• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Flow Rate Impact in Water Cooling (Summary)

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

rogerdugans

Linux challenged Senior, not that it stops me...
Joined
Dec 28, 2001
Location
Corner of No and Where
The post(s) that follows is an attempt to condense the Flow Rate Sticky into a form slightly shorter than 17 pages of posts.......

I have tried to gather the conclusions from that thread and represent them in a short, accurate and easy to read fashion.
I also tried to verify that my conclusions were correct by googling, reading and getting some proof reading done by other members here.

If you find anything that is wrong or anything you think should be included- PLEASE LET ME KNOW!

This is part of the attempt at re-organizing/redoing our water cooling stickies, so it needs to be right folks ;)
 

rogerdugans

Linux challenged Senior, not that it stops me...
Joined
Dec 28, 2001
Location
Corner of No and Where
Does more water flow = better cooling?


I wrote this in an attempt to reduce a large amount of data (which was largely in one very informative thread which was 17 pages long) to one simple and fairly easy to read document.
The original thread is HERE for anyone who wishes to get more in-depth knowledge on this subject.



Heat Transfer

Heat transfer is the basis for ALL computer cooling systems; in water cooled computers we make this more complicated by using multiple heat transfers:
cpu core to water block
water block to water
water to radiator
radiator to air.

*Heat transfer works best with the biggest temp differential: ideal would be cold water to cpu and hot water in the radiator. We cannot achieve this because a closed loop will achieve equilibrium at some point. The principle holds true however- the most efficient transfer happens at the greatest temperature differential, therefore higher flow rates will always help with all other variables remaining the same.
* It is the Heat Transfer that we want to maintain as efficiently as possible, and that is best done with a higher flow rate. Rather than thinking that there won't be enough time for heat to move towards the cool water, and therefore compromising heat loss, it is better to think that there is more fresh water moving onto the CPU and therefore, there is increased cooling.

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.

This is true even though a system with a lower flow rate will have more time to heat the water in the block and also more time in the radiator to cool the water: since the heat exchange works best with the greatest temperature differential, longer “stay time” is counter-productive.
--------------------------------------------------------------------------------------

Sources of Confusion

There are some variables that have made this more confusing in practice though:
Pump Heat
Type of Flow: Turbulent or Laminar
Friction
Component Flow Resistance
Pump Design

I will attempt to end the confusion on these points next.

Pump Heat:

Pumps generate heat; rather than explain why, let it suffice to say that if you put your hand on a running pump it WILL be warmer than one that is not running.
This heat has to go somewhere: a submerged pump (inside of a reservoir) must add all of the heat it generates to the water; inline pumps are usually designed to use the pumped fluid as a coolant, so most of the heat is going into the water. There will be some amount of heat being conducted to the outer surface of an inline pump, but this should be considered a fairly small amount of the heat produced.

Now, in a simple water cooling system (cpu water block only) we actually have two sources of heat: the cpu and the pump.
Pumps with a higher Flow Rate will generate more heat than pumps with a lesser Flow Rate, and there lies our first bit of confusion:
It is possible to add more heat from a larger pump than will be removed by the higher Flow Rate.

Flow Type:

Water moving through a vessel (tube, block or radiator) meets resistance at the walls; this causes the water at the walls to move more slowly than the water in the center of the tube: this is Laminar Flow.
Laminar flow is bad for heat exchange because the water against the vessel’s walls is slower than the water in the center. Flow rate at the heat exchange surface has diminished.
This is where turbulence comes in - if we can get fluid from the center of the vessel to mix with fluid toward the walls, we end up with more efficient heat removal. Turbulence increases heat transfer significantly over slower Laminar flow.

Turbulent flow occurs naturally in a pipe when the fluid velocity exceeds a certain point, which is dependent on a lot of factors. Also, turbulence isn't an on/off thing - you can have more or less of it. Moving faster will result in more turbulence.
So, in short moving water through the block faster improves heat transfer between the block and the water, which reduces the temperature differential between the block and water required to move an amount of heat.
It is NOT intended to reduce the temperature increase in the water as it travels through the block, but rather to allow more heat to be removed.

Faster flow means more turbulence, and that is a good thing.
--------------------------------------------------------------------------------------

Friction

I am not going to say much on friction here: it generates heat in the pump, and it also generates a minute amount of heat as water flows through system components (a system with greater Head, either Friction or Static, will produce more heat however.)
The main area friction is involved in a water cooling system is in flow resistance- the next area to be covered.
--------------------------------------------------------------------------------------

Component Flow Resistance

Pumping a liquid through a tube creates resistance. The resistance is determined by the cross section of the tube, the length and all fittings in the line.

