Flowrate...

[kusojiji]

My mistake about swiftech.com. It's actually Swiftech @ swiftnets.com.

Yeah, I'd agree with you that manufacturer's findings are suspect, but that test was done at this site: http://forums.procooling.com/vbb/sh...rpage=25&highlight=pump and test&pagenumber=1

There was a good number (subjective ) of people in on that test: Phaestus, BillA, and others.

I see the error that you made note of on swiftech's website. Vewy stwange!

But I disagree with you on the second statement. All waterblocks benefit from higher flow. Thermal-management-testing.com and overclockers.com websites show how the waterblocks perform under a controlled environment. I'm not going to go into their testing procedures because it has no pictures.
But here's the website: http://thermal-management-testing.com/methodology.htm
I made no such assumptions about water temps.

All results show that more flow = better waterblock performance, up to a certain point, where the graph flattens out and any increase in flow results in a miniscule increase in performance. That's where the decision on pump performance versus price comes in.

Yes, in a home system, where temps would vary from the test temps, results would sway in either direction. But, if a side by side test were done under the same conditions, higher flow would work better than lower flow.

Off the subject, if flow is a problem, such as with a white water waterblock, where backpressure is high, then I'd go with a bigger radiator that is free flowing in both air flow and coolant flow. That'll help the pump keep what flow rate it can make and the bigger size will allow for more surface contact. Luckily (well designed, actually) the white water block works very well with low flow. I've noticed a lot of people going to that block, recently.

 

[Graystar]

But I disagree with you on the second statement. All waterblocks benefit from higher flow.

I never said they didn't. In fact, I agreed. What I am saying is that you cannot take a property of one component, apply it to the entire system, and expect the system to behave as the component did. There are interactions that need to be accounted for.

There is one thing for sure. If you've got some watts to burn on cooling, then burning watts on a fan, mounted on a second radiator, is going to provide far more benefit than burning those watts in a bigger pump.

 

[JPSJPS]

JPSJPS, I believe it is you and the others who are missing the overall picture. The examples I’ve seen posted here show a clear misunderstanding of heat energy and heat transfer, as well as an across-the-board disregard for how the components involved affect the properties of the fluid.

BillA is considered by most on this board to be the most knowledgeable when it comes to the thermodynamics of water cooling. He says that a slower flow through the radiator is better.

I will try again!
Liquid cooling theory and design has been analyzed and developed for over 100 years by scientists like Reynolds, Nusselt, Prandtl, Grashof. Today sophisticated CAD programs based upon that and other research are available for professional liquid cooling system design and analysis.
Huge multi million/billion $ liquid cooling systems like steam power plants (both nuclear and fossil) have been designed based upon this knowledge. Liquid cooling is used extensively in many manufacturing processes, electrical/electronic/mechanical machinery and the obvious things like cooling internal combustion engines. This is a mature technology that is well documented and understood by professionals involved in that field.

But a professional cooling system for a PC would cost many times over the cost of the PC. So, clever experimenters have designed and built PC liquid cooling systems on the cheap. A lot of data has been taken to try to characterize the individual pieces like water blocks and radiators. Because expensive sophisticated test equipment and test procedures are generally required to obtain extremely accurate data, some of this data is obviously faulted. This is no big deal unless you draw conclusion from this faulted data that contradicts known science. Then these ridiculous arguments start.

A few educated knowledgeable posters here obviously understand the basic theory. But for each of them, there are many posters that want to argue their opinions or "gut feel" that contradict that proven theory.

These knowledgeable posters have already many times stated a basic cooling system fact:
Higher flow rate produces lower CPU temperatures because this results in a more uniform liquid temperature through the system!
Caveats:
1) Obviously increasing flow rate will eventually provide diminishing returns that are not worth the increased cost.
2) Most or all of the energy applied to power the pump will be transferred to the water. When this added energy is higher than that removed by the higher flow rate, temps will increase.
3) Higher flow rates can cause secondary affects like cavitation which can be a real problem in industrial cooling systems.

The best non-technical explanation that I can provide is this:
High flow rates and uniform liquid temps mean minimal liquid temp increase through the water block (as well as minimal temp decrease through the radiator).
Stated conversely, low flow rates will mean a significant liquid temperature drop through the radiator. This then means that we will have an corresponding *UNDESIRABLE LIQUID TEMP RISE* through the water block and the CPU will run hotter.

