Flow Rate Impact in Water Cooling (Summary)

[Graystar]

Guys I'm absolutely brand spanking new to water cooling and as yet I have to build a system, so tell me to butt out if you want

Oh please…join right in!

Ok, I can understand the bit about high flow rate being a good thing for a CPU water block (that seems pretty simple, faster flow means that means bigger delta of the coolant at the CPU's water block yes?)

Yes, but also remember the concept of more mass. The amount of water you have in your system to do this job is not based on how much is in the tubes or reservoir, but on how much water flows past any given point in a set amount of time. When speaking of mass, gallons mean nothing and gallons per minute means everything.

And I think I get what he is saying for radiator (low delta here is bad as the heat transfer is less?)

Yes. But remember, it’s a combo deal. You can’t talk about the heat transfer rate without stating how much mass you’re talking about. A low delta, with more mass, can be okay. The only question is whether the rad can dissipate the heat.

If I have got this right it means that you need to increase your radiator's capacity if you increase your flow rate through it?

Yes. Well, you MAY need to increase the capacity. If you had the capacity to begin with, CPU temps will drop.

My logic is this…
So in this case we need a rad of twice the capacity when using high flows than that of low flows?

I think you have the general principle. The situation is not as bad as you described, because, once again, and like many people do, you forgot to account for increased mass. But that’s okay, as long as you got the basics.

You actually are drawing more heat energy when you increase flow (and if everything else works out.) Some people argue with this, saying that the CPU is a constant heat source, and you can’t get more heat from it.

It is true that you cannot cause more heat to be generated by the CPU. However, heat transfer through the materials involved is not instantaneous, nor is it resistance free. The materials used do not allow 100% of the heat generated to be transferred to the water. The reason the CPU temperature is elevated is because some heat energy wasn’t removed by the cooling system. So, for every 100 watts of heat generated, there is actually a division of where the energy goes. For example (mind you, these numbers are being used because they’re easy, and are not indicative of actual wattage) say that 90 watts are being removed by the cooling system, with 10 watts remaining to heat the CPU to its current operating temperature. When you increase flow, and the rad has the capacity, 91 watts will be removed, and only 9 watts will remain to heat the CPU to it’s operating temperature. Hence, CPU temperature drops.

 

[geoffman]

Thanks guys for not laughing at me, and thanks for the responses it is most appreciated.

I'll have a think over your responses and see where this leads me.

For information I'm an electrical engineer, and thus tend to try and break things down into analogies of eletrical nature, which probably leads to my misunderstandings.

It works like this:
Say I have a device that generates 100W of thermal energy and operates in an environment of 30 degrees ambient, but can tolerate a case temperature of 80 degrees, I do a calulation to work out the "thermal resistance" in degrees per Watt of the heatsink I need as follows:

(Case temp - ambient temp) / power = "thermal resistance required, eg (80-50)/100 = 0.3 degrees / Watt.

(Actually I'd bung in a margin of safety on the case temperature as you should never run devices at max case temp, and that is a little oversimplied from what I do, but you get the idea).

This concept of "thermal resistance" is pretty common in the electrical engineering world, and being a really simplified system probably doesn't help me understand the nature of heat transfer all that well (and causes me to misunderstand it).

After I do my bits I usually give it to a fluids and thermo "johnny" to check and refine for me.

But hey I wanna learn how to do it properly for myself!

 

[geoffman]

problem with Darcy's law is that you need the Moody's diagram to determine the friction coefficient and to use the Moody diagram you need be able to compute the Reynold's number, so I think that will be diving a little too deep.

Prandtl, just as a matter of interest I found this online calculator for the Darcy-Weibach formula, is it any good?

 

[Graystar]

rogerdugans, let me try to wrap up this first round of explanation...

You should now be able to see the error you made when you wrote:

The reason higher flow rates work better in computer water cooling is this:
There is more water with a larger temperature differential moving through the water block- this removes more heat.
There is more water with a larger temperature differential moving through the radiator- again removing more heat.

In our trips through the water block we realized that when flow is increased the temperature differential that the radiator will see is smaller…not larger. This is so because the water temperature of the water leaving the water block has actually dropped.

Previously, those trying to demonstrate that faster flow is always better assumed that the water temperature of the water entering the radiator remains the same as the temperature of the slower-flow rate. You should now see that this assumption is the cause of the misunderstanding. They failed to account for the effect the water block has on temperature. That’s why I say that you have to consider the effect of flow rate on the entire system as a whole.

When you increase flow rate and just assume the temperature to be same, you have increased the heat energy in the fluid. If you now try to calculate the heat transferred in the radiator, you will, of course, transfer more heat. The problem, however, is that the extra heat is there because you put it there through a slip-up. That is, through a failure to keep in mind the relationship between Delta T, Mass, and the energy stored in the fluid…and not because the transfer rate has somehow increased.

