Simple measurement of RADIATOR PERFORMANCE

[DaemonForce]

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

Aluminum rads suck, plain and simple. I started out 20 years ago with a Koolance 240mm rad with three scooped 80mm fans that would engage middle and upper presets whenever anything. Didn't help that I was literally putting it to task against the Pentium 4 Prescott either but that was the least of its issues. The dual inline pumps that were provided were quiet but troublesome because they were early designs prone to cracking or other failures. Then there was the issue of fans being annoying at any dB.

Aluminum radiators are not your friend in PCs. We have very high thermal output components that put high demands on weak cooling systems and you don't need the complication of mixed metals in the loop just to save a few dollars. Stick to copper + nickel or copper + gold if you're serious or aluminum + aluminum if you're a masochist. Also, bigger fans at a low speed help a lot.

I'm not going to take any of the opening data here seriously. It's a good try but there's no ambient temp listed, no clear delta and no information on other performing factors. There are a LOT of moving parts to water cooling and it's the reason I went with the data provided by Koolance to make things better. Now if only they could provide me with a nice low profile chipset block for X570.

 

[A Computer Guy]

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

I have artic P12 slims in my NR200P and at 100% they sound terrible. I may need to rethink my plan.

 

[BoggledBeagle]

I have artic P12 slims in my NR200P and at 100% they sound terrible. I may need to rethink my plan.

I was talking abour regular P12. I do not find them sounding terrible even at 100%.

I'm not going to take any of the opening data here seriously. It's a good try but there's no ambient temp listed, no clear delta and no information on other performing factors.

Ambient temperature is listed in each calculation sheet, as one value or as a time serie when I observed it changing during the measurement.

All performance figures depend on temperature delta.

Methodology and calculation is here clearly described and calculation sheets available.

Water flow is listed only in a few measurements, but from the professional specs sheets I posted is clear that cooling performance does not strongly depend on water flow.

 

[ShrimpBrime]

Aluminum rads suck, plain and simple. I started out 20 years ago with a Koolance 240mm rad with three scooped 80mm fans that would engage middle and upper presets whenever anything. Didn't help that I was literally putting it to task against the Pentium 4 Prescott either but that was the least of its issues. The dual inline pumps that were provided were quiet but troublesome because they were early designs prone to cracking or other failures. Then there was the issue of fans being annoying at any dB.

Aluminum radiators are not your friend in PCs. We have very high thermal output components that put high demands on weak cooling systems and you don't need the complication of mixed metals in the loop just to save a few dollars. Stick to copper + nickel or copper + gold if you're serious or aluminum + aluminum if you're a masochist. Also, bigger fans at a low speed help a lot.

I'm not going to take any of the opening data here seriously. It's a good try but there's no ambient temp listed, no clear delta and no information on other performing factors. There are a LOT of moving parts to water cooling and it's the reason I went with the data provided by Koolance to make things better. Now if only they could provide me with a nice low profile chipset block for X570.

Aluminum isn't a horrible choice, you just need a little more surface area than a copper radiator. If the load is low enough, aluminum would suffice. AIO coolers sometimes have aluminum rads.

None of the testing our friend does is accurate. He's not in a lab, but probably is good fun.

So far I've learned that the old saying

"For every 120 rad, you can dissipate 100w of heat."

Still holds true after nearly 20 years of everyone else already doing similar testing, but most of the time in real world PC environments. Not from a 500w pond heater and pump in a 5 gallon bucket.

 

[A Computer Guy]

Aluminum isn't a horrible choice, you just need a little more surface area than a copper radiator. If the load is low enough, aluminum would suffice. AIO coolers sometimes have aluminum rads.

None of the testing our friend does is accurate. He's not in a lab, but probably is good fun.

So far I've learned that the old saying

"For every 120 rad, you can dissipate 100w of heat."

Still holds true after nearly 20 years of everyone else already doing similar testing, but most of the time in real world PC environments. Not from a 500w pond heater and pump in a 5 gallon bucket.

For awhile I was holding down 5950x and Rx5700 with only one EK Coolstream P360 medium fan speed only so 100w per 120 sounds about right.