If that explanation does not make sense to you and you have enough stamina, a Google search using terms like "Reynolds Number" liquid cooling" "heat transfer" will bring you more information than you will ever want read. Also, vendor websites that design or sell liquid cooling parts and systems or CAD software will provide some good stuff.

For these pressures we can assume that water is incompressible which will greatly simplify the analysis.

In fact, a study the Reynolds Number and how it effects laminar flow & boundary layer vs. turbulent flow will provide the answers for why high flow is desirable in a water block with its small water to metal heat transfer surface area.

A lot of the arguments/discussions here already have valid proven answers. Often a little technical research/study will eliminate some of the beating a dead horse discussions here.

Anyhow, thanks to the folks that post their test results. Even if they are not perfect, plenty of info is available to build a good working system.

[/B][/QUOTE]

 

[Graystar]

I will try again!

You can try as often as you like. Unless you can mathematically model a PC water-cooling system and demonstrate how changing flow rate affects the system, I'm not simply going to accept what you say. As far as I know, there are no pump designers, or radiator designers, or water-cooling designers posting in this forum. So to me, no one here is knowledgeable on the subject.

The best non-technical explanation that I can provide is this:
High flow rates and uniform liquid temps mean minimal liquid temp increase through the water block (as well as minimal temp decrease through the radiator).

That’s the best? A minimal temperature change means nothing because the starting temperature could be higher. That is a fact that you simply can’t seem to acknowledge. You also accept, without question, that a radiator will dissipate more heat at a higher flow rate. However, BillA’s testing shows that this is not always true. Sometimes the heat dissipation capacity can go down. How you can expect such a radiator to transfer more heat, when testing shows that it can’t, is beyond me.

 

[rogerdugans]

I have no degrees.
I have no highly accurate test equipment.
I do have a number of water cooled computers and have used a number of pumps, water blocks, types of tubing, radiators and tube routing schemes, as well as a few forays into refrigeration systems and other methods of further enhancing a water cooled pc.

I have yet to see any combination that did NOT provide better results with a higher flow pump.
Can I explain this mathematically?
No. I can balance my checkbook, but not much beyond that with math I'm afraid.

I do agree that there is definitely a matter of diminishing returns involved as flow rate goes up, and there is no question that a system must be looked at as a whole: the best water block in the world will be in a badly performing system if the radiator consists of a six inch copper tube with no fins or fans.

Stronger fans/more airflow over the radiator is usually the most effective way to improve a given systems performance, but it usually comes at the cost of increased noise levels.
Many are not willing to endure that.
A higher flow pump may not get as good a return on the financial investment but it is quieter.

Perhaps I have just never had a badly executed radiator in my possesion and this is why I am convinced that higher flow works.
But even a radiator made with about 6 feet of 1/4" ID copper coil with fins (from an old water cooled weldeing machine) performed better with a higher flow pump. And that is a notoriously bad design for a radiator.

While I believe I may have a grasp on the points made here (and elsewhere) about flow rate, I also know what has worked for me in many trials and tests I have done.
The two don't agree- I do NOT fault your logic or math (and I may need help with the checkbook after Christmas.....), but I do think that there must be something that is missing from these discussions that will make the math and the experience agree.
What it is I haven't a clue.

 

[Graystar]

Well I just tested my system again and I get the same results...no change when I increase flow.

I happen to have a 12v pump that I run at 7v. Running at 7v has the effect of cutting the flow rate in half, even though it's not half the voltage. Don't ask why...I don't know. All I know is what I've measured in the past. Anyway, simply by moving a wire on the Molex connector, I can set the voltage back to 12v. So it is very easy for me to test between two flow rates.

I get the same exact temperature readings with either flow rate.

To me that just means I don't have a great radiator setup. I have a single 120 fan, no shroud and I run the fan at 5v. And mind you, this is a dual 2200+ processor setup. Maybe if I boosted the voltage up to 12v, thereby increasing the cooling capacity, I may then have the spare cooling capacity required to see a temp drop with increased flow rate. However, the fan is too loud at 12v, which is why I have it at 5v. I like things quiet.

So I think the most useful thing we can say about flowrate is this: If you have a flow rate of around 1gpm, invest the watts you want to spend in extra air cooling. 8 watts of air will do much more than 20 watts worth of increased flow could ever do.