I said before that increasing flow is to increase the mass cooling the water block. It is nearly the same as using a larger heat sink in an air-cooled system. That is why the trend in air coolers has been larger heat sinks, and larger fans (more air.) More mass needs more cooling capacity. When we increase flow, we have more mass, and that mass needs more cooling capacity.

The real question you want to ask is…does increased flow increase the cooling capacity of the radiator? The answer is, it depends on the radiator design. Some rads experience an increase, some rads don’t. Some rads even have a decrease.

 

[rogerdugans]

Hmmmm.

I do think I begin to comprehend your points, somewhat anyway, but parts of it are very difficult for me to correlate with my own experiences.

I do recognize the biggest hurdle I have: we (or at least I) started off discussing just the radiators performance. I have never been able to really isolate components well in an actual system: best I can do is replace one part with something else, ie: block swap, different pump or tubing change. Any part I have ever changed does require at least minor changes in other areas, tubing if nothing else, which makes the validity of the results more questionable I admit (although 4 inches of tubing is not a major change.)

What I have always been forced to do is look at the performance of the entire system and try and understand what is happening- without the math.

Perhaps my error comes from all the radiators I have used having extra capacity: chevette heater cores, a 6"x12" chrysler core and a Swiftech dual 120mm fan radiator as well as one old welding machine radiator.
Because in each and every case of swapping pumps that I have tried, cpu temperature has gone down at a given cpu speed/voltage with radiator fans at a given speed. Many of my mad-scientist experiments were carried out with XP 1600 Pally chips @2Ghz or so, which is about 120 watts of heat or so.

There is much that I do not know about the subject, and little that I know about the principles behind it, but real-world results are something I am pretty familiar with, and I have never, ever seen a lower flow pump give better results in a given system, although I have heard of a few instances.

I cannot explain this- your explanation appears to make sense and I am unable to refute it, but it disagrees with what every experiment I have tried has shown me as far as the use of higher flow rates.

Perhaps you can help me to understand....

 

[Graystar]

but parts of it are very difficult for me to correlate with my own experiences.

Why? I never said that CPU temperatures wouldn’t go down. My point is simply that you cannot assume that CPU temperatures will go down.

The temperature of your CPU is based upon the radiator’s ability to cool the water. Nothing else matters. That is simply the bottom line.

BillA's radiator testing shows that the cooling capacity of a radiator may go up or down with increased flow. Therefore, the effects of increased flow will depend upon the design of your radiator, and it's overall capacity to cool.

 

[JPSJPS]

With all other factors being equal however (which is a real-world impossibility) a higher flow rate will give better performance.

rogerdugans - When you make a statement that is generally true without stating limits or conditions, you sort of set a trap for yourself. Most folks will know exactly what you mean but you still leave yourself open to argument.
Obviously as the flow rate increases you will reach a point where the performance increase is too small to measure. And, at some point, the energy added by the pump will be larger than the energy removed due to the higher flow rate. (I realize that I am not adding anything that you don't know)

As you know the amount of energy transferred from the radiator to the air is directly proportional to the temperature difference between the two. Thus, we want a high enough flow rate that we get a minimal temperature drop across the radiator. If the flow rate is low, the output end of the radiator will be cooler and we transfer less heat to the air on that end.

High flow rate ALSO means that we will have a minimal water temperature rise through the water block. Thus, by the same logic, we have the largest temperature differential between the water block and the water and the heat transfer to the water will be the greatest.

Conversely, if a low flow rate allows a water temperature drop through the radiator, obviously we will have a temperature rise in the water block. Thus, this warmer water will not cool the water block & CPU as well.

Some folks here are trying to analyze the radiator cooling for specific conditions and then draw conclusions from that. BUT, this is a closed loop system. Required is a CLOSED LOOP analysis which includes the transfer function (thermal resistance etc) of all components in that system. The analysis results will include a transient (depending upon initial conditions) and then will decay to the final steady state condition.

For a non technical type, this closed loop theory/analysis will probably be hard to understand. The "how it works" of closed loop systems is generally NOT intuitively obvious even to a technical type.

 

[JPSJPS]

My logic is this (may be faulty so please correct me if I'm wrong.) Ok in both cases we are removing the same amount of heat energy...
In a low flo system this may raise the coolant temp by 5 degrees, and in a high flow system it may raise it by 2 degrees...
Now for our system we want to get it our water as close to ambient as possible, so in the first case (low flow) we want a rad capacity of 100W/5 degrees or 25W/degree, and in the high flow case 100W/2 degrees or 50W/degree capacity.

