Flowrate...

[I.M.O.G.]

AngryAlpaca, you are wrong on part of that - ONE DOES NOT WANT A LOW FLOW RATE THROUGH THE RADIATOR. Don't take this personally, this is just a common falsehood that A LOT of people still claim as true.

The graphs derived from Bill Adams radiator dissipation testing clearly show that most models of radiators perform considerably better with higher flowrates.

Higher flow rate makes the radiator dissipate more heat, this is shown in the graphs below. Dissipateing more heat is better. Water doesn't have to sit in the rad in order to cool down, it just has to continually move through the radiator so heat is constantly taken out as fast as possible.

Water sitting in the rad means the waters temperature differential between the air is less which leads to slower heat transfer. Sitting water is bad.

The graphs can be seen here. Higher flowrate is a lot better for some radiators, and a little better for others. Slower flow is almost NEVER better:

http://www.thermal-management-testing.com/corrected heat dissipation.htm

You can also look at some of my explanations that are more thorough in the 30-40 posts before this one.

 

[I.M.O.G.]

Sorry no msn man.

This has absolutely nothing to do with how long it stays in the radiator. This is the part of what you said priorly, and said again, that is wrong "Admittedly, with the water sitting in there for a while, it becomes less efficient, but better overall."

Not better overall either.

This is why:

The faster the water moves through the radiator, the more heat the radiator gives off.

You don't need the same water sitting in their longer to get the water colder - you need all of the water in the loop moving through the loop as fast as possible. This will lower the temperature of all the water faster, not just lower the temperature of the water sitting in the radiator more.

This is explained better in this thread:

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.

Finally, the advantage that you may be thinking of can only be had through using a larger radiator so that the water is in the radiator longer.

 

[Mithril]

Although it sounds counter intuitive, given two radiators that have the same water temp in and water temp out, you want the one in which the water spends the least amount of time.


Heres why, most people look at the temperaterature of the water at the input of the device as the temperature from which to base all heat transfer equations. If that were true, low flow would be good.
Unfortunatly, as the water passes through the water block or radiator, it heats up or cools down, and becomes less efficent, until it reaches the output temperature. Since the rate of heat transfer varies linearly with water temperature the more important temperature is the average temperature!

Average temperature = (temp in + temp out)/2 right?

WRONG!!!!

The temperature of the water changes fast at first but then slows down as the water become less efficent. Its an exponentially decaying relationship. So because the water spends more time at the output temperature the average temperature is always closer to the output temperature than the input.

Consider the following two cases:
In the case of very low flow the water heats up to the output temperature very fast and thus has a very high average temprature.

Now with very high flow the temperature doesn't get to change very much. Now if anyone knows anything about Taylor series or calculus in general you can see that the temperature change would be nearly linear. This would mean that the average temperature would be very close to (in+out)/2 which would be the most optimal average temperature possible.


So you actually want the water to spend the least amount of time possible (hence the high flow rate) in the radiator or waterblock.

Remember, there is always water in both the radiator and the wblock. Both are constantly transfering heat according to the water temperature, regardless of the time an individual water molecule spends in the the device. I find that it helps to think of the water as a whole instead of little pieces.

Anyway, just my 2cents, hope that clears up the confusion for someone.

 

[I.M.O.G.]

It's reassuring Mithril to see people get it.

A lot of people think the water needs to be in the radiator longer.

 

[kusojiji]

Mithril, nice write up. IMOG too.

Now, what does "exponentially" mean? Nah, only joking!

You know, even if people don't understand the physics to heat transfer, the graphs say it all. It's been 20+ years since I took physics in college, so it's nice to read these posts to jar those dead brain cells loose (provided they weren't already culled out by the dreaded beer machine!).

I used the test results from overclockers.com in making my decision on the design of my system and it works/worked very well (wish it had included my dang pump, which died too quickly!).

The only mistake I made/make is that I keep forgetting that you cannot compress water. I always worry about flow and designed my system to go from pump to waterblocks to radiator to reservoir. My thinking was that I'd maintain the highest flow to the waterblocks first and then it would slow down at the radiator. Well, the resistance of the system would affect the flow right at the pump outlet.

Now if I can get it, there may be hope . . .

 

[JPSJPS]

AngryAlpaca - EDIT: Negative response removed.

IMOG's and Mithril's analysis are correct. Nothing is gained by increasing the volume of water in the radiator so that the water stays there longer *if the surface area for cooling air is the same*. Multiple radiators will improve cooling only because the surface area is multiplied.

 

[I.M.O.G.]

Still wrong Angry.

Efficiency doesn't matter. How cool the water gets, does.