 

[BoggledBeagle]

I have some radiators of the same size to test:





Ideally, I would like to test all 3 fan configurations, 3 different fan speeds, that is 45 measurements. So far the preparation and measurent itself lasts 40 minutes, with me writing down temperatures each minute, and then I have to type all the temperatures in the excel sheet, which unfortunatelly does not work automatically, because I need to input the parameters of the regression functions three times. It is over one hour of total time to get the result.

The above plan is something like a week of testing a 45 buckets of warm water wasted, and it will not happen. I already tested all five radiators with fans at 100% speed pushing the air, any subsequent comparison which may come later will probably not use all five radiators.

Radiators (480mm):
Alphacool NexXxoS HPE-30: 14 water tubes in 1 layer - highest fin density
Alphacool NexXxoS ST30: 12 tubes in 1 layer
Alphacool NexXxoS XT45: 2x 12 tubes in two layers
Alphacool NexXxoS UT60: 3x 12 tubes in three layers
Alphacool NexXxoS Monsta: 4x 10 tubes in four layers - lowest fin density

I measured a water flow rate with the same pump:

Alphacool NexXxoS HPE-30: 4,6 l/min
Alphacool NexXxoS ST30: 5,1 l/min
Alphacool NexXxoS XT45: 5,8 l/min
Alphacool NexXxoS UT60: 6,4 l/min
Alphacool NexXxoS Monsta: 6,7 l/min

These numbers really do not reflect what was happening in the bucket, 4,6 l/min meant almost no visible water surface movement, while 6,7 l/min caused lively movement.

Looking at these numbers, I would really like to test flow rate through both 30mm thick rads in serie compared to Monsta with UT60... Those thick rads should better redeem themselves after my cooling power testing, which I will post later. (HPE-30 performed the best)

 

[A Computer Guy]

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

Food for thought. My NR200P Testing yesterday 1h Homeworld 3 gameplay attempting to balance performance and noise. Sorry I forgot to get wattage readings at the time TPU review indicates gaming should be about 220W.
( Only GPU in loop, 1 xspc 240 slim + 1 xspc 240 x-flow , RX 6700 XT 230W max , 450ML distilled + EK Cryofuel mix)

 

[BoggledBeagle]

Water flow rate (l/min)

Both slim 30mm rads in serie: 3,5
Monsta + UT60 in serie: 5,2

The difference is significant.



I have no mood to prepare all the calculations screenshots and measurement photos (will add them later), but the results with P12 fans at 100% and 10°C water over ambient are as follow:

Cooling power (W):

Alphacool NexXxoS HPE-30: 455
Alphacool NexXxoS HPE-30 PUSH PULL: 481
Alphacool NexXxoS ST30: 431
Alphacool NexXxoS XT45: 404
Alphacool NexXxoS UT60: 411
Alphacool NexXxoS Monsta: ***
Alphacool NexXxoS Monsta PUSH PULL: *** I removed the Monsta rad from cooling power testing, it has some problem, which I will investigate further.

As I mentioned earlier, significantly higher water flow is the only saving grace of the thick radiators.

I wonder what the results will be with reduced fan speed.

450ML

450 000 000 liters? That is a lot of water.

Please follow lower and upper cases strictly. There is 1 000 000 000 ml in one Ml.

MM is nothing, mm is a millimeter.
KG is nothing, kg is a kilogram.
G is giga-, g is a gram.

Etc.

 

[ShrimpBrime]

Water flow rate (l/min)

Both slim 30mm rads in serie: 3,5
Monsta + UT60 in serie: 5,2

The difference is significant.

View attachment 350778View attachment 350779

I have no mood to prepare all the calculations screenshots and measurement photos (will add them later), but the results with P12 fans at 100% and 10°C water over ambient are as follow:

Cooling power (W):

Alphacool NexXxoS HPE-30: 455
Alphacool NexXxoS HPE-30 PUSH PULL: 481
Alphacool NexXxoS ST30: 431
Alphacool NexXxoS XT45: 404
Alphacool NexXxoS UT60: 411
Alphacool NexXxoS Monsta: 378
Alphacool NexXxoS Monsta PUSH PULL: 412

As I mentioned earlier, significantly higher water flow is the only saving grace of the thick radiators.

The Monsta gets close to other rads only with two layers of fans.

I wonder what the results will be with reduced fan speed.


