some more rad data

While these charts are good for picking the better rad when several different rads are tested, I’ve always found the actual data in these heat-dissipated-vs.-flowrate charts to be quite meaningless and even misleading.

In a real WC loop, radiator dissipation ability always drops with increased flow. Once you reach around 1 GPM, dissipation is practically going to stay the same no matter how much you increase flow. So when a chart shows dissipation increasing with flow, it’s misleading.

What would really be useful is heat dissipation vs CFM for a given flow rate. So pick a rate (1.25 GPM would be a good typical rate) and then vary the CFM and record dissipation.

Another useful modification would be to change the delta T for the different classes of rads...10C for a 120, 6C for a 240, and 3C for a 320. That’s more in line with the actual differences recorded in real systems.

Actually, a good way to compare the 120s, 240s, and 320s against each other is to maintain flowrate and CFM, and vary delta T until dissipation is equal. Then you can see what kind of temperature to expect when moving from a 120 to a 240 or 230.

I would love to test these things myself but I don’t have the equipment necessary or even space to setup a testing center. It’s very expensive to perform reliable testing.

Graystar said:

While these charts are good for picking the better rad when several different rads are tested, I’ve always found the actual data in these heat-dissipated-vs.-flowrate charts to be quite meaningless and even misleading.

In a real WC loop, radiator dissipation ability always drops with increased flow. Once you reach around 1 GPM, dissipation is practically going to stay the same no matter how much you increase flow. So when a chart shows dissipation increasing with flow, it’s misleading.

What would really be useful is heat dissipation vs CFM for a given flow rate. So pick a rate (1.25 GPM would be a good typical rate) and then vary the CFM and record dissipation.

Another useful modification would be to change the delta T for the different classes of rads...10C for a 120, 6C for a 240, and 3C for a 320. That’s more in line with the actual differences recorded in real systems.

Actually, a good way to compare the 120s, 240s, and 320s against each other is to maintain flowrate and CFM, and vary delta T until dissipation is equal. Then you can see what kind of temperature to expect when moving from a 120 to a 240 or 230.

I would love to test these things myself but I don’t have the equipment necessary or even space to setup a testing center. It’s very expensive to perform reliable testing.

Click to expand...

see bolded in quote: That's what I would like to see. When keeping the flow constant, the change in cfm would give us all a good idea of how it might work in our system.

only intended as a basis for comparison
heat exchanger sizing necessitates the temps and mass flow rates on both sides, far beyond the information available to the WCer

- specifically the air side, fan ratings are not a description of the actual CFM being moved through the rad; nor will the same fan on different rads be moving the same CFM
(I have occasionally plotted this same data by fan, assuming that the CFMs are a constant; I think such is even more misleading.)

the take-home message is really quite simple:
bigger rad = more cooling (and size)
more air = more cooling (and noise)
lower fin density = better performance with low noise (output) fans

sandwiching a rad with pairs of fans produces no benefit
a deep shroud produces no benefit
both tested several times

BillA said:

only intended as a basis for comparison
heat exchanger sizing necessitates the temps and mass flow rates on both sides, far beyond the information available to the WCer

- specifically the air side, fan ratings are not a description of the actual CFM being moved through the rad; nor will the same fan on different rads be moving the same CFM
(I have occasionally plotted this same data by fan, assuming that the CFMs are a constant; I think such is even more misleading.)

the take-home message is really quite simple:
bigger rad = more cooling (and size)
more air = more cooling (and noise)
lower fin density = better performance with low noise (output) fans

sandwiching a rad with pairs of fans produces no benefit
a deep shroud produces no benefit
both tested several times

Click to expand...

I can live with those simple rules. Unfortunately I get carried away with the fine differences sometimes (I work in IT as an engineer).

This is still useful to get some idea of heat dissipation that can be expected when selecting a rad for a setup given a particular heatload.

voigts said:

This is still useful to get some idea of heat dissipation that can be expected when selecting a rad for a setup given a particular heatload.

Click to expand...