This is where you are thinking wrong - you seem so convinced that you are right that you continue to refuse reconsidering this, then we will not get anywhere and the problem of this lousy thread will be perpetuated.

How cool the water gets doesn't matter. Repeat that.

How much heat the loop removes does. Repeat that.

More heat is removed when the water spends less time in the radiator because it is moving through faster. Repeat that.

Now that I stated this. Again. I want to state a quote from earlier. Again. Tell me what you think of this quote:

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 quote explains why that quote is true, along with a plethora of other correct information supporting this fact throughout the rest of the forums:

Now with very high flow the temperature doesn't get to change very much. Now if anyone knows anything about Taylor series or calculus in general you can see that the temperature change would be nearly linear. This would mean that the average temperature would be very close to (in+out)/2 which would be the most optimal average temperature possible.


So you actually want the water to spend the least amount of time possible (hence the high flow rate) in the radiator or waterblock.

 

[I.M.O.G.]

It's the perspective taken that made them different.

Case 1: Water staying in the radiator for 10 seconds will get colder than Case 2 : water staying in the radiator for 5 seconds.

The first five seconds of heat removal with each Case is equal, with equal amounts of heat being removed.

The last five seconds of heat removal with each case is not equal. Case 1 will remove less heat than it did in the first five seconds. Case 2 starts over and removes the same amount of heat it did in the first five seconds.

So in a ten second interval, performing two Case 2's will be better than performing one Case 1.

In Case 1 the water gets colder before it leaves the radiator than in Case 2. However, two iterations of Case 2 can occur in the time it takes one iteration of Case 1 to occur.

Thus Case 2 removes more heat from the loop and is superior.

The way you were thinking of it, the water getting colder had a somewhat inverse relationship to heat removal - not the same thing.

 

[I.M.O.G.]

It's the perspective taken that made them different.

Case 1: Water staying in the radiator for 10 seconds will get colder than Case 2 : water staying in the radiator for 5 seconds.

The first five seconds of heat removal with each Case is equal, with equal amounts of heat being removed.

The last five seconds of heat removal with each case is not equal. Case 1 will remove less heat than it did in the first five seconds. Case 2 starts over and removes the same amount of heat it did in the first five seconds.

So in a ten second interval, performing two Case 2's will be better than performing one Case 1.

In Case 1 the water gets colder before it leaves the radiator than in Case 2. However, two iterations of Case 2 can occur in the time it takes one iteration of Case 1 to occur.

Thus Case 2 removes more heat from the loop and is superior.

The way you were thinking of it, the water getting colder had a somewhat inverse relationship to heat removal - not the same thing.

 

[Mithril]

AngryAlpaca, IMOG's right. The only important thing is heat removal. After any heat you remove has to come from somewhere right?



Imagine an ideal cooling loop, just a radiator and a waterblock.

Now the temperature going out of the radiator is the same as the temperature going into the cpu and visa versa.


The efficiency of both the radiator and the water block depend upon the average temperature of the total amount of water in each unit. So we want lower average waterblock temperature and higher average radiator temperature.


Now imagine an infinite flow rate. This implies that the water is a constant temperature at any point in the loop.

Thus the average radiator temperature = average waterblock temperature.

Now the water is not going to freeze becase the waterblock doesn't have any time to transfer heat to the water, nor is the water going to boil because it doesn't have time to cool in the radiator. The water will reach an equilibrium temperature depending on the ratio of the ablity of waterblock to heat the water and the radiators ability to cool it.



Having considered that example, lets decrease the flow to some finite amount. The input temperature to the water block will drop and the input to the radiator will rise, call them Ti and To.

If the average temperature of the water in the waterblock was equal to (Ti+To)/2 then flowrate would not matter. However, average temperature of the water in a device is ALWAYS closer to the output temperature than the input temperature as I mentioned earlier. Consequently, both the radiator and the waterblock are slightly less efficient with a finite flow over a infinite flow.



The real tradeoffs start with radiator design. First, having two radiators in series is just like having on radiator that's twice as long. Likewise, two radiators in parallel can be equiviquently represented by a radiator that's twice a thick. This can be viewed another way, a radiator can be considered to be a bunch of very thin tubes all stacked in parallel.

1). Radiators cool by having a large surface area. More surface area = more cooling potential.

2). To get more surface area you could either increase the thickness of the radiator (more tubes) or make it longer (longer tubes).

3). Increasing the length of the tubes increases their resistance to flow.

4). Increasing thickness (number of tubes) of the radiator reduces the resistance of the radiator to flow but it also decreases the flow within each tube.