Static Head (or Lift) - number of feet of elevation that the pump must lift the fluid regardless of flow rate.
.
Friction Head- measure of resistance to flow (backpressure) provided by the pipe and its associated valves, elbows and other system elements:
A smaller tube diameter will have greater resistance: even with identical fittings, pumps and water blocks, a system with larger diameter tubing will have a higher flow rate.
A longer tube will also have greater resistance: even with identical fittings, pumps and water blocks, a system with shorter tubing lengths will have a higher flow rate.
A straight length of tube will have less resistance to flow than one that is bent. A partially kinked tube easily proves this point. Any bend at all introduces some restriction to the flow: a sharper bend is more restrictive than a gradual bend.

I was not able to get data on the most commonly used tubing in water cooling- Tygon, Clearflex and vinyl, but I was able to get data on copper tubing which illustrates the point nicely.
Copper Tubing Flow Resistance by Linear Foot (loss expressed in psi) Conversion: closest I have come to is 1 foot = 1 psi.
Copper Tubing Flow Resistance for Fittings (loss expressed in foot equivalence)
(NOTE that these figures are NOT accurate for the tubing used in most water cooled systems- I include it only to show the relationships between Flow Rate, tube diameter and the use of sharp bends or fittings.)

Head- the entire amount of flow resistance in a system. Static Head + Friction Head = Head
This is what pump head capacity must overcome and is entirely responsible for the reduction of flow rate in a system.
--------------------------------------------------------------------------------------

Pump Design
Positive displacement pumps will maintain constant flowrate but increase pressure as line restrictions interfere.
Most pumps used in water cooling are centrifugal pumps and these are NOT positive displacement pumps.
Getting the pump with the highest flow rating is NOT necessarily the best answer: centrifugal pumps tend to be extremely sensitive to flow restriction.
A pump with a higher Head Capacity will be less sensitive to restriction and be more suitable for computer use.
Which brings us back to the issue of pump heat ;): a pump with more head capacity and higher flow rate will add more heat to the system.
--------------------------------------------------------------------------------------

A Bit about Fans

Just as higher flow rates remove heat from the cpu faster, greater air flow rates through the radiator will improve performance at any given temperature. The actual equations differ since the fluid characteristics- water and air- differ, but the same principles apply.
Similar to a water pump, the pressure needs to be taken into consideration and axial fans don't usually provide a lot of pressure; thicker fans generally will provide more air pressure.
Example: a 120mm fan 38mm thick should be better for our purposes than a 120mm fan 25mm thick.)

The advantage to more airflow is that it provides a good “bang for the buck” improvement in performance; the disadvantage to this is that a fan will create more noise than a water pump, and noise is often one of the areas we are trying to improve with water cooling
--------------------------------------------------------------------------------------

Conclusion

A higher flow rate will give lower temperatures as long as NO other variable is changed.
Heat exchange is improved with turbulence- higher flow rate THROUGH A GIVEN DIAMETER TUBE will flow faster and be more turbulent.
Flow rate can be increased by using larger diameter tubing, the shortest total length of tube possible, fewest bends and fittings possible, lowest restriction radiator possible and by using a pump with a higher head capacity and flow rate.
Maintaining good airflow through the radiator is probably the easiest and noisiest way to improve performance.

Choosing an appropriate pump may be the hardest part: a high head capacity is best, high flow rating important but less so; close attention MUST be paid to the amount of heat generated by a pump as this heat will be added to the system.
--------------------------------------------------------------------------------------



Formulas used by the sources of data below:

Head Loss formula:
Williams and Hazen: Head Loss H= 0.2083(100/C)1.852 * R1.852/D4.8655
Where H=feet of water per 100’, C= Constant for different materials, R=Flowrate (gpm) and D= Inside Diameter. (NOTE: copper is 130. I was unable to find vinyl, Tyron and Clearflex; if anyone knows a verified source for this information, please let me know!)


Q=M x c x Delta T.
There is an elementary equation from basic thermodynamics that states that the rate of heat transfer (Q) equals the mass flow rate (M) times a constant (the specific heat of water) times the delta T (fluid temp out minus fluid temp in).

Links to source information:

Old Flow Rate Sticky
www.pump.net
www.copper.org/
Google



Much of this information is from the original Flow Rate Sticky and credit is due the original authors.
Thank you for your contributions.

Any inaccuracies are mine ;)
Please PM me with any errors you see and include source information.

EDITED- 3 July 03: added fan pressure consideration to "A Bit about Fans"
 
Last edited:

bigben2k

Member
Joined
Jun 17, 2002
Location
Texas
A good idea, thanks Roger!


Under "Heat Transfer", you mention delta-T as being the biggest factor to improve efficiency. Although that's probably true, turbulence will also lower the resistance of heat travel. It's mentionned in the next paragraph, but not in those words.

I don't know if you want to mention it, but turbulence can be achieved in two ways: high pumping action, or "turbulators" (including jet inpingement). The first is not attainable, because the pumping pressures required are beyond our normal pumping abilities.

