Simple measurement of RADIATOR PERFORMANCE

[BoggledBeagle]

I seriously started considering building a custom watercooling loop for my computer, which with current 14900K and a 4080 I am thinking about getting would need to cool approx. 500W.

I decided I would like to know the approximate cooling power of the radiators, at this moment I am thinking about using two slim 280 rads or one slim 280 and one thick 360.

Since I do not have any configurable power supply and heating sources, I decided to measure the performance of the rad simply with a bucket, aquarium pump, and a tourist 500W water heater. I simply put a warm water in the bucket, let the rad cool it and recorded the temperature of the water at certain time points and then calculated average cooling power in these time intervals. Then I just let the water be warmed by the heater and waited for the temperature of the water to stabilise.

Here are some pictures from the measurement:





The correct measurement should first establish the self cooling of the system and then subtract this from the results, however I believe that this will not be worse than 20% difference and I adjusted the results with this coefficient. I am not looking for an exact lab measurement, just some ballpark cooling power numbers. There are also better ways to measure water temperature than with an IR meter measuring rad core.

Here is a calculation table and the results table:





The slim 280 rad with cooling water 10°C above ambient could have cooling power of about 200W, and with 15°C over ambient 350W. The control measurement with the tourist heater confirmed the cool down measurement pretty well.

With better fans (I just used some crappy case fans that came with my Fractal case) the performance should improve. I think I could get away with using two of these slim 280 rads, but I am still curious how the thick 360 rad will perform.

How important are the fans for the radiator performance, could for example Arctic P12 fans perform significantly better than the case fans I used?

Why is not the cooling performance of the radiators somehow specified by the manufacturers?

I had to buy the aquarium pump and the heater for this measurement and also spent few hours getting the things together and measuring it, so I hope this will be useful for somebody else than me as well.

 

[Vya Domus]

They usually do specify some "cooling power" figure however it doesn't mean anything outside of whatever parameters they used to come up with that number, it's kind of a waste of time trying to measure this in an objective way. I'd get two 360s for your system.

but I am still curious how the thick 360 rad will perform

It's generally better to get more slimer rads than fewer thicker ones because that way there is more airflow/area. For instance two 3cm rads will perform better than one 6cm rad because you're going to have twice as many fans so twice the airflow.

How important are the fans for the radiator performance

They matter to some degree but it's mostly just up to RPM and noise. I use P12, they're pretty quiet up until 50% speed.

I had to buy the aquarium pump and the heater

You could have just heated the radiator with a hairdryer, much simpler. I do that when I leak test the loop, it makes the tubing softer so you can tighten the fittings more.

 

[BoggledBeagle]

I am limited by the optical drive, without it I would have much more options.

BTW I was surprised that the water manages to cool in the radiator just by 0,6°C or so.

Are not the thicker rads less restrictive?

 

[Vya Domus]

Are not the thicker rads less restrictive?

Not in any significant manner.

 

[KedsDead]

I am limited by the optical drive, without it I would have much more options.

BTW I was surprised that the water manages to cool in the radiator just by 0,6°C or so.

Are not the thicker rads less restrictive?

Restrictive would come down to fin density, I believe.

As for it only cooling a certain amount water in your bucket, you would also have to include the volume of water you're circulating. If you are drawing water from the bottom of the bucket and returning it in the top, you need to have enough circulation to move the water from the top to the bottom. Cold water will drop the bottom while the warming water will stay on top, thus throwing off your measurements.

** Edit

In your pictures, your return is right next to your pump. so your pump is sucking in the warm water from the return.

 

[Bones]

Have you thought about an evaporative system instead?

Those can cool water to sub-ambient temps (About 5c or so from ambient) and I had created one before, it worked well and handled heat loads just fine, even from my FX-9590 in an overclocked state in the middle of August.
It's efficient and most can handle large heat loads with ease BUT you must set it up correctly or it won't work well and there is the additional factor of it being evaporative, meaning it will lose coolant over time because the water/coolant is constantly evaporating to atmosphere from it.

I'll refer you to shop fans that have fins and water in them - That's what they are and those have to have a small water supply line (Garden hose in most cases) ran to them for keeping them in operation.... But they do work.

As for thicker vs thinner rads, you also have to think about the amount of material absorbing and passing heat to atmosphere, in some cases a thicker rad is better because it has more material to absorb heat and pass it along.
Also, you don't need fans capable of creating a mini-hurricane but you do want some decent fan CFM capacity regardless.

