Simple measurement of RADIATOR PERFORMANCE

[BoggledBeagle]

I had two more rads on the bench... I mean washing machine!





Those two rads working together were kicking out so much heat, that I cannot presume constant ambient temperature and I had to adjust my calculation.

More details to follow.

I would need to cut off bits of the case to fit the large one...

Spoiler: the mid thickness smaller one performed nearly the same as the thick 360 I measured previously. The two rads in series still allowed more than 5 l/m with my wimpy 6W aquarium pump.

 

[mechtech]

Get a rad for a HD 3/4ton pickup truck.

 

[BoggledBeagle]

I know that there are powerful (and cheap!) rads for vehicles available and you could build yourself an extremely effective external cooling box with these rads, but I do not have space for additional box beside a computer, I need to fit everything inside.

 

[Lei]

I know that there are powerful (and cheap!) rads for vehicles available and you could build yourself an extremely effective external cooling box with these rads, but I do not have space for additional box beside a computer, I need to fit everything inside.

You don't need to fit things inside. They can just stay outside, it lowers the risk of leakage.

Your pump is inside the bucket, so you don't have to keep rads at a higher position than reservoir.
Use more coolant to benefit from water internal heat capacity.
Use thinner longer rads instead of thicker. A potato cut into halves cools down faster.

 

[BoggledBeagle]

You don't need to fit things inside. They can just stay outside, it lowers the risk of leakage.

What? I said that I have no space to use anything outside of my computer and therefore I have no interest in measuring anything that cannot fit in it!

Your pump is inside the bucket, so you don't have to keep rads at a higher position than reservoir.
Use more coolant to benefit from water internal heat capacity.

I am using 8 litres of water for the masurement and I spend more than 20 minutes sitting there writing the numbers down! I have no intention to use more water for the measurement to last longer!

Use thinner longer rads instead of thicker. A potato cut into halves cools down faster.

Rads are not uniform objects that are left to cool down, they a massive surface for heat exchange folded into a compact object. Thicker rad means larger surface for heat exchange.

 

[Lei]

I am using 8 liters of water for the measurement and I spend more than 20 minutes sitting there writing the numbers down! I have no intention to use more water for the measurement to last longer!

more water means temps take longer to rise. fans run quieter to cool.
you're designing a loop that will be cool and quiet. you won't be writing down forever

according to Q=mcΔθ your coolant will get warmer later than sooner if you have more mass of water.
if you use 8x2 liters, it will take twice the time for your temp to go up. resulting in fans running twice as slow to cool it down.
if you use 2 liters instead of 8, you'll spend 5 minutes to write things down under quadruple amount of fan noise.

Rads are not uniform objects that are left to cool down, they a massive surface for heat exchange folded into a compact object. Thicker rad means larger surface for heat exchange.

thicker rads need fans to rev up to pass the air through. I said use multiple thinner rads than a single thick one. hence you will have 2 halves of potato instead of 1 big one, which obviously has more surface area.

 

[BoggledBeagle]

Here are the new results:

By chance the performance of the EK P280M rad is exactly the same as of the Corsair XR7 360, the 280 is IMO more practical thanks to its compact size:



Thick EK X420M:



Both rads in serie. This time the change of temperature of the room was so significant, that I had to include it in the calculation (marked blue). The measured performance WITH THIS CORRECTION was even slightly better than performance of individual rads summed, which suggests that not correcting the measurement for increase of ambient temperature slightly undervalues the real cooling performance of the rads.

The water flow in 280 rad was approx. 6 l/m, 420 6,5 l/m, both rads in serie 5,3 l/s. The real difference of performance of individual rads summed should not be that much higher than measured performance of rads in serie, because the drop in water flow is not that high.



Summing up my findings so far:



I believe these values are not that far from the real performance. It would be really interesting to learn the performance of the rads with say 50% fan speed, but I am not planning to make this measurement. I think it would be safe to presume that the cooling power of the rads at this reduced fan speed or even overcoming some dust filters would be 30-45% of the above values.