 

[JPSJPS]

Graystar:
I suggested that if my explanation did not make sense that some time spent chasing Google hits would be educational and beneficial to you. You did not do this but instead want to continue to argue your baseless notions. If you would Google and study, you will learn the following:
1) Liquid cooling systems have been mathematically modeled by professionals and scientists for at least 50+ years and the affects of flow rate are well known. These are facts are based upon mathematics and physics rather than opinion, guesses, etc.
2) Same answer as 1) for my acceptance of flow rate vs radiator dissipation.
3) Starting temps are irrelevant - Final steady state conditions are all that matter. That is what I am describing.
4) I am not going to get into BillA and his data. He has done a lot of good work and the reasons for his minor test data mistakes are not important.
5) You have a long way to go before any mathematical explanation would make sense to you.

rogerdugans:
The only mistake you made is your last statement. In fact, the math and theory agree exactly with your experience!

 

[Graystar]

I suggested that if my explanation did not make sense that some time spent chasing Google hits would be educational and beneficial to you.

I already did that nearly a year ago, the last time this thread flared up. That's why I believe as I do.

4) I am not going to get into BillA and his data. He has done a lot of good work and the reasons for his minor test data mistakes are not important.

Oh I see...if the test data doesn't agree with you, then it's wrong.

Interesting...

 

[rogerdugans]

Here is one thing that is missing from all the actual real-world testing that has been mentioned, myself included:
no two systems are exactly alike.

This, I realize, invalidates my results to at least some degree: I do not have an endless supply of hardware or money to purchase same.

In various systems I have seen seemingly minor changes make huge performance differences and vice versa.

There are many factors that must be taken into account when assembling a water cooling system, and for most of us I think that there is a lot of guesswork involved, even if it is often (sometimes Highly) educated guesswork.
And the system can be more than the sum of its parts, I think, or less, depending on how good your guesswork and experience at this is.

And a gentle reminder to all to please avoid flaming other members in the forums- I don't think we are there yet, but I do see the potential for it to happen.
Please don't go there.

 

[JPSJPS]

That's why I believe as I do.
What you believe is only important to you! Beliefs are meaningless. The facts/conclusions based upon technology are what is important.

Oh I see...if the test data doesn't agree with you, then it's wrong. Interesting...
No the test data does not agree with well established and proven technology. I did not develop any of this technology but only learned it from experts that came before me. Your inane argument is not with me; You are arguing with all these experts.

There are several ways to go through this world.
An ignorant uneducated person sometimes believes his opinion based upon nothing more than nonsense is very important. He will argue this until he is blue in the face.
An intelligent person will try to learn from the experienced experts so that he has a knowledge based upon facts. This learning requires work.

The choice is yours.

I planned to offer a semi technical explanation (as others have already done on this thread) but you argued with a very simple explanation.

I give up!

Have a nice day!

 

[Graystar]

An intelligent person will try to learn from the experienced experts so that he has a knowledge based upon facts. This learning requires work.

I'd be happy to learn from experts. Please let me know if any ever start posting here. Until then, I'll follow my research on the subject.

 

[kusojiji]

The information I was quoting from are from people I consider to be experts. I was not making any assumptions based on my information and I backed it up with data coming from these experts. While I don't consider myself to be anywhere near an expert on this subject, I do know how to read and extrapolate the data and make it work. My own results proved it to work.

As for BillA's findings, I agree that he has done a lot of good work and will continue to do so, and I'm not busting on him. But if someone is going to put up some data that others are going to use and especially if that person is recognized as the technical expert in this field, then that information has to be correct. Anything other than that would be unprofessional. Swiftech uses him as a technical advisor and uses his statements to promote their pump.

The radiator tests were done a few years ago and I wouldn't expect him to go through and rerun the tests, but correcting the writeup and the final graph is not that difficult. Or pulling the article completely instead of having his name on misleading information would be better.

 

[I.M.O.G.]

I wish a bomb would be dropped on this thread. It is the worst thread on the entire forum IMO. It is so full of misinformation.

Graystar, nice to see you around again. I will concede any victory to you on the matter of the specific example I used because it simply isn't worth rehashing. I haven't even had any time to be on the forums at all lately let alone argue some marginal example of whats going on in a futile attempt at re-explaining the same old thing in this same old thread with so much mixed information. This is my first stop into OCF in a couple weeks atleast I think.