Under "Pump heat", you make reference to the heat induced by the wire windings, but not the pumping action: it's covered lower, under "Friction", which is semi-accurate. The smoothness of the tubing is not a factor in flow resistance, for laminar flow, but the effect that you describe is there. I'm not sure how I could describe it better. The Hazen Williams formulae should be replaced by Darcy's, but if you want backup on H-Z, you'll find it at: (I'll dig up the link).

Under "A Bit about Fans", you ought to mention that, similar to a water pump, the pressure needs to be taken into consideration. Axial fans don't usually provide a lot of pressure. Personally, I've opted to use an automotive blower, because the squirrel-cage design can provide that pressure, and the flow that I'm looking for, and can do it quietly. I'll report on that later.
 

rogerdugans

Linux challenged Senior, not that it stops me...
Joined
Dec 28, 2001
Location
Corner of No and Where
Glad to do it: it has been a lot of work so far, but my understanding of all the factors involved has deepened greatly, and should continue to do so. ;)

The points you make: not that I don't believe/trust you, but I have tried to verify each conclusion I make with at least two sources, so please explain your points deeper and include any sources I may need to comprehend.
And I may need a lot: I have absolutely NO background in engineering and none in science since high school- 20 years ago!

Turbulence improves heat transfer by allowing more water molecules to contact the heat conducting surface (ie.: water block and radiator.) Is that what you mean? Or are you referring to water molecules having less resistance to transferring heat between each other?

I did not go into turbulators, although I did run across them numerous times, maily because they hurt flow rate severely. Since the pumps used in water cooling are extremely sensitive to flow restrictions, use of turbulators in our application would most likely be counter productive: any gain in efficiency from turbulence would be lost (and more) by the loss of flow rate. So for simplicity, I did not mention them.

Ok, your next comment caused me hours of research, and I am still not entirely clear on the subject....
I did find some actual data but most of what I found indicated to me that while heat generated by flow through components (including tubing) is real, at our low flow rates and low pressures it is pretty small and probably not much of a concern.
The amount flow is reduced by fittings and tubing however is pretty large, especially with our low pressure pumps.
I saw nothing on Darcy's formula but I will look it up in a bit. Any links you suggest will be good :)

Fans: You are unquestionably correct sir :D
I wanted to keep talk of fans minimal (as this is about coolant flow) but it has such a large impact that its kind of silly to not mention it. That will be added.

Eddicate me more, bigben2k :D
 

will_maltby

Member
Joined
Nov 22, 2001
Location
UK
IMHO maybe you could add in some implications to real world situations and/or provide a few examples of setups with regard to the actual measurable affect of changing a pump in a given system.

For pragmatics like me, formulas can be a bit abstract and possibly overkill given the potentially small effect an upgrade of pump may have.

Basically, it'd be nice to see what impact going from say, and Eheim 1048 to a 1250 pump would have on measured cpu temps in a typical watercooling system. This would be a great aid to watercooling newcomers so they'd instantly be able to see what kind of pump they'll be after for a new system of theirs.

Just my 2 cents, hope I'm not ruining the intention of this thread :)
 

bigben2k

Member
Joined
Jun 17, 2002
Location
Texas
Kewl:cool:

Ok there's a fella that wrote an article (Mike ?) right here at OC, on turbulence (water cooling section). It isn't quite complete, but a good primer on turbulence.

The idea is that turbulence breaks up the boundary layer normally found in laminar flow. The more turbulent the flow is, the more that boundary region is reduced, and you then have a lower resistance for heat to be transferred to the coolant. Coolant speed is a factor, but not by itself.

Turbulators are used in just about everyone's rig: why do you think that heatercores look the way that they do? ;) Also, there's a turbulator in Hayden tranny coolers, which makes them awfully efficient, for their size.

The heat from the restrictions is indeed small, but it becomes a factor when people get into these huge pumps, like Bill's Iwaki #70, or any of the other wackos with a pump that's got more than a 50 feet head (no offense Bill ;) ).

Here's a bunch of links on turbulence:
(Hopefully, most still work)

http://www.mas.ncl.ac.uk/~sbrooks/book/nish.mit.edu/2006/Textbook/Nodes/chap07/node9.html
("Turbulent Reynolds stress")

http://www.wikipedia.org/wiki/Reynolds_number
(basic definition)

http://www.ichmt.org/abstracts/MECT-02/Abstracts/2-2.pdf

http://www.nag.co.uk/simulation/Fastflo/Documents/ToolBox/html/node8.htm
("Theory of laminar and turbulent flow")

http://psdam.mit.edu/2.000/Administrative/handouts/Hydraulics/Turbulent-Flow.pdf
(simplistic, but there)

http://www.efm.leeds.ac.uk/CIVE/CIVE1400/Section4/laminar_turbulent.htm
(dye experiment)

http://www.sigmaxi.org/amsci/articles/97articles/Hademenos-5.html
(a medical perspective!)