I have at this moment two Corsair 280 slim rads in the PC. Upgrading them to these EK rads would improve the performance by about 30%, possibly more at lower fan speeds. These rads would be less water flow restrictive than the current ones (air probably too). These rads cost 80% more money than the current rads.

A NOTE ABOUT THE CALCULATION SHEETS: The second graph is not the end result, it is based on columns 8 and 9. There is one more graph based on columns 10 and 13 invisible in the screenshots, which is then used to determine the thermal resitance.

The final table with the results just uses the determined thermal resistance numbers, that is why the cooling power numbers are slightly different from those in column 13 of calculation sheets.

Again, whis was not a lab grade measurement, just something I threw together quickly, however with datalogging thermometer with at least two probes (ambient air going into rads and water temperatures), this simple method has a potencial to produce reliable numbers, and it is quite quick measurement, you just monitor the pretty quick process of the water cooldown, you do not need to wait extensive periods of time for the system to equalize or heat up in another possible measurement set-ups.

 

[BoggledBeagle]

... It would be really interesting to learn the performance of the rads with say 50% fan speed, but I am not planning to make this measurement. I think it would be safe to presume that the cooling power of the rads at this reduced fan speed or even overcoming some dust filters would be 30-45% of the above values.

I installed the X420M rad in my system (as the only rad), and from one combination of CPU and GPU power draw and water over ambient temperature derived thermal resistance of the rad. I am very happy to report that with the same fans at full speed I got VERY SIMILAR RESULT to what I got from my cooldown measurement. This VALIDATES MY RESULTS! Two totally different methods lead to a nearly identical result.

(The real life in case cooling power should be of course a bit lower than the "lab results", but I have no lab and the shortcomings of my cooldown measurtements have been already explained previously, so the opposite relation between these results come down to uncertainty of my measurements, not to any significant flaw.)

I am also positivelly surprised that running fans and pump at reduced speeds does not have that bad effect on the cooling performance!

Running both pump and fans at 70% speed dropped cooling performance by 36%, at 50% only by 54%! I expected much worse decline in performance.



I will add the screenshots with all the details of the above three measurements in my build log.

RATIONAL custom loop for 14900K and 4080S

Ok, so I guess it is really happening, I got a hose and water and other stuff: I plan to make a simple loop in Fractal Define 7 XL case: ... on these premises: 1) Use cheap stuff, no unreasonable and unnecessary parts - I can reuse some case fans, can mix my own cooling fluid, can use...

www.techpowerup.com

(The real life in case cooling power should be of course a bit lower than the "lab results", but I have no lab and the shortcomings of my cooldown measurtements have been already explained previously, so the opposite relation between these results come down to uncertainty of my measurements, not to any significant flaw.)

I just realised that this may have happened due to the fans cooling a bit better when pulling the air than when pushing it.

 

[BoggledBeagle]

I got this little 120 puppy on the bench:



The calculation:



Results so far:



It is interesting to compare the performance of P120 and X420 rads.



If you see them side by side, you may guess that the little one will cool ten times worse than the large one. No, the large one is just about 3,3 times better. And that is with one layer of fans pushing the air. I guess the difference could increase somewhat if you ran both rads in push-pull fan configuration.

Anyway, P120 + P280 should be roughly equivalent to X420 with much better placing options in the case.

 

[Ferrum Master]

Great, where you have the time for it tbh.

 

[BoggledBeagle]

Great, where you have the time for it tbh.

Now when I have the calculation sheet ready, self cooling of the system measured, I can set up the measuring in a few minutes and then spend 20-25 minutes measuring it, this one took longer because it is the smallest rad and it was cooling the water slowly.

The water heater method I used in the beginning would take longer, because it takes some time for everything to equalize.

 

[BoggledBeagle]

I accumulated a few new results:

EK P280M in push pull fan config:






EK X240M with normal P12 fans:






EK X240M with strong Coolermaster fans (0.36A vs 0.1A of P12 fans)





And finally EK X420M in push pull configuration:







The results so far:



Applying the second fan layer in a push pull configuration added 24% performance to the medium thickness rad and a third more to the thick rad.