The flaws you have highlighted are slight misinterpretations of the points the example makes - but that is my fault for not taking the time to explain it more thoroughly, it's not yours. It's not the clearest example and it just shouldn't be used any further - it was just meant to explain something to AngryAlpaca and it finally worked at communicateing the point that was trying to be passed along. It is a poor example when applied in general but worked for it's specific purpose. So no one should waste their time examining that example, its not worth your time.

Regarding Bill Adams on this matter - his own tests available through www.thermal-management-testing.com show that higher flow often improves radiator heat dissipation by a small but easily noticeable margin. Go to his site and click on the "corrected" heat dissipation graphs.

It may be wise to be careful which comments of his one quotes on this matter - some of his statements that are out there on this subject are based upon the erroneous original results and conflict with the evidence portrayed in the updated results.

That's the problem when someone bases one's opinions on direct experimental applications - flaws in the experiment can lead many of the opinions one has formulated to be wrong. That's why I prefer looking through to the governing dynamics that produce the results, and explaining things from a thorough understanding of these.

Obviously I haven't taken the time to dig through all of the responses since I have participated in this thread, but I noticed you were around and I wanted to atleast respond to where you directly addressed me. See ya'll around the forums. If I'm not back sooner, I'll be around the week of January 12th.

 

[Graystar]

Hi IMOG! Good to see you too!

Hey, you know, we're a couple of smart guys, and there are a lot of other smart people here. We must be able to create a simple model of this and work out the math to see what's up. Then there simply can't be any question (well..no question as to the model, at least! ) It would be nice to get something like that in a new sticky and get rid of these two that are here. THEN we can argue about how the model relates to a waterblock and rad!

Here's what I was thinking...

for the model we could assume two square pieces of copper, of some thickness. One would have 100 watts of heat energy applied to the underside. The other would have a constant temperature on its underside. On top we have a sheet of water that flows from one copper block to the other. Pretty simple.

From this model we should be able to calculate the various temperature deltas, transfer rates, and fluid temperature.

Make sense so far?

 

[JPSJPS]

Hi IMOG! ... THEN we can argue about how the model relates to a water block and rad!

This is a technology that is mature and the theory has been well known and documented for something like 50 years. The models are well known but understanding them and how they affect the system analysis requires a fairly extensive education in physics and math. In this scenario, technical folks generally do not "argue", but instead "discuss". The discussion will involve the principles of thermodynamics (heat transfer), the mathematics behind this, and how it all applies to the subject at hand.

It is clear that IMOG has at least has a good basic theoretical understanding of this subject. (Maybe much more)
It is equally clear that you DO NOT!

He first pointed out that this thread is full of misinformation!
Skimming the thread, I see that this misinformation and the bickering about is about 99% of the thread. The few posts from knowledgeable folks get lost in the rubble of posters that believe that their opinions based upon nothing are very important. Do I have to name one of these rubble posters?

IMOG is a VERY nice guy (he chewed me out for one of my earlier blunt posts so I edited it).
He first points out in a nice way that it is not worth his time arguing with an ankle biter that misinterprets his posts and wants to argue. Do I have to name that ankle biter?

He next points out that BillA took data and drew conclusions from that data. Then, he later took data that contradicted the first and drew later conclusions that also contradicted the first.
IMOG's following quote tells it like it is:

Originally posted by IMOG That's the problem when someone bases one's opinions on direct experimental applications - flaws in the experiment can lead many of the opinions one has formulated to be wrong. That's why I prefer looking through to the governing dynamics that produce the results, and explaining things from a thorough understanding of these.

To a technical person it is obvious that BillA knows as little about the governing dynamics (theory) as you. If he had possessed even a "little bit" of knowledge, he would have known that his first data was faulted! He would have researched his measurement procedure to determine a way to get valid data instead of posting the bad data and the bad conclusions he drew from it. Instead of listening to the experts that critiqued his faulted results, he argued with them and made a fool of himself.
When you quote BillA, it is the case of the blind leading the blind.

Originally posted by Graystar
Here's what I was thinking...

for the model we could assume two square pieces of copper, of some thickness. One would have 100 watts of heat energy applied to the underside. The other would have a constant temperature on its underside. On top we have a sheet of water that flows from one copper block to the other. Pretty simple.

From this model we should be able to calculate the various temperature deltas, transfer rates, and fluid temperature.

Make sense so far?

