I don't think this is correct. I know of other instances involving automobile cooling systems where increasing the flow rate has lead to overheating of the engine because the radiator couldn't cool the water fast enough.

It is not a misconception that water needs to spend time in the radiator for optimum performance. The reason people may think this is because they are confusing heat transfer rates with what we ultimately want the radiator to do.

The radiator works best at 0 flow. At 0 flow, the water in the radiator will be cooled down to ambient. That the coolest water we can get out of the radiator. As the flow increases, the temperature of the water, at the outlet, increases. This is exactly what we don't want. It does not matter if the heat transfer is occuring with better efficiency. The fact is that the output temperature is increasing, and that's a bad thing.

It is also a mistake to discount the cause-&-effect relationship between the waterblock and radiator. Increasing flow lowers the temperature of the water coming out of the block, and that greatly affects the performance of the radiator.

It is an absolute truth that faster flow increases the heat transfer rate. However, you cannot blindly apply this truth to an entire system. Reality is far too complex to allow that.

you are a bit of a puzzlement Greystar
really ignorant and obstinate about some things,
yet having a good grasp of some rather difficult counter-intutitive topics
- lapping and rad performance for eg.

can't see why you guys are beating your gums, and ignoring actual data

http://www.thermal-management-testin...issipation.htm

am just completing 2 wks of (private) rad testing and crunching the numbers now
will post some heater core data in the next several days
most enlightening - to me anyway

Greystar: you are indeed comparing apples to oranges ;)

The amount of surface area of the heat source in a car is much closer to the amount of surface area of the rad. In fact, I wouldn't be surprised if the engine block had more surface area, in some cases!

In any case, an increase in flow would permit a decrease of the resistance in the heat transfer, from the engine to the coolant, as well as into the rad, from the coolant.

But in the case of a car, the rate at which the engine's heat is expected to be dissipated in the rad rises much, much quicker with a flow rate increase. In a PC, doubling the flow rate is something that the heatercore will always be able to handle.

In fact, if you think about the function of the heatercore, in a PC, it's only to drop the temperature of the coolant a few degrees (usually less than 10 deg C) above the ambient temperature. In a car, the temperature differential is much, much higher!!!

This goes back to the balance point, between heat extracted, and heat dissipated. The theory is the same in a PC, but the proportions and numbers are different.

Heatsource -> water -> radiator....seems like an apple to me!

The reason behind why increasing flow in an automobile can cause it to overheat is this. First when you increase flow you transfer more of the engine's heat to the water. Since you haven't changed the radiator or the air flow over it, the radiator can't handle the extra heat being transfered. It's size and the airflow limit how much heat it can remove from the water. When you excede the limit of the radiator the water just continues to heat up until it's boiling point is reached. Overheating.

I think it's a bad analogy because of this:

Car: 125,000 Watt heat source --> 900 cubic inch radiator

PC: 70 Watt heat source --> 70 cubic inch radiator

AND

A car has a different design goal for the cooling system.

Ever had a car that didn't run right until it warmed up?

Isn't the design goal to cool?

I think I can defend myself on this one.

"The radiator works best at 0 flow."

Very wrong!!!

What you are stating is that the water will be cooled further if it is left in the radiator for a longer period of time.

SO WHAT!!

Let us suppose that we have a radiator with an internal volume of 1 litre. This is attached via a pump to a reservoir (no heat load), with a capacity of 9 litres (including the tubes between res, pump and rad.

Lets say the water in this "system" is at 50 degrees C and we want to cool it down to ambient with the rad. There are TWO ways of doing this. The are extremes at either end of the scale of reality, though they are reasonable for the argument here.

1. Pump one litre of water into the radiator, then leave it until it is cooled down to ambient. Then pump the next liter in and so on ten times until all of the water has been cooled to ambient.

2. Now pump the water at such a speed that the turbulence in the radiator reduces the barrier thickness (slow moving/stagnant water at the interfaces between copper and water) effectivel to zero. Keep pumping the water around until the water reaches ambient.

