powerful pump (rated at 27W), but i doubt it dumps all its 27W in water. Actually it's quite hot to the touch, and since it has fins i suspect it dumps a good part of its heat in the air. Besides the engine (hot part) is totally insulated from the water chamber, since it's a mag drive. The percentage of heat dumped by the pump into water has been a hot topic but until now i've seen no convincing answer.

Item 1, you are correct. Sudden drops only occur when there are sudden changes in the flow path. Fittings, bends, Ts, valve bodies, etc., cause "step" changes in pressure. Tubing causes a linear "loss per length" based upon flow and internal surface roughness along with Reynold's number.
Item 2, tough to answer. Most motors run on the order of 80% efficiency. Most centrifugal pumps that we use peak out around 60% efficiency. This peak occurs when discharge resistance is moderate, neither a minimum (peak flow) or a maximum (dead headed with zero flow). Because of our systems, we tend to run higher head pressure on the pump than the "peak efficiency" head pressure. So figure your pump is ~40% efficient overall. It also doesn't run at 100% rated motor current, say maybe 80%.

Take all this together and you probably dump ~27*.8 (load)*.2 (motor inefficiency) as heat from the motor fins. The remainder of energy input (27 * .8 * .8) gets put into the water. Of this, flow rate * delta-P (outlet - suction pressure) is "useful work" and the remainder is pure heat. Even the useful work is still energy, though, so is energy that enters the system and must be removed by the radiator.

Tell ya what, I'm part way through a series for AMDMB discussing all facets of water cooling. There's also a couple of articles discussing heat transfer and fans. For the water cooling parts, all that is currently posted is the introduction and pump section. The fluids section should be up within a week or so. The real fun starts with the next one after that covering radiators and blocks. There's sure to be material there that throws some people into a tizzy fit. If you haven't already seen it, feel free to stop on by for a look. The pump article is "stickied" in the liquid cooling forum and the others are "stickied" in the cases&cooling forum.

[Sirpent]

myv65, is it true that as long as we think in terms of delta-Ts for the radiator and waterblock (ignoring frictions in the system), changing the flow rate does not make any difference in the cooling efficiency?

[myv65]

Sirpent,

Your question represents a potential can of worms. Strictly speaking, efficiency is generally defined as useful work produced divided by required energy input. In cooling, I would interpret this as minimum cooling system power consumption while keeping the computer operational. Under this strict definition, lower flow will always win because chips can run a lot hotter (at stock speeds) than we tend to let them. Again, strictly speaking you can cool a top end XP at stock speed with room temperature water flowing at a rate of about 1-2 gph provided you've got a block designed to work with that flow and can stomach the temperatures you'll have.

If you've got some other definition for "cooling efficiency", describe it and I'll give my best shot at an honest answer.