The dissipation always equals the load. That's the primary problem with the tests.

If you’ve got 100 watts from the processor and 100 watts from a GPU, then your radiator is going to dissipate 200 watts. It has to...no “if”s, “and”s, or “but”s about it. The only question is at what temperature above ambient is this going to happen.

This was exactly the point David at Cooling Masters made with his comment on the testing procedure in his Black Ice GTS tests.

(translated by Google)
“Attention, we do not work at constant dissipated power, but with constant water-air variation, it is different!”

If we could test with constant load, constant flowrate, and constant CFM, then the variable will be operating temp over ambient...which is what we really want to know.

If heat dissipation always equals the load, then why do the graphs show an increased heat dissipation as flow increases? If the applied heatload is constant, then why do the graphs show increased heat dissipation? Where is the extra heat coming from?

voigts said:

If heat dissipation always equals the load, then why do the graphs show an increased heat dissipation as flow increases? If the applied heatload is constant, then why do the graphs show increased heat dissipation? Where is the extra heat coming from?

Click to expand...

Because Bill keeps a 10C air/water delta at all times. So, as each rad dissipates more heat, more heat must be added to the water to keep the 10c air/water delta. I also wish that Bill would test in a different matter - not give up the current just do another test. For example, keep a constant heat load of say 300w then record the air/water delta as your points of interest. So, a true real world comparison could be made between say the PA120.3 and the BIX3 with various fan speeds.

no, no
the 10°C coolant / air temp difference is based on the coolant inlet, in effect it is a constant 35°C as the air is held at 25°C
the heat dissipated is based on the coolant inlet / outlet delta T x the flow rate, the higher the flow the smaller the delta T even though the dissipation is increasing

look at the dissipation vs. air flow curves here http://thermal-management-testing.com/ThermoChill.htm
the problem is that the "air flow" is actually the fan spec under laboratory test conditions

I do not have the equipment to test at a 3°C delta T, I would need resolution to 0.001°C, and the scatter would be a horror

take what you can get and live with it, the relative performance would be not a bit different than with the present procedure

Ahh, thx for the explanation Bill.

Nice work Bill!

When discussing results like this isn't it usually done with non-dimentionals in academia ie Reynolds numbers? It's been awhile, but when I was working on my Masters Thesis I recall using Reynolds number for pressure drop correlations and I think it was Prandlt number for heat rejection? I'll dig out my thesis and take a look.

In any case using a consistent methodology is what is key because then the only "differing effect" comes from the differences in products.

Hi Owen
yes indeed, those and more
you trying to get me banned ?
- GOT to speak to the audience, I am already considered an arrogant pedantic fathead
jeez, it has taken me years to understand this, and I still 'overspeak'

nikhsub1 said:

Because Bill keeps a 10C air/water delta at all times. So, as each rad dissipates more heat, more heat must be added to the water to keep the 10c air/water delta.

Click to expand...

Though oddly worded, this is technically correct...if the rad is dissipating more heat, more heat must obviously be added to the circulating water to maintain the inlet temperature.

BillA statement is, of course, correct as well.

But neither answered voigts’ question, so here it is.

voigts said:

If heat dissipation always equals the load, then why do the graphs show an increased heat dissipation as flow increases? If the applied heatload is constant, then why do the graphs show increased heat dissipation? Where is the extra heat coming from?

Click to expand...

The extra heat is coming from the water, which is delivering twice the amount of dissipate-able heat when flowrate is doubled.

Remember the quote:
“Attention, we do not work at constant dissipated power, but with constant water-air variation, it is different!”

PC water-ccoling loops works with constant power dissipation. It is (nearly) equal to the power consumed by the components that are being water cooled. When radiators are tested the heat transferred to the water is not constant. It varies in order to maintain a constant delta T between inlet water and ambient. Understanding this, we can move forward.

When flowrate is increased but delta T between inlet water temp and ambient air is kept the same, you are transferring a greater amount of energy to the radiator. It is exactly the same as turning up the flame on a simmering pot. The pot now boils, but you wouldn’t say it’s performing better...it’s just transferring more heat.