Now I ask you.

Which one will reach ambient first. In both cases, we are removing EXACTLY the same amount of heat energy from the water.

Answer - 2

Your approach to this is wrong.

I am correct in saying that you need to consider the whole system to be able to understand the implications of what happens when you change something like the flow rate.

In fact, rereading your post, you have already shot yourself in the foot. Your argument, be it already flawed, that the water leaving the block will be colder is preceded in the previous paragraph by a statement that water leaving the radiator will be warmer. These will cancel each other out.

For the sake of argument, lets assume that the water temperature does not vary throughout the loop. This is a reasonable assumption as it actually doesn't vary by much more than a degree anyway.

Now lets think in the opposite direction. Lets start from the air.

Lets assume we have a cpu producing 70W. The radiator only has to dissipate 70W No more, because there IS no more. The performance of a radiator as we require it, is determined by the efficiency.

C/W - the temperature difference between the coolant and the air required to transfer 1 Watt of heat.

This value decreases with increasing flow, as the barrrier region is broken down at the surface of the water. This is a layer of water which "sticks" to the surface. This is not good, as heat must get through this barrier from the faster flow of water in the center of the tube to the copper tube. Water is not a good conductor. What breaks it down - turbulent flow from faster flowing water.

So we have reduced the C/W value by increasing the flow.

Can you now see, that to remove a fixed value of heat, we will need a lower temp difference.

So the average coolant temp will be reduced.

Equally, the C/W ratio for the waterblock will also be reduced by the increased flow. This means the temperature diffeence between the cpu and the water can be lower while still having the block remove 70W of heat. But the coolant is already cooler than before (on average) and the cpu will be cooler relative to the water than before.

In essence we have reduced the temperature difference required at the two heat exchanging steps, and the temperature is fixed at one end, the ambient air temp.

So now, can you see that you HAVE to consider the whole system otherwise you will fall into the trap of believing that slow flow in a rad is good.

The only other thing to consider, is the "point of diminishing returns". This is the point where by additional heat produced by the larger pump, and friction in the tubing from increased flow, means that the overall heatload goes up.

At first, the increased efficiency will more than compensate for the additional heatload, but there will come apoint where the heatload will become to much, and more flow will result in and increase in cpu temperature. However, this opint is beyond what most people's systems are capable of, particularly with the heatercores and high flow blocks which dominate the market at the moment.

I hope this clears things up a little.

8-ball

It's to keep the engine at a specific range of temperature, not to remove all available heat.

8-ball: Nice explanation.

So let me see if I understand 8-Ball correctly. Please pound me mercelessly if I'm incorrect :dome:

So, with the water moving too slowly, we get a barrier of water itself that keeps the heat from properly getting to and from the flowing water to transfer the heat out of the system at the radiator and into the system at the source (CPU).

If we move the water too fast there is a point at which we will begin to add more heat than we are removing with the added efficiency.

So the idea is to get that balance where we move the water just fast enough to break down that barrier of stagnant water with turbulance and cavitation, but not so fast that we are adding more heat with the need for a stronger pump that only contributes to the problem.

Is that the basics of what you were saying 8-Ball? I feel this thread has taken a toll on my grey matter... :drool:

That's probably the best summary of the principles behind watercooling I have seen. Short and to the point.

Though you might want to remove "cavitation" as this is where negative pressure either at pump inlets are around ship propellers causes the water to vapourise. This is not really a good thing as it will wreck efficiency. Though I knew what you meant. Edit that out and it will be a good summary to this thread which I now have linked in my sig for the benefit of anyone who believes otherwise.

Glad I could help

8-ball

Yeah, you're on the right track now!

But the balance point is so far up there, that you're not likely to reach it: if you ever get a pump that puts in enough heat as to make a significant impact, post a picture of it, OK? ;)

BTW, the more "turbulent" the flow is, the easier it will be for the water to shed its heat (or pick it up). The measurement of turbulence is in Reynolds, where <2'000 is considered laminar, 2'000 to 4'000 is in the transition zone, and >4'000 is fully turbulent.