http://wuche.wustl.edu/~sato/flowtrans/flowtrans2a.html
(Reynold's experiment)

http://www.icase.edu/Dienst/Repository/2.0/Body/ncstrl.icase/TR-99-33/pdf
("Streamwise Vorticity Generation in laminar and turbulent jets")

http://home.olemiss.edu/~cmprice/lectures/turb.html
(some basics about turbulent flow)

http://www.uts.com/products/tkintro.html
(A software)

http://www.me.mtu.edu/courses/me328/scans/notes/me328formula.pdf
(A formulae roundup)
 

bigben2k

Member
Joined
Jun 17, 2002
Location
Texas
will_maltby said:
IMHO maybe you could add in some implications to real world situations and/or provide a few examples of setups with regard to the actual measurable affect of changing a pump in a given system.

For pragmatics like me, formulas can be a bit abstract and possibly overkill given the potentially small effect an upgrade of pump may have.

Basically, it'd be nice to see what impact going from say, and Eheim 1048 to a 1250 pump would have on measured cpu temps in a typical watercooling system. This would be a great aid to watercooling newcomers so they'd instantly be able to see what kind of pump they'll be after for a new system of theirs.

Just my 2 cents, hope I'm not ruining the intention of this thread :)
That depends on the design of the block, so it's an individual question. In general, it helps, but your question ought to be: is it worth the expense? If it's an upgrade, no. If it's a first build or replacement of a failed pump, yes.
 

Mull

Member
Joined
Jan 20, 2002
Location
northeast ohio
Great Job!
I have designed cooling systems on stock cars using the same
principles but could never put it in a article as good as you have.

Mull
 

rogerdugans

Linux challenged Senior, not that it stops me...
Joined
Dec 28, 2001
Location
Corner of No and Where
BB2K:
Keeping me honest- I have more reading to do soon, but not today- (I haven't looked up Darcy's Law yet either, lol.)
Happy 4th of July!

I will read through all the info you linked and see if you have improved my understanding of turbulence. ;)

will_maltby-
As bigben2k said, it is yet another question that relies on many variables and the answer may not always be readily apparent.

I have read quite a few reports of people with higher flow pumps getting WORSE temps, and while I can jump to a conclusion(s) about what is happenning in such a situation, I would probably be wrong. :D

But this document is NOT intended to answer ALL flowrate and pump questions in any case (neither did the original thread.)
My hope was to simply provide (correct!) conclusions about the factors that are discussed most often and/or that have the most impact. And supply links to more in-depth info.

In other words, provide the basics necessary for greater understanding in order to achieve better performance from a home-brewed setup.
 
Last edited:

UnLoadeD

Member
Joined
Dec 1, 2002
Location
miwest, usa
rogerdugans said:
I did not go into turbulators, although I did run across them numerous times, maily because they hurt flow rate severely. Since the pumps used in water cooling are extremely sensitive to flow restrictions, use of turbulators in our application would most likely be counter productive: any gain in efficiency from turbulence would be lost (and more) by the loss of flow rate. So for simplicity, I did not mention them.

Not all turbulators are gonna be counterproductive. Have a look at Hoot's pin block. The pins introduce quite a bit of turbulence without hurting the flow. Also the dimples in Swiftie blocks serve the same purpose. I'm sure when you were reading about turbulators pins, fins, dimples and such didn't come to mind, but this is what they are.

peace.
unloaded
 

rogerdugans

Linux challenged Senior, not that it stops me...
Joined
Dec 28, 2001
Location
Corner of No and Where
Ok, an update about research on the comments made so far:

I am an idiot :D

lol, not really but I definitely do NOT have the education to understand the mathematical equations involved. And that is the name of that tune.
Unless someone can render that information into English ;) , my understanding will remain fairly limited.

BUT- I do think I see an error in the categories: I do not have anything on Water Blocks or Radiators.
The turbulators (being anything in the system that intentionally creates turbulent flow) in water cooling systems are concentrated in wbs and rads, which are also the biggest flow restrictions in most systems.
I do not mention them at all and few of my comments concerning flow rate considers them. Yet these are where the heat is actually transferred!

I am going to toss together a section for Blocks/Radiators and edit that in.

I think I am also going to revise the intro. to reflect two things:
1) This is a simplified explanation tailored to our purposes and does not intend to delve into all the details of fluid dynamics.
2) This post should help in figuring out how components work together: it is NOT a primer on how to design a waterblock, etc. although it does consider some aspects of block design and how it impacts flow rates.
 

rogerdugans

Linux challenged Senior, not that it stops me...
Joined
Dec 28, 2001
Location
Corner of No and Where
Okay- I am posting the changed sections here with notation as to where there go:

Between the first intro paragraph and the Heat Transfer section.
CHANGE
This post is not intended to be the “final word” on Flow Rates, nor do I plan on getting heavily involved with the mathematics of fluid dynamics- the equations are far beyond my abilities.
What this IS:
1) A simplified explanation of Flow Rate tailored to our purposes. Components will be taken into account and briefly discussed.
2) Understand a little more about how the components work together- what impact to expect with a change in tube or barb size, etcetera.
END CHANGE