At 1 liter per minute, a 10C delta means that a perfect radiator (cools water to ambient) would dissipate a max of 10,000 calories per minute. If you now double the flow to 2 liters per minute, but maintain the 10C delta, a perfect radiator would dissipate a max of 20,000 calories per minute...twice as much heat will be transferred. The radiator isn't working any better...it's just transferring more heat because there's more heat available to transfer.

Remember the heat transfer formula: Q = M x C x Delta T
Q = Heat
M = Mass
C = Specific Heat (water = 1)
Delta T = Temperature difference

In these tests the Delta T component is kept constant and Q changes with M (flowrate). In a PC water-cooling loop the Q (CPU power) is constant and Delta T changes with changes to M.

To put it another way, in the test when flow is doubled and delta T remains the same, Q is doubled. In a PC WC loop when flow is doubled Q remains unaffected and Delta T drops in half.

So as David at Cooling Masters says...it is different.

So, if I understand this correctly, the reason that rads are tested with a 10c difference between coolant inlet and air temp, instead of allowing for a variable Delta T, is because of the sensitivity needed in the testing equiptment to read results below the 10c difference in Delta don't exist or are two costly to be realistic?

This seems a shame as what most everyone wants to know is given a constant heatload, i.e certain CPU, GPU, etc., what is the actual heat dissipation of a given rad at a given flow rate, and how this heat dissipation translates into temperatures of said CPU, GPU, etc. One frequent question I read is the "how much of a temp improvement can be expected from a dual to a triple rad", etc. If we can't test given a constant heatload with a variable Delta T per rad, then how do we reliably get an answer to this kind of question?

extrapolation, + a grain of salt

voigts said:

So, if I understand this correctly, the reason that rads are tested with a 10c difference between coolant inlet and air temp, instead of allowing for a variable Delta T, is because of the sensitivity needed in the testing equiptment to read results below the 10c difference in Delta don't exist or are two costly to be realistic?

Click to expand...

This is a personal opinion, but I think the issue is the difficulty in having a system that imparts a specific heat load into the water. I think it’s trickier than it sounds. The current method is easier and more controllable. Also, the results from the current methods are perfectly valid as a comparative measure. If the test says rad A performed better than rad B, then it’s a very safe bet that rad A will perform better than rad B in your setup. But we would also like to know how much better...and that answer isn’t available.

voigts said:

This seems a shame as what most everyone wants to know is given a constant heatload, i.e certain CPU, GPU, etc., what is the actual heat dissipation of a given rad at a given flow rate, and how this heat dissipation translates into temperatures of said CPU, GPU, etc. One frequent question I read is the "how much of a temp improvement can be expected from a dual to a triple rad", etc. If we can't test given a constant heatload with a variable Delta T per rad, then how do we reliably get an answer to this kind of question?

Click to expand...

You can’t. At the moment such questions can only be reliably answered by the people who have the experience. When Cooling-Masters says:

Orders of magnitude of the deltaT water/air for 100 W dissipated

Simple radiator (1x 120 mm) 5 to 9 °C
Radiator doubles (2x 120 mm) 3 to 5 °C
Radiator triples (3x 120 mm) 2 to 3 °C

Click to expand...

...you pretty much just say “thank you very much for all your hard work” and believe it. At the moment I think that's as good as it gets.

Graystar said:

This is a personal opinion, but I think the issue is the difficulty in having a system that imparts a specific heat load into the water. I think it’s trickier than it sounds. The current method is easier and more controllable. . . . .

Click to expand...

yes
with an automated system its doable, but extremely time consuming;
running manually, instead of weeks testing say 3 rads, it would take several months because one is 'fishing' for a specific heat load that cannot be accurately predicted beforehand, or calculated until the system has stabilized

given that data points must be confirmed, it could take several days work to get a single validated datum

Well Thank-you BillA!
Nice to see you're still at work