If you can run the numbers, you'll find out that no one is achieving fully turbulent flow (jet inpingement excluded).

and the toll would be far higher were facts used instead of 'theoritical' assumptions

8-Ball is wrong
those rads that I have tested show a C/W 'peak' (a minima actually) at 1 to 1.5 gpm
(will try to post this afternoon)

now what ?
as a friend of mine (expert in CFD modeling with a 30 teraflop computer) says:

another beautiful theory destroyed by facts

Isn't higher flow better for a rad? That's what I was told.

That's not too inconvenient: it's really hard to push 1.5 GPM through a heatercore anyways. It makes the case for running two cores.

I'm very much looking forward to your article, Bill. Your contribution to the watercooling world is absolutely amazing. I thought you were going to drop the whole radiator issue, where we still needed more info.

Thank you.

That would probably be one of the posts from last night. Imeant to read over those and check for wine related mistakes.

However, it is still the case that 1.5Gpm is still quite highfor a watercooling loop. Add to that the fact that most waterblocks will continue to gain efficiency beyond that flow level.

BillA, can you tell me which post that was in as I haven't the time to reread them all. Too much revision to do.

As for the minima at 1.5GPm, If you were recording the temperature of water as it entered the radiator, could friction within the rad contirbute to additional heat while not appearing on your measured temp? Just a thought as I'm trying to get my head round it all.

8-ball

the coolant inlet temp (offset) is held at 5.0°C (±0.6) wrt the rad inlet air
coolant temp measurement resolution 0.01°C (±0.02 inst uncertainty) taken at inlet and outlet
C/W calculated using mean temp between inlet and outlet as the difference between this and the "logarithmic mean temperature difference" (normally used for heat exchanger calculations) is close to nothing for these deltaTs

note that the values are quite small
I am just observing that 'more is not necessarily better', even though that is what the theory says

and if you observe the graphs linked to it is seen thet some behave VERY differently
lots of reasons for such

I agree that few WCing systems achieve a 1.5gpm flow rate,
but the serious can do so without much difficulty
- just a LOT of attention to detail

Not from where I'm standing! :p

You're example, where you suggests that pumping water around will cool faster, is indicative of the general problem of understanding this issue....you continue to take components out of the system to review their behavior.

Pumping water around and around will NOT cool the water faster. Why? Because of every loop you have a waterblock adding 6 therms of energy to the water!

You will never understand what is happening if you separate the components.

The zero flow example was an extreme. Just like saying that at infinitely fast flow, the waterblock is the same temperature as the water and the water temperature never rises. It simply illustrates what the pertinent physics formulas will tell you.

Finally, when you reduce the temperature of the water, you increase the difficulty in extracting the same amount of heat energy from it. The closer the water temperature gets to ambient, the harder it is to cool it. That's why the benefit from faster flow is countered in the radiator. The water from the block is at a lower temperature, so the radiator will have a more difficult time performing it's job.

The easiest way to see this is to just create a simple hypothetical system and use the heat tranfer formula to see how transfer rates and loads are affected.

It is true that you will NEVER actually reach a CPU temperature that matches the coldest air temperature measured going into your radiator.

As stated the closer you get to ambient temp the harder it is to get to that point... kind of like a half-life or half-step... if, with every step, you cover half the distance to your goal... you will never reach your goal.

Practiacally speaking, you would get close enough that you wouldn't care about that last 2 to 5 degrees.

The CPUs we are trying to cool have an EXPECTED operating temperature of what... 40° C to 50° C? Average room temperatures in a home are what? 21° C to 25° C? That's a pretty big difference... if you can get your CPU core temperatures down under 35° C you are doing well in my opinion. If you are under 30° C you are doing really well.. in fact I am more than happy with just being under 30° C. If you get under 25° in a room that is 21° C without a peltier or some chilled water setup... you are doing EXTREMELY well.

Again, I was always under the impression that the goal was to keep our CPUs cool under load, especially when we overclock. If they look good and function quietly in addition... then so much the better.