A few changes here and the WB/Rad section is New ;)
CHANGE
Flow Type:

Water moving through a vessel (tube, block or radiator) meets resistance at the walls; this causes the water at the walls to move more slowly than the water in the center of the tube: this is Laminar Flow.
Laminar flow is bad for heat exchange because the water against the vessel’s walls is slower than the water in the center. Flow rate at the heat exchange surface has diminished.
This is where turbulence comes in - if we can get fluid from the center of the vessel to mix with fluid toward the walls, we end up with more efficient heat removal. Turbulence increases heat transfer significantly over slower Laminar flow.

Turbulent flow occurs naturally in a pipe when the fluid velocity exceeds a certain point, which is dependent on a lot of factors. Also, turbulence isn't an on/off thing - you can have more or less of it. Moving faster will result in more turbulence.
So, in short moving water through the block faster improves heat transfer between the block and the water, which reduces the temperature differential between the block and water required to move an amount of heat.
It is NOT intended to reduce the temperature increase in the water as it travels through the block, but rather to allow more heat to be removed.

Faster flow means more turbulence, and that is a good thing- and brings us to the next topic…..

Water Blocks and Radiators
Water blocks are one of the most restrictive components in a water cooling system- this is due in part to the relatively massive amounts of turbulence in these components caused by turbulators such as pins, dimples, changes of direction, fins and impingement jets.
A good radiator is in the same situation : by design they maximize surface area and induce turbulence.
These turbulence inducing components more than compensate for the restriction they impose by increasing heat transfer dramatically.

Please note that I said a GOOD radiator: the best radiators for use in water cooled computers have been found to be automotive heater cores. Many of the “radiators” sold by online vendors are of the bent-tube-and-fin type. These add flow restriction without the benefit
of effective surface area or sufficient turbulence.
END CHANGE

A question on flows myself:
Turbulent Flow is better for heat transfer- will laminar flow have less restriction? Not looking at the heat transfer part of the system- would laminar flow provide greater flow rate with the same pump?
 

bigben2k

Member
Joined
Jun 17, 2002
Location
Texas
Yes, turbulent flow is restrictive. This is where we get into a more complex explanation.

Like I wrote earlier, you can achieve turbulence in three ways:
#1: pump the fluid at a very high rate. This is not achievable with a normal pump.

#2: add turbulators. This is the most simple way to do it, with a minimal flow restriction. It's also where we get into fluid dynamics, and tuning of the turbulators, to maximize the effect of the turbulence. Very tricky, very hard to do (ref: CFD, Computational Fluid Dynamics), and infinitely difficult to tune, if you're a designer. I'm pretty sure that heatercore designs weren't just a fluke.

#3: jet inpingement. This is by far the easiest way to do it. It maximizes the pumping power into a small concentrated area. It is fairly restrictive, but only because of the pre-jet tubing required. On a seperate note, and mostly irrelevant to most, I found out that it's also referred to as an "irrecoverable pressure drop", because you can't recover the flow speed to the outlet. But I digress... ;)


You ought to take a look at Hayden coolers: it looks deceptively simple, with a simple U shaped tube, and fins. What they did is include turbulators inside the tube, making it most efficient. You'd have to dig up one of Bill's threads, to see the results. If you're interested, I can post a link to a ProCooling thread about it.
 

rogerdugans

Linux challenged Senior, not that it stops me...
Joined
Dec 28, 2001
Location
Corner of No and Where
Okay, I am going to add a few more things (later though):

1) A mention that laminar flow is preferred in the tubing to keep flow rates up, but turbulent flow is better in blocks and radiators for Heat Exchange.

2) bigben2k's list on turbulence inducers-
#1: pump the fluid at a very high rate. This is not achievable with a normal pump.

#2: add turbulators. This is the most simple way to do it, with a minimal flow restriction. It's also where we get into fluid dynamics, and tuning of the turbulators, to maximize the effect of the turbulence. Very tricky, very hard to do (ref: CFD, Computational Fluid Dynamics), and infinitely difficult to tune, if you're a designer. I'm pretty sure that heatercore designs weren't just a fluke.

#3: jet impingement. This is by far the easiest way to do it. It maximizes the pumping power into a small concentrated area. It is fairly restrictive, but only because of the pre-jet tubing required. On a seperate note, and mostly irrelevant to most, I found out that it's also referred to as an "irrecoverable pressure drop", because you can't recover the flow speed to the outlet.
 

bigben2k

Member
Joined
Jun 17, 2002
Location
Texas
rogerdugans said:
1) A mention that laminar flow is preferred in the tubing to keep flow rates up, but turbulent flow is better in blocks and radiators for Heat Exchange.
Right, but it's not so much about turbulence, as it's about pressure drop.