Thinner rads in many cases will cool better but with a really high thermally active (BTU) application are also subject to becoming "Heat Soaked" if your cooling requries alot of thermal capacity.
All that means is heat is building up faster than the system itself can remove it.

It's the same effect as cooling an stock engine vs a high performance engine in an auto. Run a thinner rad with a high performance engine, it's going to run hot.
Thicker rads in an auto do better than thinner ones will and are used with larger displacement engines for that very reason and it does apply because it's by the same principals of operation a computer watercooling setup works by itself.

Thinner rads are really not as good as thicker ones are because a thicker rad has more/larger flues or "Rows" to expose the coolant to it's cooling medium (Which is moving air) while air is passing through it. That means by volume in a given area/surface, you have more coolant being cooled at the same time in that area of the system which does increase it's BTU capacity overall but know this increase in efficiency diminishes as heat loads climb.
Any cooling system has a certain BTU capacity and once you're there - That's it and heat soak begins.

However.....
Linking two or more rads as suggested above should do the trick. I'd suggest a pair of 360's myself linked and they can be thinner ones, you won't be getting as much heat to the second rad anyway so it can make up for the lack of thickness a single thinner rad would have.

 

[BoggledBeagle]

Restrictive would come down to fin density, I believe.

As for it only cooling a certain amount water in your bucket, you would also have to include the volume of water you're circulating. If you are drawing water from the bottom of the bucket and returning it in the top, you need to have enough circulation to move the water from the top to the bottom. Cold water will drop the bottom while the warming water will stay on top, thus throwing off your measurements.

** Edit

In your pictures, your return is right next to your pump. so your pump is sucking in the warm water from the return.

I meant restrictive for water flow. I weighted the amount of water in the cool down measurement, it is in the calculation picture. The water was well agitated in the bucket.

The real problem of this masurement is that I was lazy to measure self cooling of the water in the bucket without a rad.


ANOTHER POST ADDED TO THE PREVIOUS ONE:

Another trip to hardware store, I got a thermometer and a hose connector.

This time I did it properly, adjusted the measurement for self cooling of the system, I also used Arctic P14 fans at full speed. Here you can see shots from the measurements:




The calculation sheet:

CORRECTIONS NEEDED, WILL UPLOAD FIXED MEASUREMENT, CALCULATION AND RESULTS LATER

The results:

CORRECTIONS NEEDED, WILL UPLOAD FIXED MEASUREMENT, CALCULATION AND RESULTS LATER

 

[ShrimpBrime]

Are these radiators aluminum? How do they compare to copper rads of similar dimensions?
And are these single pass or double pass rads?

 

[BoggledBeagle]

I used this radiator:

Hydro X Series XR5 280mm Water Cooling Radiator (corsair.com)

 

[ShrimpBrime]

I used this radiator:

Hydro X Series XR5 280mm Water Cooling Radiator (corsair.com)

Nice.

Just set a loop up last night. Looks similar to this, but using a much smaller bucket and a much larger radiator. The water level is only to the top of the pump. Maybe half a gallon total including filled pipes and block and rad.

Only have 1 fan running, some old sun on fan, has good pressure though.

Went 3/8" ID this time. Had a 1/4" barb in the mix. Extremely difficult to clamp 3/8" hose to a 1/4" barb. I tired, it leaked lol. Went up to 1/2" on one side of the rad and heat up the tube and force it on. Haha.

I'm using tap water with a few drops of bleach.

Was thinking I should go back to blue windshield washer fluid. Always seems to cool well in a loop. Have you thought to compare a different liquid to the water?

 

[Vya Domus]

Are these radiators aluminum? How do they compare to copper rads of similar dimensions?

99% of the radiators you can buy are copper.

 

[ShrimpBrime]

99% of the radiators you can buy are copper.

I suppose. Don't buy them often honestly.
Mines a double pass fatty. 120.3. All copper. real nice.

I know Alphacool makes (made) Aluminum Radiators.
Don't think I've ever had an aluminum rad come to think of it though. You're probably 100% right!

 

[Vya Domus]

I know Alphacool makes (made) Aluminum Radiators.
Don't think I've ever had an aluminum rad come to think of it though. You're probably 100% right!

AIOs usually have aluminium rads even if the jetplate is copper because they're sealed units and they just put in the required anticorrosion agents in the liquid. The standalone radiators you can buy are usually copper, DIY watercooling people are pretty averse to anything made out of aluminium because of the risk of corrosion, so aluminium rads/blocks are pretty much nonexistent.