My goal is to have cool and quiet... not just a pressure head that could lift a car and cool a volcano. :p

this is absolute garbage
do some (remedial for this site) reading

then make a list of each thermal impedance between the CPU and the air
then do some more research and ascribe a value to each
then add them up
-> then you will understand what tripe you just posted

get serious

I don't think I've ever been accused of posting "tripe" but thanks... :D

I'm just quoting what I know... my system temps are BOTH under 30° C... if this is NOT considered low... well I apologize to all the readers and will post no more comments on actual temperatures. :shrug:

What are you basing these "actual" temps on?

If you reread my post, though it is possible I didn't make the point clear enough, there is no waterblock in the situation I have described.

I was merely describing two situations where we are attempting to cool the same 10 litres of water from the same temperature with the same radiator and no additional heat load. The only difference is that in one scenario, we have water moving, breaking down the barrier layer, and in the other, we have the water stationary with effectively a full barrier layer, ie the thermal energy from the water at the center of the tube/channel must conduct through the water rather than turbulent flow randomly carrying it into the copper wall.

And yes, I agree that the water entering the rad will be cooler, but each "packet" of water is carrying less thermal energy.

It is plain to see that the radiator will dissipate more heat if the water is warmer, but it doesn't have to dissipate more heat. The heat load is set by the processor.

We know that the processor will rise to such a temperature that the thermal gradient across the WB baseplate transfers 70W of heat. However, this does not determine the water temperature directly. The water temperature will be set by the efficiency of the radiator and the resulting delta T required over the ambient air temp so that it can remove the 70W.

If you suddenly slow the flow rate down in a watercooling loop, it is the reduction in the efficiency of the radiator which will cause the overall "average" water temperature to rise. This in turn will cause the cpu to heat up to a temperature where the delta T is sufficient to transfer 70W to the water.

I'm not separating components as such, merely working back from the only fixed variable we have, the air temp, at each stage considering the efficiency of the heat exchanger and the corresponding delta T required to move 70W, which is all we are asking the system to do.

Ask yourself, this, how is it that increasing the efficiency of both the waterblock and the radiator would result in an increase in the cpu temp, as this seems to be what you're implying.

8-ball

Well, not having access to a "Lab quality" temperature meter... I have simply used several commercially available thermometers and averaged the results.

Once that was done at idle and under load I "adjusted" my MBM5 software to compensate. As stated before the sensors on most motherboards are neither accurate or placed in very good spots.

I then remeasured my temps with the external thermometers (yes they had thermister probes) and verified that my MBM5 temps were within 1 degree C.

I will see if I can obtain a temp meter from the lab and bring it home to verify.

My ambient air temperatures in my apartment range from 21° C to 25° C this is also based on those same external thermometers.

I have not directly monitored the water temperature, so I guess I can't give numbers for that. :shrug:

If my methodology is flawed... please relate the proper proceedure for measuring these temps without a lab quality meter. :confused:

BillA,

Am I being daft somewhere along the line because it appears that the theory which I am describing is sound, yet it is still causing trouble.

I would definitely like to have a discussion about why the radiators have a max efficiency around 1.5GPm.

The point I tried to make earlier was that no matter how accurate your measuring devices, you would not be able to detect any additional thermal load reducing from increased friction WITHIN the radiator itself as this is AFTER your initial temperature reading.

I'll have a think about this this afternoon and try and explain my thinking a little better. That's provided you don't point out any obvious mistakes which I'm making and have not seen.

8-ball

I take NO exception to the 'theory' of thermo or fluid mechanics
(even I am not so daft)
but 'reality' (in the form of un-accounted for influences) often skews the results

http://www.thermal-management-testing.com/6ddcube.gif

can you describe what is occuring here ? (pop quiz, lol)
note that max efficiency is unique for each rad

re the frictional heating
come on, more than offset by the internal pressure drop
WAY too small
and the readings are at the inlet and outlet, any such is included (as is the effect of the pressure drop - several threads on this)