450 000 000 liters? That is a lot of water.

Please follow lower and upper cases strictly. There is 1 000 000 000 ml in one Ml.

MM is nothing, mm is millimeter.
KG is nothing, kg is kilogram.
G is giga-, g is gram.

Etc.

Haha, easy there buddy. We all know what he meant.

That bucket of water is closer to 150ML than almost all AIO and custom loops. Not even a close representation of how these radiators would be used in the natural environment.

 

[A Computer Guy]

Haha, easy there buddy. We all know what he meant.

LOL ... to late to edit now.

Water flow rate (l/min)

Both slim 30mm rads in serie: 3,5
Monsta + UT60 in serie: 5,2

The difference is significant.

View attachment 350778View attachment 350779

I have no mood to prepare all the calculations screenshots and measurement photos (will add them later), but the results with P12 fans at 100% and 10°C water over ambient are as follow:

Cooling power (W):

Alphacool NexXxoS HPE-30: 455
Alphacool NexXxoS HPE-30 PUSH PULL: 481
Alphacool NexXxoS ST30: 431
Alphacool NexXxoS XT45: 404
Alphacool NexXxoS UT60: 411
Alphacool NexXxoS Monsta: 378
Alphacool NexXxoS Monsta PUSH PULL: 412

As I mentioned earlier, significantly higher water flow is the only saving grace of the thick radiators.

The Monsta gets close to other rads only with two layers of fans.

I wonder what the results will be with reduced fan speed.

All this is all interesting but when putting together a custom loop being able to cross reference radiator performance (cooling power) at let's say 30%, 50%, 75% fan speeds would be really useful as it's unlikely one assembles a custom loop with the desire to run pumps or fans at 100% speed due to noise.

 

[BoggledBeagle]

All this is all interesting but when putting together a custom loop being able to cross reference radiator performance (cooling power) at let's say 30%, 50%, 75% fan speeds would be really useful as it's unlikely one assembles a custom loop with the desire to run pumps or fans at 100% speed due to noise.

I agree with you, for a lot of measurements it would be useful to have a datalogging multi-input thermometer, in the meantime I found one useful from Omega Electronics, I would need to pay $500 for it, as an investment for a momentary interest it seems like a lot of money, I also have no further use for such a device. My hot water tank is also small and does not allow for many measurement, I have no confinuous supply of hot water.

I want to test at least one rad with multiple fan speeds anyway, I was originally planning of adding 33 and 66% speeds.

 

[Ferrum Master]

Are you good with arduino/ESP32 or RPi coding?

To build a constant heat supply I recommend using TEC as in those small fridges. Basically you can regulate precisely the variables with it using dedicated log function. It will take some tinkering but it will be cheap. The data can be gathered via UART or even in better fashion, run a webserver on the thing and write some code that spews variables out there and save.

 

[BoggledBeagle]

I am not working with momentary measurement for which I would need steady supply of hot water, this measurement would require a precise lab grade thermometers and flow rate meter, as I mentioned earlier, you would just measure temperature of input and output water and quickly calculate the cooling power using flow rate.

I just work with cooling a body of water (8 kilos) by 15 or 20°C over a period of time (usualy it lasts 10-20 minutes), but even with this slow method a datalogging thermometer would help immensely. It would also solve the weakest point of my measurement, accuratelly measuring the temperature of the air sucked in the rad.

Therefore I do not need any constant heat supply.

 

[Ferrum Master]

I am not working with momentary measurement for which I would need steady supply of hot water, this measurement would require a precise lab grade thermometers and flow rate meter, as I mentioned earlier, you would just measure temperature of input and output water and quickly calculate the cooling power using flow rate.

I just work with cooling a body of water (8 kilos) by 15 or 20°C over a period of time (usualy it lasts 10-20 minutes), but even with this slow method a datalogging thermometer would help immensely. It would also solve the weakest point of my measurement, accuratelly measuring the temperature of the air sucked in the rad.

Therefore I do not need any constant heat supply.

How much precision you need? 9 bits of DS18B20 is not enough? ~0.5C?

 

[BoggledBeagle]

I need proper small thermocouples, which could accuratelly and quickly react to changing temperature of air (reacting to temperature of water is much smaller problem). I do not think that 0,5°C accuracy is enough.