 

[ShrimpBrime]

AIOs usually have aluminium rads even if the jetplate is copper because they're sealed units and they just put in the required anticorrosion agents in the liquid. The standalone radiators you can buy are usually copper, DIY watercooling people are pretty averse to anything made out of aluminium because of the risk of corrosion, so aluminium rads/blocks are pretty much nonexistent.

I've made aluminum blocks before. Not the greatest results for a water loop. But was fun experimenting.
I just use ethylene glycol and distilled water. it's really good against growths.

You're speaking of electrolysis?

 

[BoggledBeagle]

TRAGEDY: the adjustment for self cooling is all wrong !!!! I will have to redo the measurement and calculation.

 

[Vya Domus]

You're speaking of electrolysis

Yeah

 

[BoggledBeagle]

Everything would have been so much easier if I had a thermometer with data logging.

I finally did the thing properly and measured the self cooling of the measurement system.

Here are the results collected over 4 hours and calculation:



Then I remeasured the cooling with radiator again, here are the results and calculation with correction for self cooling:



And the final result, which should be pretty correct:



So some thought about my intended loop:

Using one 280 rad to cool 500W would lead to over 20°C temp over ambient water, two of them acceptable 8°C, three 6°C - so only a marginal improvement.

Using 360 rads: one 17°C, two 7,5°C, three 6°C.

We could do the calculation more precisely by thinking about temperature loss in each radiator: what if the first rad cools the water by 0,5°C?

For example 280 + 360 rad combination:
280 with cooling water 8°C handles 251W and 360 with 7,5°C 256W.

So this combination of radiators should be able to handle 507 W with such water temperature. Or course, cooling water temperature is a result of the heat fed into the cooling system...

 

[Bobaganoosh]

Koolance provides performance graphs on their radiators. It does not translate directly to irregular loads from a PC, but it is still helpful information. One example. I find it a helpful reference in combination with some reviews that have tested it and others when comparing radiators with different fpi and thicknesses. You can get an idea for the impact of flow rate as well as an expected delta-T for consistent heatloads.

 

[BoggledBeagle]

Koolance provides performance graphs on their radiators. It does not translate directly to irregular loads from a PC, but it is still helpful information. One example. I find it a helpful reference in combination with some reviews that have tested it and others when comparing radiators with different fpi and thicknesses. You can get an idea for the impact of flow rate as well as an expected delta-T for consistent heatloads.

This is interesting, unfortunately the specs have limited usability IMO.

Here are the specs of different sizes of 140mm rads, which show, that the heat dissipation really nicely depends on cooler surface, so the estimation of the cooler performance I made should be quite useful. It also shows that the heat dissipation is not that much dependent on the flow rate.

Their 280 rad for a given flow rate can dissipate 500W with water to ambient 15°C difference, I measured 427W with my rad.



Here are complete specs for 360 rad:


The first two graphs are water flow / pressure diagrams which would be useful if you were designing some system with a lot of these rads, so that you would know what pump to use, for a normal user it is useless.

The last graph for delta 15°C and for example 4l/m flow rate presumes around 700W performance, which is weird in comparison with their 280 rad - their surface areas difference would indicate much lower difference between their performance. Something may be wrong here.

The second to last graph for a given fan and flow rate combination gives you heat dissipation (or heat resistance), which is a LINEAR relation between delta temperature and heat flow.

That is weird, because my results for the 280 rad show logaritmic curve:



Truth to be told, my curve seems leading to something like 0W at 3°C delta, which is a nonsense. The curve should start in the origin of the graph.

If the dependence should really be linear, I may be off in the table by no more than 15%. If my errors concentrate just to tha small deltas, I may be correct in the higher values, but values for small deltas may be off by 50%.

Blue are data from the result table, I added the origin.

 

[damric]

99% of the radiators you can buy are copper.

The reverse is true where I shop. When building cheap loops I buy the 240mm aluminum with built-in barbs direct from China on ebay. They are cheap asf and perfect for my low cost folding@home farms and HWBOT benches.

I'll have to show you guys some of my builds and show a parts list and how well they perform compared my high dollar stuff in my personal rig.