I think all of this implies that given flow vs head curve and heat transfer rating of a radiator, there should be an optimal radiator design. However, (ideally) higher flow would allways improve the systems performance

 

[I.M.O.G.]

Case 2 is not impossible, and air chilling is not necessary.

It's not like the second and third part are not important, its that Case 2 makes the first most efficient part happen constantly so they aren't necessary.

 

[I.M.O.G.]

You said that it did the first 5 second cycle, where maximum heat is removed, AGAIN. That cannot happen, unless some really weird stuff is going on.

As long as it takes ten seconds or more for the water to make a full cycle, it is not impossible, and the relative comparison holds true. The actual amount of heat removed would not be exactly identical, but it would be virtually indifferent.

Regardless, that wasn't the strongest description of what happens, it was the last of a numerous amount of descriptions in order to explain as completely as possible. If you want to microanalyze something, microanalyze my earlier explanations because they stick closer to the actual principles... My last explanation was an abstraction of what happens so that I could explain the same thing a different way.

 

[Graystar]

Whoa! Another flare up of this thread and I missed it! Well, better late than never.

Mithril, I don’t know what you’re talking about.

Although it sounds counter intuitive, given two radiators that have the same water temp in and water temp out, you want the one in which the water spends the least amount of time.

Um…no. Given two radiators that have the same water temp in and out, it doesn’t matter which one you use, as the two radiators are dissipating the same amount of heat, regardless of their different efficiencies. Also, average temp means nothing. Temp out is the ONLY temperature that matters. Or more accurately, temp into the WB is all that matters. The water can be boiling throughout the entire loop, but if it is cooled down to room temperature one nanosecond before going into the WB, it will cool better than a system that maintains a very low average temperature, but lets water enter the WB at 1C over room temperature.

The temperature of the water changes fast at first but then slows down as the water become less efficent.

I didn’t know water had an efficiency rating. Can you please explain how that works?


IMOG, we’ve been through this before.

In your “Case 1/ Case 2” post, you seemed to have forgotten to add heat from the waterblock to the water on the second loop. But more important than that, you have totally missed the fact that if the water is traveling twice as fast, then it will not be as hot as it was in Case 1. Therefore, your observation, that the first 5 seconds of each iteration removes the same amount of heat, is wrong. That is simply impossible. Faster water means a lower water temperature. Therefore, for the first 5 secs, Case 2 will remove less heat.

But then, you seem to imply that you’re not going through WB again

As long as it takes ten seconds or more for the water to make a full cycle

And then you say

The actual amount of heat removed would not be exactly identical, but it would be virtually indifferent.

This, of course, is not only wrong, it is out of character, as I’m sure you can perform the calculations that demonstrate that the heat removed by the second 5 sec pass through the radiator is but a fraction of the heat removed the first time. You have, in fact, simply described two radiators in series. Both cases should transfer nearly the same amount of heat.

And herein lies the biggest problem, is that no one ever accounts for how the Waterblock and rad affect each other with changes in flow rate. In fact, no one even really considers what is happening to the water temp when flow rates are changed. There is a big difference between knowing things like “faster flow transfers more heat” and understanding how the different parts are working together in a system.

I’ve reviewed my last post on page 11 (near the bottom.) After considering all the new posts since then, I still think it is an accurate description of the process.

http://www.ocforums.com/showthread.php?s=&threadid=78055&perpage=30&pagenumber=11

 

[JPSJPS]

Graystar - You are missing the overall picture. Huge, complex and very expensive water/liquid cooling systems are used extensively in industry today. The design principles are well known and are based upon simple physics/thermodynamics. This stuff is not any new type of "rocket science".
Mithril and IMOG are trying to explain some of these principles. Their responses and over-simplified explanations are in response to other posts and specific situations and were not intended to be taken out of that context as general statements.
A professional cooling system design will involve developing a thermal model of each component based upon its specs, measured data, or a combination of both. Then the system calculations are relatively simple. Cooling performance can be calculated as parameters like liquid flow rate, CPU dissipation etc. are varied.
There is no such thing as "interactions between different components or one having an effect on another". However obviously one component can be the "weakest link" which will limit overall system performance.
A "seat of the pants" explanation based upon opinions is just not meaningful compared to these well proven explanations of "old well known stuff" based upon physics.

 

[Graystar]

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 watercooling. He says that a slower flow through the radiator is better.