So in this case we need a rad of twice the capacity when using high flows than that of low flows?

Dunno If I explained that right or if I am right, but that seems to be what Graystar is saying?

Welcome geoffman!! - Glad to see another technical type here. I am a retired EE myself so I am pretty sure I can come up with an explanation that will make sense to you. I learned a fair amount of thermal stuff working with MEs. They modeled the individual devices and I converted those models to the electrical equivalents and performed the analysis for them. This allowed the use of powerful electrical/electronic CAD analysis programs like spice etc.

Your quoted analysis is faulted because you performed an individual component analysis (correctly) and then used those results to predict the operation of a *CLOSED LOOP SYSTEM*.

An analysis of these cooling systems is analogous to that of an electrical closed loop feedback circuit. The analysis requires control system or servo theory because the output of the radiator is fed back to its input through the transfer function of the water block.

Lets try to look at your analysis intuitively. Obviously you are correct that a low flow rate would result in a higher coolant temperature so a given radiator would transfer more heat to the air. BUT, a higher coolant temperature would mean a higher water block temperature which would mean a higher 100W CPU temperature! This is the opposite of what we want. And in fact, even this attempt at intuitive analysis is also completely faulted because we have a closed loop system.

Following is a simplified analysis that will (I hope) demonstrate the superiority of highest flow rates:
We build a system with an infinite flow rate. Thus, the water temperature will be the same throughout the system. This says that the water provides a perfect no loss thermal conduction path between the water block and the radiator and they will be the same temperature. For sake of analysis we can even eliminate the water path completely and connect the water block directly to the radiator. Obviously, this is as good as it is gonna get. If we lower the flow rate, the effective thermal resistance between the water block and the radiator will increase. Thus, all temperatures will increase. As we reduce the flow rate things get hotter and hotter. If we lower the flow rate enough the CPU will get too hot and burn up! This will occur even if we use a HUGE radiator.

The bottom line is that a feedback system is not easy to analyze. Intuition and "gut feel" generally do not produce correct results. Ya gotta do the closed loop analysis which you probably know is usually not simple.

BTW - I bet you do not recognise the word "Spirule". Times sure have changed!

 

[geoffman]

The bottom line is that a feedback system is not easy to analyze. Intuition and "gut feel" generally do not produce correct results. Ya gotta do the closed loop analysis which you probably know is usually not simple.

BTW - I bet you do not recognise the word "Spirule". Times sure have changed!

Ah dear, feeback loops and transfer functions, now it's becoming clearer and was silly that I didn't really recognise it when I should have, and yes I was attempting to understand it in a piecemeal fashion. I really have no experience at all with closed loop thermal systems.....

As for a spirule, are you refering to the root locus plotting device ie round slide rule on an arm, that was great for vector maths?

 

[JPSJPS]

Ah dear, feeback loops and transfer functions, now it's becoming clearer and was silly that I didn't really recognise it when I should have, and yes I was attempting to understand it in a piecemeal fashion. I really have no experience at all with closed loop thermal systems.....

I bet you would have no problem with the analysis once everything got converted into electrical equivalents

As for a spirule, are you refering to the root locus plotting device ie round slide rule on an arm, that was great for vector maths?

I am impressed! Most engineers today have seen a slide rule but they think you are joking when you mention Spirule. I wish I had not thrown mine away although I hated to have to use it. And to tell the truth, I never completely understood what I was doing when I used one. Glad those days are over.

 

[rogerdugans]

JPSJPS-
Your point about me making a fairly broad-based open statement with giving any conditions is a very good one.
Not the first error I have ever made and surely not the last however.

There are many areas in which I was/am deficient as far as truly understanding all of what I wrote about in the original summary, but that in a way, was why I wrote it in the first place:
The original sticky on this topic contained a wealth of valuable technical information....much of which is far beyond my level of understanding- and possibly beyond others as well....
But the information inside is very valuable in computer water-cooling and I wanted to both better understand it myself, and try to simplify it for others without the technical knowledge required to really comprehend it fully. And I had hoped to do it without too many errors....!

This most recent discussion on the matter has both broadened my knowledge of water cooling (and MUCH more!) and taught me to be even more careful in just how I make broad statements.
I thank all who have participated for that. (Me more learn stuff. Me think that GOOD!)

 

[JPSJPS]

JPSJPS-
Your point about me making a fairly broad-based open statement with giving any conditions is a very good one.
Not the first error I have ever made and surely not the last however.

I was not being critical - I have made the same mistake many times!

 

[__TRONIK__]

time to bu a stronger pump for my whitewater and bury this home depot 200gph monstrosity!

 

[FrogBite]

Has anyone actually done an experiment where you decrease the voltage supplied to the pump and recording that against CPU temperature?