The results of a thick 240 in comparison with medium thickness 120 were really disappointing. Twice larger area and thicker rad brought improvement of just 70%! Only once I used strong fans, the performance jumped to the expected level.

Thick radiators probably should be used with strong fans or with push pull fan configuration.

I believe that the old CM fans could be a little bit comparable to the P12 Max fans.

 

[BoggledBeagle]

I noticed that I forgot to post the last measurements I performed more than a month ago:

Corsair XR5 420






Corsair XR5 420 push pull





The results so far:



I think I can draw some conclusions, even though I tested some eclectic mixture of rads and some measurements were not that precise (you can judge how each meausrements went in the second graph - it should be linear and crossing zero point. The last mesurement is an example of measurement that did not go perfectly, the line is not straight and you can see that in the calculation I had to ignore that error of 45W visible in the line equation.

1) Rad cooling power mostly depends on rad surface, the rad with the same core will scale almost perfectly with number of fans

Commentary: You can see that a 420 rad with 50% more fans than 280 rad (same type) 46% higher cooling power.

2) Rad thickness has only limited effect on cooling power and the potential of thick rads needs to be enabled by using push-pull fan configuration (or very powerful and noisy one layer of fans)

3) Cooling power or the rads of the same size can vary by maximum of 40% - difference between slim rad with one layer of fans and thick rad with two layers of fans.

4) Adding second layer of fans to slim, medium and thick rads in a push-pull configuration improves cooling performance by 13, 25 and 32% respectively.

Commentary: Push-pull fan configuration has almost no sense for slim rads.



When I divide my results by number of fan positions I get these results:



5) Cooling power of rads for 1°C temperature of cooling water over ambient can be estimated:

For one 140 fan (P14 100% speed): 15-21 W, depending on rad thickness and fan configuration.

For one 120 fan: 11-15 W

Commentary: 120 fan performance is estimated from 140 fan results using rad core dimensions and corresponds well to measured results.



Rad performance can be easily estimated, for example:

One slim 360 and one slim 420 rad with one fan layer and acceptable water temperature over ambient 10°C: 10*3*11 + 10*3*15 = 780W.

One thick 280 rad in push-pull fan configuration and acceptable water temperature over ambient 15°C: 15*2*21 = 630W

These numbers are with normal fans at 100%, so with need for more silent operation these cooling power numbers would decrease somewhat. On the other hand, using extremely noisy and powerful fans the performance can be further increased.

 

[QuietBob]

Thanks for taking the time to share your findings! It was interesting to read about all these trials. I'm not into liquid cooling myself, but found this to be a learning experience as well.

 

[BoggledBeagle]

Thanks for taking the time to share your findings!

Well I got the information I wanted, I hope it will be useful for others too.

BTW thanks to taking a lot of measurements and then smoothing them with regression function I believe the results are pretty reliable.

Of course if I had an accurate flow meter and temperature meters, this measurement is doable much more easier, just with a quick calculation using the water flow and temperature of water entering and leaving the rad. I just checked again and again had no luck finding any flow meter claiming any sort of accuracy for acceptable amount of money.

 

[BoggledBeagle]

I got an idea to test difference between parallel and serial connection of two coolers, but I expect the cooling performance to be nearly identical.

However for example Corsair XR5 coolers are pretty restrictive for water flow, so using two paralelly connected 240 rads may be less water restrictive than using one 480 rad or two 240 rads in serie.

Anyway, here is a test of TWO Corsair XR5 240 in parallel connection:







I already tested XR5 420 rad, which has the same core as XR5 240 rad, and has nearly identical core area as two 240 rads. I am really happy to see, that I got also almost IDENTICAL COOLING PERFORMANCE (thermal resistance 0,0221 vs 0,0220), which again enforces my opinion that my measurements are consistent and pretty accurate.