 

[lilhasselhoffer]

So....
500 watts = 500 joules/s
The thermal capacity of water is 4.2 joules/gram/degree C. The average 5 gallon bucket of water is 18,927 grams of water

This means on average (1/4.2 gram*degree C/joules)*(1/18927 1/gram)*(500 joules) = 0.0063 degrees C per second


That means you are only increasing the temperature of the water in that bucket by about 0.0063 degrees C per second. This is not equivalent to a 500w load on the heat plate...and this entire experiment fails to be more than an exercise in showing that heating up water is expensive. You lose energy in the bucket. The steady state temperature of the bucket is unknown. The energy being dumped into the loop is not equivalent to 500 watts, because the delta in temperature between the loop and your huge thermal capacitance bucket is silly.


Let me suggest that instead of this bad measurement you buy a cheap thermocouple. You then buy a cheap space heater. Crack the space heater apart, and set its quartz heating element against the block. Make sure it reaches 90 C with the thermocouple...and remove the bucket entirely. Now you have something useful. The problem is that heat transfer is deltas in temperature...and you start this crude measure out by instead measuring the steady state with a huge thermal question mark assuming that 500 watts of heating into a large thermal mass is the same as a functional point load of 500 watts...which is actually thermally limited instead of wattage limited.

What you are measuring now is not equal to a 500 watt input load on the plate. It is in fact a much lower load...so your math proves that the radiator can come to a steady state with a huge energy loss into a water bucket...so the math to show it's x watts of capacity is bogus GIGO.

 

[BoggledBeagle]

In principle there is nothing wrong on this measurement, but unfortunately a number of things made my measurements not very realiable.

My calculation was flawed, I will post detailed information and corrected calculations and results in about 5-6 hours.

 

[damric]

Thermal power = mass flow rate times specific heat capacity times the difference of outlet and inlet temperatures. At least that's how we calculated it as reactor operators. The problem with adapting that to pc cooling loop is that the temp differences in the inlet and outlet are quite small and you would need precise probes on each, and it might also be difficult to measure your coolant mass flow rate. But if you can, that equation tells you how much thermal power is being dumped into the loop from the CPU and how much thermal power the radiator is providing.

 

[BoggledBeagle]

I corrected my calculation method.


A simple measurement of radiator cooling performance – description of the method.

Let the rad cool a body of hot water, record the temperature in certain time points, then calculate the heat removed from water in the intervals between these time points and get cooling power from that.

A free standing body of hot water will cool by itself, so you first need to establish, how the measurement systém cools down without the radiator in the loop, and then subtract this self cooling the the total cooling power results.

Details from the calculation are visible in the calculation sheets:









The measurement would provide good reliable results, if you had a thermometer with datalogging, which could not only record the temperature of water, but also ambient temperature. In my measurement I presumed constant ambient temperature, which made my results less reliable.
Also if you wanted to compare different radiators, if would be practical to have an adjustable pump and a way of measuring flow rate, so that you could ensure the same flow rate for different radiators. The body of the water in the bucket should be well agitated and mixed, preferably not just with the flow of water from the pump, but also some other agitator.

I used primitive thermometer, aquarium pump without regulation, presumed constant ambient temperature, and the water in the bucket may not have been optimally agitated, which made my measurements not very reliable. For example, results for a thick 360 rad are worse, than a 360 slim rad would perform after being estimated from core areas difference from performance of 280, which I measured.

I measured single rads: thin 280 and thick 360, and combination of both in series.

Resulting Thermal resistance:

Thin 280: 0,030

Thick 360: 0,028

Both in series: 0,017

For example, radiator with thermal resistance 0,017 would with coolant temperature 25°C over ambient would have cooling power 25/0,017 = 1470W

On the these calculation sheets you can see absolute parts of the power dep. on delta temp graphs, which are results of my imperfect measurement:
Self cooling: 9W
280 slim: 1,25W
360 thick: 17W
280+360: 0,5W.

These errors have been ignored and the final performance results calculated just with the above mentioned thermal resistence parameters.

Comparison of the results in one graph, 360 thick rad is only slightly better than thin 280 rad, I tried to add the cooling power of these two rads and then compare it to the measured results, and the difference it not that dramatic.



With proper equipment:

thermometer with datalogging to record water and ambient temperature
device for measuring flow rate
adjustable pump
water agitator
insulated body of water

you could with the above explained method get very realiable results.

On the other hand, my results are probably not that far from correct values and they contain some practical usable information.

 

[BoggledBeagle]

Just adding photos from 360 thick and combination of rads measurement, which I forgot to add.