A good rule of thumb is to keep the flow speed in the tubes at about 3 fps (feet per second).

That's why I've had to switch to 3/4" tubing ;)
 

cpufan

Member
Joined
Jun 17, 2003
Location
Iowa City, IA USA
bigben2k said:

...Right, but it's not so much about turbulence, as it's about pressure drop.

Bigben2k has it right, of course. Keep in mind that the frictional loss (pressure drop) along the length of the tubing is proportional to the fluid velocity squared, and the fluid velocity is inversly proportional to the radius of the tubing squared. So, ultimately the friction factor is inversely proportional to the diameter to the fifth power. I hope I got that right :)

What does it all mean with respect to cooling?

*Use large-diameter tubing to minimize the frictional loss along the length of the tubing. Of course very large tubing is unweildy, so we must settle on a reasonable size.

*3/8' tubing will result in approximately 4x the frictional loss of 1/2" tubing.:eek:

*High velocity and the resulting turbulent flow is desirable in the components in which we want heat transfer, but low velocity is desirable in the rest of the system, where the goal is to reduce frictional (and local, or minor) losses.

Great job on the post Roger. Please feel fre to ask questions if you do not understand what I am saying, or if there are flaws in my reasoning.

Jeff
 

rogerdugans

Linux challenged Senior, not that it stops me...
Joined
Dec 28, 2001
Location
Corner of No and Where
Okay folks - I'm back at it again! I have committed the last few changes and I think this is done.......
For ease of reading I am going to repost the entire thing below (and add some colors too .)

Does more water flow = better cooling?


I wrote this in an attempt to reduce a large amount of data (which was largely in one very informative thread which was 17 pages long) to one simple and fairly easy to read document.
The original thread is HERE for anyone who wishes to get more in-depth knowledge on this subject.


This post is not intended to be the “final word” on Flow Rates, nor do I plan on getting heavily involved with the mathematics of fluid dynamics- the equations are far beyond my abilities.
What this IS:
1) A simplified explanation of Flow Rate tailored to our purposes. Components will be taken into account and briefly discussed.
2) Understand a little more about how the components work together- what impact to expect with a change in tube or barb size, etcetera.

Heat Transfer

Heat transfer is the basis for ALL computer cooling systems; in water cooled computers we make this more complicated by using multiple heat transfers:
cpu core to water block
water block to water
water to radiator
radiator to air.

* Heat transfer works best with the biggest temp differential: ideal would be cold water to cpu and hot water in the radiator. We cannot achieve this because a closed loop will achieve equilibrium at some point. The principle holds true however- the most efficient transfer happens at the greatest temperature differential, therefore higher flow rates will always help with all other variables remaining the same.
*It is the Heat Transfer that we want to maintain as efficiently as possible, and that is best done with a higher flow rate. Rather than thinking that there won't be enough time for heat to move towards the cool water, and therefore compromising heat loss, it is better to think that there is more fresh water moving onto the CPU and therefore, there is increased cooling.

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.

This is true even though a system with a lower flow rate will have more time to heat the water in the block and also more time in the radiator to cool the water: since the heat exchange works best with the greatest temperature differential, longer “stay time” is counter-productive.

Sources of Confusion:

There are some variables that have made this more confusing in practice though:
*Pump Heat
*Type of Flow: Turbulent or Laminar
*Friction
*Component Flow Resistance
*Pump Design

I will attempt to end the confusion on these points next.

Pump Heat:

Pumps generate heat; rather than explain why, let it suffice to say that if you put your hand on a running pump it WILL be warmer than one that is not running.
This heat has to go somewhere: a submerged pump (inside of a reservoir) must add all of the heat it generates to the water; inline pumps are usually designed to use the pumped fluid as a coolant, so most of the heat is going into the water. There will be some amount of heat being conducted to the outer surface of an inline pump, but this should be considered a fairly small amount of the heat produced.

Now, in a simple water cooling system (cpu water block only) we actually have two sources of heat: the cpu and the pump.
Pumps with a higher Flow Rate will generate more heat than pumps with a lesser Flow Rate, and there lies our first bit of confusion:
It is possible to add more heat from a larger pump than will be removed by the higher Flow Rate.

Flow Type:

For our purposes I am dividing flow type into two categories- there are more, but since I have no degree in Fluid Dynamics two is enough…
High velocity and the resulting turbulent flow is desirable in the components in which we want heat transfer, but low velocity is desirable in the rest of the system, where the goal is to reduce frictional (and local, or minor) losses.

*Laminar Flow*

Water moving through a vessel (tube, block or radiator) meets resistance at the walls; this causes the water at the walls to move more slowly than the water in the center of the tube: this is Laminar Flow.
Laminar flow is bad for heat exchange because the water against the vessel’s walls is slower than the water in the center. Flow rate at the heat exchange surface has diminished.; however: it is better for the tubing in our system because it has less frictional loss (or pressure drop) than turbulent flow.