Honestly when I measured that 2+3x 140mm rads in series in the small bathroom, the results are VERY dubious, because of very rapid air temperature increase in the room.

So far I found a very suspitious looking cheap 4 input thermometer (I would be worried to instal its software on my computer) and then normal 8 input thermometer by Omega, I could find use for at least 5 thermocouples.

I am trying to get around this air temperature measurement problem as best as I can, but for example yesterday I started with 23,1°C, then after 5 minutes I remembered to measure it again, and it was already 23,7°C, then it stayed the same for the rest of the measurement. In the calculation sheet I linearly interpolated the increase in the first 5 minutes, but is it probably not very accurate.

 

[BoggledBeagle]

Alphacool NexXxoS ST30




Alphacool NexXxoS HPE-30




Alphacool NexXxoS HPE-30 PUSH PULL




Alphacool NexXxoS XT45




Alphacool NexXxoS UT60




Alphacool NexXxoS UT60 PUSH PULL



The results so far:

 

[ShrimpBrime]

Would like to see spacers in place between rad and fans to eleviate the dead spot from the center of the fans to see if it increases or decreases performance. Not a must, but has been tested in the past with some positive results.

 

[A Computer Guy]

The results so far:

View attachment 351024

From the chart I don't understand how many watts your putting into the system that leads to your tiered temp results over ambient.

If I understand your chart correctly your saying the rad can dissipate x number of watts at y degrees C over ambient given 100% fan speed correct?
Also the number of watts is based on reaching steady state?

(I missed something or I'm not understanding something so my questions might be dumb.)

 

[BoggledBeagle]

From the chart I don't understand how many watts your putting into the system that leads to your tiered temp results over ambient.

The cooling power in results table is calculated from the thermal resistance number written in the table.

The measurement is simple: when the ambient is for example 23°C and water temperature in minute 6 is 35°C and in minute 7 is 33°C, the average is 34. Temperature delta in this time period was 11°C - this is the first input of the measurement. Using the specific heat of water, mass of water and the difference of 2°C, I get expelled energy in this time period (66880 J), which I then divide by the number of seconds in the time interval and get cooling power in watts - 1115 W. Using these two numbers I get the thermal resistance (0,0099).

I perform whole series of these measurements and that way I am getting more reliable result with my imprecise temperature measuring method. I am smoothing out the water temperature measuremens with regression function. How I mentioned earlier, the biggest weakness of my measurement now is imprecise measurement od ambient air temperature.

If I understand your chart correctly your saying the rad can dissipate x number of watts at y degrees C over ambient given 100% fan speed correct?

Yes.

Also the number of watts is based on reaching steady state?

What do you mean? Cooling power in watts in a momentary intensity value, which expresses how many Joules of heat you can expell during the time interval of 1 second.

"steady state" in my measurement is when the water temperature in the bucket reached ambient temperature and I am getting no data anymore.

Steady state in a PC is when the water temperature and the increasing cooling power of the radiators reaches a heat load of your components. If you were using UT60 rad and you knew that your heat load will be 370W, you will read in the table that the steady state cooling water temperature will be 9°C over ambient.

Or if you knew that your heat load will be say 550W, you can use the thermal resistance value of 0,0243 and compute, that you water temp over ambient will be 13,4°C.

Would like to see spacers in place between rad and fans to eleviate the dead spot from the center of the fans to see if it increases or decreases performance. Not a must, but has been tested in the past with some positive results.

That is a good idea for measurements, it could be also possible to test different spacer thicknesses.

 

[A Computer Guy]

What do you mean? Cooling power in watts in a momentary intensity value, which expresses how many Joules of heat you can expell during the time interval of 1 second.

I think you answered my question clearly , it just took some time to wrap my head around it. Thanks.

I might have a follow-up question later as I try to compare your numbers to my pc setup temps I posted earlier.

(edit)

Ok in my setup I'm putting 130w on average (188w max) into 2 x 240 slim rads reaching steady state of 7c @ 50% fan speed which seems is in the ballpark of what you said of using a single 240 slim rad at 100% fan speed at 220w. Basically using the 2 rads allows me to drop the fan speed by 1/2 and benefit from the noise reduction although it seems my temp over ambient seems a tad high for only 130w average. Perhaps because I'm using slim fans, case blockage (vent holes), and water additives prevents a better results closer to 5c over ambient?