 

[BoggledBeagle]

I got an idea to test difference between parallel and serial connection of two coolers, but I expect the cooling performance to be nearly identical.

However for example Corsair XR5 coolers are pretty restrictive for water flow, so using two paralelly connected 240 rads may be less water restrictive than using one 480 rad or two 240 rads in serie.

I am not sure about that now, because I just measured water flow in the above parallel setup to be 4,6 l/m and the serial connection of the two rads (with a thicker hose) resulted in just 2,1 l/m, that is less than half!

 

[BoggledBeagle]

Two Corsair XR 240 in SERIE

I connected the two rads in serie with a thick hose and got these results:





Before I presumed that the results will be about the same, or the serie connection will be something like 3-5% worse, then I saw the huge drop in flow described a post above and I had no idea what will happen, but in the end the drop in performance in comparison with the parallel connection was about 7%.

 

[damric]

So, after all the experimenting, does the the old rule of thumb of 100W per 120mm radiator come close to being accurate?

 

[A Computer Guy]

Rad performance can be easily estimated, for example:

One slim 360 and one slim 420 rad with one fan layer and acceptable water temperature over ambient 10°C: 10*3*11 + 10*3*15 = 780W.

One thick 280 rad in push-pull fan configuration and acceptable water temperature over ambient 15°C: 15*2*21 = 630W

These numbers are with normal fans at 100%, so with need for more silent operation these cooling power numbers would decrease somewhat. On the other hand, using extremely noisy and powerful fans the performance can be further increased.

Well this sounds interesting. I'm prepping to make my 2nd custom loop in my NR200P this summer and will be using an XSPC slim 240 to cool RX 6700 XT possibly without doing the necessary custom mod work needed to get a 2nd slim rad to fit on the top. I was starting to think I was crazy only using 1 slim 240 rad.

 

[BoggledBeagle]

So, after all the experimenting, does the the old rule of thumb of 100W per 120mm radiator come close to being accurate?

Is should be clear from what is written here that a crucial thing is what water temperature is acceptable. I can Imagine that some hardcore enthusiast with unlimited space and budget will be aiming fot the coolest water possible, for example max. 5°C above ambient, and some other guy who just wants the loop to work and is not aiming for some crazy low component temperatures will be ok with water 15°C above ambient. The second guy will need THREE TIMES SMALLER RAD AREA than the first guy.

You cannot formulate any rule of thumb, before you know, how warm you want the water to be. That is why I formulated the rad performance per 1°C of water temperature over ambient.

Well this sounds interesting. I'm prepping to make my 2nd custom loop in my NR200P this summer and will be using an XSPC slim 240 to cool RX 6700 XT possibly without doing the necessary custom mod work needed to get a 2nd slim rad to fit on the top. I was starting to think I was crazy only using 1 slim 240 rad.

Slim 240 rad can comfortably cool 220W with water just 10°C over ambient!

 

[A Computer Guy]

Is should be clear from what is written here that crucial thing is what water temperature is acceptable. I can Imagine that some hardcore enthusiast with unlimited space and budget will be aiming fot the coolest water possible, for example max. 5°C above ambient, and some other guy who just wants the loop to work and is not aiming for some crazy low component temperatures will be ok with water 15°C above ambient. The second guy will need THREE TIMES SMALLER RAD AREA than the first guy.

You cannot formulate any rule of thumb, before you know, how warm you want the water to be. That is why I formulated the rad performance per 1°C of water temperature over ambient.


Slim 240 rad can comfortably cool 220W with water just 10°C over ambient!

At what fan speed?

 

[BoggledBeagle]

100% Arctic P12. You could run the fans in push-pull to make it more quiet.

 

[damric]

Yeah even at 100%, P120s are pretty quiet.

I'm someone who needs coolant temp near ambient, and that's because my ambients are always high because of climate. So yeah I try to overkill radiators.

 

[Keullo-e]

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?

Why you don't just get an external drive (or an enclosure for your current one) if you need one?