A couple of pointers to help pick tubing size…..
1. Use large-diameter tubing to minimize the frictional loss along the length of the tubing. Of course very large tubing is unwieldy, so we must settle on a reasonable size.
2. 3/8' tubing will result in approximately 4x the frictional loss of 1/2" tubing.

*Turbulent Flow*

If we can get fluid from the center of the vessel to mix with fluid toward the walls, we end up with more efficient heat removal. Turbulence increases heat transfer significantly over slower Laminar flow.

Turbulent flow occurs naturally in a pipe when the fluid velocity exceeds a certain point, which is dependent on a lot of factors. Also, turbulence isn't an on/off thing - you can have more or less of it. Moving faster will result in more turbulence.
1. Pump the fluid at a very high rate. This is not achievable with a normal pump.
2. Add turbulators. This is the simplest way to do it, with a minimal flow restriction. It's also where we get into fluid dynamics, and tuning of the turbulators, to maximize the effect of the turbulence. Very tricky, very hard to do (ref: CFD, Computational Fluid Dynamics), and infinitely difficult to tune, if you're a designer. I'm pretty sure that heater core designs weren't just a fluke.
3. Jet impingement. This is by far the easiest way to do it. It maximizes the pumping power into a small concentrated area. It is fairly restrictive, but only because of the pre-jet tubing required. On a separate note, and mostly irrelevant to most, I found out that it's also referred to as an "irrecoverable pressure drop", because you can't recover the flow speed to the outlet.

So, in short moving water through the block faster improves heat transfer between the block and the water, which reduces the temperature differential between the block and water required to move an amount of heat.
It is NOT intended to reduce the temperature increase in the water as it travels through the block, but rather to allow more heat to be removed.

Faster flow means more turbulence, and that is a good thing- and brings us to the next topic…..

Water Blocks and Radiators
Water blocks are one of the most restrictive components in a water cooling system- this is due in part to the relatively massive amounts of turbulence in these components caused by turbulators such as pins, dimples, changes of direction, fins and impingement jets.
A good radiator is in the same situation : by design they maximize surface area and induce turbulence.
These turbulence inducing components more than compensate for the restriction they impose by increasing heat transfer dramatically.

Please note that I said a GOOD radiator: the best radiators for use in water cooled computers have been found to be automotive heater cores. Many of the “radiators” sold by online vendors are of the bent-tube-and-fin type. These add flow restriction without the benefit
of effective surface area or sufficient turbulence.

Friction:

I am not going to say much on friction here: it generates heat in the pump, and it also generates an amount of heat as water flows through system components (a system with greater Flow Rate or Head, either Friction or Static, will produce more heat.)
Perhaps the main area friction is involved in a water cooling system is in flow resistance- the next area to be covered.

Component Flow Resistance:

Pumping a liquid through a tube creates resistance. The resistance is determined by the cross section of the tube, the length and all fittings in the line.

Static Head (or Lift) - number of feet of elevation that the pump must lift the fluid regardless of flow rate.
.
Friction Head- measure of resistance to flow (backpressure) provided by the pipe and its associated valves, elbows and other system elements:
A smaller tube diameter will have greater resistance: even with identical fittings, pumps and water blocks, a system with larger diameter tubing will have a higher flow rate.
A longer tube will also have greater resistance: even with identical fittings, pumps and water blocks, a system with shorter tubing lengths will have a higher flow rate.
A straight length of tube will have less resistance to flow than one that is bent. A partially kinked tube easily proves this point. Any bend at all introduces some restriction to the flow: a sharper bend is more restrictive than a gradual bend.

Head- the entire amount of flow resistance in a system. Static Head + Friction Head = Head
This is what pump head capacity must overcome and is entirely responsible for the reduction of flow rate in a system.
As an example:
*3/8' tubing will result in approximately 4x the frictional loss of 1/2" tubing.

Pump Design:
Positive displacement pumps will maintain constant flowrate but increase pressure as line restrictions interfere.
Most pumps used in water cooling are centrifugal pumps and these are NOT positive displacement pumps.
Getting the pump with the highest flow rating is NOT necessarily the best answer: centrifugal pumps tend to be extremely sensitive to flow restriction.
A pump with a higher Head Capacity will be less sensitive to restriction and be more suitable for computer use.
Which brings us back to the issue of pump heat ;): a pump with more head capacity and higher flow rate will add more heat to the system.