Not really - This is NOT a model, but instead is a drawing. And, even the drawing is not quite there. Following is a way over simplified model:
1) I assume that you correctly put the black line to show circular flow in both directions to form a closed system. Good
2) The rest of the drawing needs some help.
The left block can be described as a water block with a known measured thermal coefficient vs flow rate.
3) Then we need to calculate the thermal resistance (heat carrying capacity) vs flow rate of the water path.
4) The right block will NOT be at a constant temperature.
Instead, it will be the radiator with a known measured thermal coefficient vs flow rate for both water and air variations.
So, when we have known individual transfer coefficient models, the closed loop calculations can be performed.
As you can see, this is NOT a simple process if we have both water and air flow rates variable!!!!!

As I posted in the other thread, we can simplify the analysis at a very high water flow rate. The thermal resistance of the water block and radiator will flatten out to a near final value. The water temperature differential throughout the system will be small compared to the CPU vs Air temp difference. Then, we can simplify the model by connecting the two blocks together since the water conduction is nearly perfect.

Of course, we need to measure the water temperature differential to verify that we are at this point. Then, the overall thermal resistance will be the two component individual resistances in series.
Then, the analysis results will be obvious and we will easily know the "weakest link".

It is late at night so if I screwed up, let me know.

 

[ThePimpulator]

Thank you

Recognizing the contribution particularly of RohXS, I e-mailed the link to this thread to myself on Thu, 1 Aug 2002 02:41:32 -0700 (PDT), knowing that one day, when finally getting the opportunity to build a WC rig, I would read it through and find it very useful. Well, today, Monday January 12, 2004 was that day. I just read the entire thing in one sitting. I must have been bored.

I would just like to thank RohXS, jimmytp, Owenator, IMOG, and Mithril, as well as others, for their excellent, excellent information. I now feel equipped to go about purchasing the components needed for my setup.

I would also like to thank GreenmanWD-40, AngryAlpaca, and particularly Graystar, as well as others, for providing the entertainment needed for getting through these 13 heated pages, even if I can't help thinking that if they had kept quiet the job would have been far less time consuming. Still, free speech and all that.

As for BillA, your efforts to remove your posts due to your flawed test results prolonging the arguing are much appreciated. However, you have a couple of posts remaining on page 8. It would be great if you could remove them

Anyway, I would just like to say one last thing: there were a few calls for this thread to be locked. That’s one thing. But please don’t remove it or its “sticky” status. I for one found it very useful, as will anyone else that is capable of understanding, and sifting through the BS, but has simply never had a need to learn this material until now. It will certainly save much Googling, much heartache, and much money in the long run.

Thanks again.

ThePimpulator.

 

[JPSJPS]

ThePimpulator - Thanks for one of the best posts on this thread!

 

[Graystar]

LOL! I think ThePimpulator is JSPJSP under another name, as both are equally confused.

If you want to believe RhoXS go right ahead. But the errors in his posts are clear. For example, lets talk about pump heat.

He would have you believe that all the wattage of a pump goes into the water. This is incorrect.

The change in temperature that water experiences while moving through the pump is calculated with the following formula:

Fahrenheit rise = ((BHP – WPH) x 42.41) / (lbs/min x Specific Heat)

Where:
BHP = Brake Horse Power (you get his number from the pump curve supplied by the manufacturer)
WHP = Water Horse Power = (GPM x 8.33 x SG x Head (in feet)) / 33,000
42.41 = Conversion of HP to Btu./min.
lbs./ min. = Gpm. x 8.33 x Specific Gravity
S.H. = Specific Heat (1 for water)

Once you have the Fahrenheit rise you can calculate the wattage required to cause such a rise:

Centigrade = Fahrenheit rise / 1.8
Milliliters/min = GPM x 3,785.4118
Calories/min = milliliters/min X Centigrade rise
Watts = calories/min x 0.06978

Watts of energy input into the water = (GPM x 3,785.4118) X (F / 1.8) x 0.06978


Lets look at the Custom Sealife Velocity T1 pump. This pump consumes 98 watts, pumps 546 GPH at 0, and has a max Head of 25 ft. From the P-Q chart we find that this pump will move 5 GPM at 20ft of head. How much heat will this pump put into the water?