“The goal of CPU watercooling is to cool the baseplate of the waterblock as much as possible. To do this, a coolant temperature of only slightly above ambient is sought, the thought being that the lowest possible coolant temperature will maximize the temperature difference - and therefore the heat transfer potential. As the efficiency of the radiator is greatest with the largest possible temperature difference, this would suggest lower coolant flow rates to maximize the heat rejection by maximizing the contact time of the coolant and the radiator tube walls.”

http://www.overclockers.com/articles481/

So anyone here that wants to claim that faster flow through the rad is better had better be able to prove BillA wrong.

 

[kusojiji]

I think BillA needs to answer this one. His posting on overclockers.com and thermal-management-testing.com was originally incorrect and had graphs that supported his statements that lower flow worked better at the different levels of air flow. It was corrected since, but only the graphs. The write up was not corrected and the final plot, comparing all of the radiators on one graph was also not corrected. If you look at the graphs of each radiator, performance does increase with flow on most of them and definitely all benefitted from higher air flow, while the cumulative graph shows the plots from the incorrect data.

I used the information in his graphs (easier to read) and chose the 68 Camaro 396ci heatercore. When I was using a 560 gph / 9.25 ft head pump, I had outstanding performance and did not really have to push too much air through it. I saw only a 2 to 3 degree C rise above ambient at full load with fans running at low speeds to be almost inaudible.

 

[kusojiji]

Take a look at this site: http://thermal-management-testing.com/ and compare the information from the original post and the corrected posting on radiator testing. Look specifically at the final plot where all the radiators are shown on a single graph and you'll see that the data used on that graph is the data from the "Radiator Heat Dissipation Testing - contains errors, revision needed (but not likely)" write up.

Also notice the "Corrected Radiator Dissipation Charts" are the ones that are posted on the overclockers.com site.

 

[Graystar]

If you look at the graphs of each radiator, performance does increase with flow on most of them...

Well, of the airflows attainable by the fans we use (the two lower air flows) 12 tests out of 22 experienced a gain in cooling capacity. A majority, but I wouldn't say "most". Also, those gains occurred at flow rates that are nearly impossible to reach with the pumps that we commonly use (even the high-end pumps.)

What IS clear is that the radiator design will determine what happens with higher flow.

 

[kusojiji]

Well, back to a deep subject:

Swiftech.com posted their "Real World Pump Testing" and the MCP600 achieved 8.4 lpm or 2.2 gpm. "The ½" ID system was made up of an MCW5002™ waterblock, a HW Labs BIack Ice Extreme radiator, and a vented (open) reservoir, all with ½" barb connectors; and the bent tubing to connect the components."

Even the Eheim 1048 was able to get 7 lpm or 1.85 gpm with the BIX.

Also note that the BIX rad has very high back pressure to water flow, as shown in the overclockers.com report.

So, I'd say that those flow numbers are not impossible to obtain. It all depends on how many tight bends and waterblocks are in the system and the equipment being used. An example of a low back pressure system would be a MCW5000 rev 1 and a caprice/chevette heatercore.

As for the air pressure side, I'll have to go by what the report indicates as far as achievable air flow/pressure (they don't have any good pictures!). So, say that the lower two bars are the practical ones, then you are correct in saying that a little more than half of the radiators exhibited better cooling. But, only one radiator showed a loss in performance in relation to coolant flow. My two AC fans are capable of .24 psi or 6.65 in H2O, so whatever that means to a heatercore.

But, waterblocks ALL benefit from higher flow. They do tend to even out at a certain flow number and the gains do not outweigh the cost in trying to get that little bit more of flow.

So, I'd say that as long as the OCWC 4 x 8 radiator is avoided, providing more flow to the waterblocks should be emphasized more as the radiators are generally more forgiving.

 

[Graystar]

Swiftech.com posted their "Real World Pump Testing"

I am always suspect of a manufacturer's test results. For example, in their P-Q chart for the MCP600 they have a discharge head of 3.2 meters, when the specs say 2.2 meters.

In this overclockers article the Eheim 1250 barely broke 1gpm.

http://www.overclockers.com/articles752/


In this overclockers article none of the pumps broke 1gpm.

http://www.overclockers.com/articles723/

But, waterblocks ALL benefit from higher flow.

Yes. But that means nothing. In making that statement you're assuming that the temperature of the water entering the waterblock is the same at both flow rates. It may, or it may not be, depending on the cooling capacity of the radiator. If you have a radiator like the Aqua Cool or Danger Den, which show a drop in cooling capacity with increased flow, then you WON'T get rid of the extra heat you transferred from the block. The heat stays in the water and the temperature of the water entering the waterblock increases to balance out the system. You experience no drop in CPU temperature.

It takes a lot of power to increase the flow of the system. To reduce CPU temperature, that power would be better spend on airflow.