A Bit about Fans:

Just as higher flow rates remove heat from the cpu faster, greater air flow rates through the radiator will improve performance at any given temperature. The actual equations differ since the fluid characteristics- water and air- differ, but the same principles apply.
The advantage to more airflow is that it provides a good “bang for the buck” improvement in performance; the disadvantage to this is that a fan will create more noise than a water pump, and noise is often one of the areas we are trying to improve with water cooling

Conclusion:

A higher flow rate will give lower temperatures as long as NO other variable is changed.
Heat exchange is improved with turbulence- higher flow rate THROUGH A GIVEN DIAMETER TUBE will flow faster and be more turbulent.
Flow rate can be increased by using larger diameter tubing, the shortest total length of tube possible, fewest bends and fittings possible, lowest restriction radiator possible and by using a pump with a higher head capacity and flow rate.
Maintaining good airflow through the radiator is probably the easiest and noisiest way to improve performance.

Choosing an appropriate pump may be the hardest part: a high head capacity is best, high flow rating important but less so; close attention MUST be paid to the amount of heat generated by a pump as this heat will be added to the system.

Formulas:

Q=M x c x Delta T.
There is an elementary equation from basic thermodynamics that states that the rate of heat transfer (Q) equals the mass flow rate (M) times a constant (the specific heat of water) times the delta T (fluid temp out minus fluid temp in).

Links to source information:

Old Flow Rate Sticky
www.pump.net
Google



Much of this information is from the original Flow Rate Sticky and credit is due the original authors.
Thank you for your contributions.

Any inaccuracies are mine ;)
Please PM me with any errors you see and include source information.

That, I think, is that.
A few days for comments and then I'll see how I can fit it in as a sticky.....or something.
 
Last edited:

Prandtl

Member
Joined
Jan 9, 2003
Location
Montreal, Canada
cpufan said:


*3/8' tubing will result in approximately 4x the frictional loss of 1/2" tubing.:eek:

Althought that might seems a lot, keep in mind that 4 times something negligeable (in regards to component head loss) might still be negligeable.

rogerdugans said:

Copper Tubing Flow Resistance by Linear Foot (loss expressed in psi) Conversion: closest I have come to is 1 foot = 1 psi.

This is WAY off!! My maxiJet has a max head of 4 psi, so thats mean that 4 feet of copper tubing will shut it off!?! Also, head loss in tubing is a factor of diameter and fluid velocity (or flow rate), but your dont give any of these two parameter along with your very high linear loss! As far as I remember (from computation i made in fluid dynamic class), unless you use alot of tubing, you can consider the loss in your tubing negligeable in regards to the loss in elbow, hard tubing bend (wich can be consider elbow) and block. In short, frictional loss (tube) is negligeable in regards to minor loss (every other source of loss but the elevation).

EDIT: I'll make some computation when i get home to prove my point further and give better explanation.
 

cpufan

Member
Joined
Jun 17, 2003
Location
Iowa City, IA USA
Prandtl said:

Althought that might seems a lot, keep in mind that 4 times something negligeable (in regards to component head loss) might still be negligeable.

Agreed. I have not calculated the tubing losses and compared them to that of the radiator and waterblock, which would have to be measured. I believe that several people here have done empirical flow-rate tests with the two popular tubing sizes, and concluded that there is a significant difference. I can't seem to find the info in a search, however.

Prandtl said:

This is WAY off!! My maxiJet has a max head of 4 psi, so thats mean that 4 feet of copper tubing will shut it off!?! ...

Yes, this is a puzzleing number. He must be refering to some sort of "rule of thumb" design calc., and i assume the 1' = 1psi should read something more like 100' = 1psi. Of course, without flow rates or pipe sizes, I am only guessing.

Roger- this section with the copper tubing doesn't really lend much to the understanding of the subject. Short of introducing the Darcy-Weisbach Friction Factor, the best thing to do might just be to point out that frictional loss in tubing is proportional to the velocity squared, or better yet, just that smaller-diameter tubing has a higher resistance to flow (which you already point out in a couple of places). You probably don't want to get into the D-W equation, since things would get complicated quickly, but not really add much to the discussion.

Prandtl said:

In short, frictional loss (tube) is negligeable in regards to minor loss (every other source of loss but the elevation).

You are likely correct. Of course, part of the minor loss is related to tubing diameter as well- e.g. entrance and exit losses of the res, rad, and wb- since such losses are also proportional to the fluid velocity.


Prandtl- Have you chosen your screen name to amuse us, or is it just coincidence that your namesake studied fluid dynamics, boundary layer effects, and heat transfer?
 

bigben2k

Member
Joined
Jun 17, 2002
Location
Texas
Pressure drop figures for standard copper piping aren't hard to find.

You need to add Darcy to the Hazen-Williams formulae: H-W was meant for turbulent flow. I've never compared the two, and the definition of the term "turbulent" gets a little hazy, when Reynolds is under 2'000.

For 1/2" ID tubing, using H-W (!), a friction factor of 130, and given a 1 gpm (60 gph) flow rate, you'll have a pressure drop of 0.44 inches. In other words, you loose 3.7 feet of pressure, over a 100 foot long tube.

1/2" Copper tubing isn't quite 1/2", so it's best to take industry standard figures.