With a BHP of 1/20HP and an efficiency of 50%:

Fahrenheit rise = ((.05 – ((5 x 8.33 x 1 x 20) / 33,000))) x 42.41) / (5 x 8.33 x 1)
_________________((.05 - (833 / 33,000)) x42.41) / 41.65
_________________((.05 - 0.025242424)) x 42.41 / 41.65
______________________0.02475757 x 42.41 / 41.65
___________________________1.049968 / 41.65

Fahrenheit rise = 0.0252 degrees

Now to calculate the power required to cause such a change

Watts of energy input into the water = (5 x 3,785.4118) X (0.0252 / 1.8) x 0.06978
__________________________________18927.059 x 0.014 x 0.06978

Watts of energy input into the water = 18.497

So we see that even a very powerful pump, working in its inefficient range, will only add about 18.5 watts of heat energy to the water. That is a far cry from 98 watts. I would imagine that an Iwaki 20 series will add about 10 watts. Something like a Danner Mag 3 will add about 7 watts.

The pumps we use are filled with epoxy. The thermal conductivity of epoxy is 0.19. Because of the low thermal conductivity of epoxy, and the small contact area between water and pump, very little of the heat generated by the pump’s motor will migrate into the water.

I would presume that RhoXS put about as much science into his other explanations as he did for his pump heat conclusions.

 

[JPSJPS]

LOL! I think ThePimpulator is JSPJSP under another name, as both are equally confused.
If you want to believe RhoXS go right ahead. But the errors in his posts are clear. For example, lets talk about pump heat.
He would have you believe that all the wattage of a pump goes into the water. This is incorrect.

You are something else! RhoXs is obviously educated in this field, has a lot of professional experience and overall is very knowledgeable.
You on the other hand have NO education, experience,or knowledge!
Why do you continue to make a fool of yourself?
Do you actually believe you know something about this subject?
Who do you think you are fooling?
You can't be stupid enough to think any knowledgeable person is going to take you seriously.

I bet you are the kid in the back of the class that causes a big commotion to get attention.

<<Long calculation that you do NOT understand (and applied incorrectly to this example) snipped>>

So we see that even a very powerful pump, working in its inefficient range, will only add about 18.5 watts of heat energy to the water. That is a far cry from 98 watts.The pumps we use are filled with epoxy. The thermal conductivity of epoxy is 0.19. Because of the low thermal conductivity of epoxy, and the small contact area between water and pump, very little of the heat generated by the pump’s motor will migrate into the water.

You can't read at a 5th grade level! RhoXS has explained many times that for centrifugal pumps, the 98 watts in this case is ONLY at 0 head and 0 Flow resistance. As head and resistance increase, pump input power decreases. The pump is NOT working at an "inefficient range", it is drawing much less power.
The power to the pump can be EASILY calculated by measuring the DC input Volts X Amps to the pump.
WARNING!!! Don't do this measurement or you will realize how full of $hit you really are!

For a submerged pump, "all the wattage of a pump goes into the water" because there is simply NO other place for this power to go. It can't get any simpler than that!

If the pump is not submerged, the body of the pump will transfer a % of its internally dissipated (wasted) power into the air instead of into the water. But, because the surface area is small, pump efficiencies are fairly high, and there is no forced cooling airflow, this % power will be VERY small.

If you want embarrass yourself by running a test, turn off your computer and the fan to the radiator but let the pump run. Measure the temperature of the water over time and report back to us.
WARNING!!! Don't do this measurement and report back if you don't want to prove yourself a fool!

I would presume that RhoXS put about as much science into his other explanations as he did for his pump heat conclusions.

You would NOT know science if it hit you in the head!
With every post, you continue to expose your complete ignorance!

 

[geoffman]

Does that take into account heat added by friction sources like friction in the pump movement itself, from seals etc?

As the bottom line of the forumla only has a mass movement (lb/min) and makes no account of total fluid mass contained in the system, I assume the formula applies to the system once it has reached a steady state and is out of start up transient conditions?

Edit asked this while JPSJPS had his "post incoming message up", now after reading his reply I am completely confused!

Edit No2 hunted around for a while and found McNally Institute Site , if I read correctly the quoted formula and example only calculates the temperature rise due friction withing the pump itself, and makes no allowance for rises due to energy input by the driving mechanism (if I read it correctly). If I read the rest of the site correctly it also seems that the formula is for centrifugal pumps that are directly coupled to the driving mechanism shaft, and not indirectly coupled magnetic drive pumps that the water cooling community seems to use. I'm not experienced enough (obviously ) to say whether it is valid for mag drive pumps or not. If anyone wants to see what what the McNally says about mag drive pumps it is here