 

[BoggledBeagle]

Ok in my setup I'm putting 130w on average (188w max) into 2 x 240 slim rads reaching steady state of 7c @ 50% fan speed which seems is in the ballpark of what you said of using a single 240 slim rad at 100% fan speed at 220w. Basically using the 2 rads allows me to drop the fan speed by 1/2 and benefit from the noise reduction although it seems my temp over ambient seems a tad high for only 130w average. Perhaps because I'm using slim fans, case blockage (vent holes), and water additives prevents a better results closer to 5c over ambient?

I see that you stealthily edited your message and I have not noticed that...

Case vents surely slow down the air flow, slim fans perform a bit worse than normal fans, but the additives may not have a huge impact, depending on how much you added, I already wrote elsewhere:

BTW I learned that ethylene glycol has specific heat capacity 2400 compared to 4180 J/kgK of water. Using 20% solution worsened the heat capacity of the coolant to about 3800, that is not even 10% worse compared to pure water.

I see no reason to use more than 20% of car antifreeze in the computer loop, it should have plenty additives in it to prevent some biological growth or corrosion.

 

[A Computer Guy]

I see that you stealthily edited your message and I have not noticed that...

Sorry wasn't trying to be stealthy also that run on sentence was rough I admit.

 

[BoggledBeagle]

I think it is pretty stupid that when you want to post some new information, your new message just appears in your previus message and nobody knows you added anything. So the forum does most of this stuff by itself...

 

[Steevo]

I see that you stealthily edited your message and I have not noticed that...

Case vents surely slow down the air flow, slim fans perform a bit worse than normal fans, but the additives may not have a huge impact, depending on how much you added, I already wrote elsewhere:


I see no reason to use more than 20% of car antifreeze in the computer loop, it should have plenty additives in it to prevent some biological growth or corrosion.

Ethylene Glycol improves waters ability to transfer heat VS distilled as it lowers the surface tension.

Radware Bot Manager Captcha

Also think about your project not in "how much data can I generate" but first in engineering terms, "how much can I remove and still get the result I need" A couple of K-type thermometers, a stopwatch and a camera. Heat the volume of liquid to X degrees, Run the same fans and measure inlet air temp, record with the camera and have the stopwatch showing. Maybe make a duct to hold the fans that you can place the radiator into.

 

[BoggledBeagle]

Ethylene Glycol improves waters ability to transfer heat VS distilled as it lowers the surface tension.

Radware Bot Manager Captcha

Just from the superficial glance over the article it is suspect, as it contains typos and data in a format not used in scientific articles. You do not write 0,00067m^3/s. NEVER. In the text they write 3 zeros, in the tables 4, in the abstract even 6. They should fix their zeros first and then publish anything.

They also test with 50:50 water glycol mixture, which has significantly lower specific heat capacity than pure water. It is physically impossible that it cools better than pure water. Pure water also flows and turbulates more easily. That improves cooling.

BTW a natural thing to test in this research was different water glycol mixtures and compare differences between them. They did not bother.

This is not a proper scientific article, but some quickly thrown together conference nonsense.

Also think about your project not in "how much data can I generate" but first in engineering terms, "how much can I remove and still get the result I need" A couple of K-type thermometers, a stopwatch and a camera. Heat the volume of liquid to X degrees, Run the same fans and measure inlet air temp, record with the camera and have the stopwatch showing. Maybe make a duct to hold the fans that you can place the radiator into.

This setup would bring some improvement to what I do, but I would need to buy more stuff than just one datalogging thermometer that I need.

I just bought one.

I will need to revise my calculation, improve physical measurement (I also have a problem that the inlet and outlet air are not separated enough, it would not harm if the bucket was a bit more isolated and covered, although I am correcting the measurement for the bucket self cooling as is).

I will then need to redo all the measurements.

Another problem is that now I have only one current 420 EK rad, 4 Alphacool 480 rads (current) and one older 420 rad. Every other rad I measured has been returned to the retailers. Retailers do not like me much...

In the meantime Excel gurus can tell me, where do I find the regression function parameters, so that I do not need to read them and input them in the sheet manually.