Comparison of Radiator Cooling Efficiency
1) Introduction
2) Construction of testing environment
3) Data collection and interpretation
4) Fans
5) Radiators
6) Test results
a) Alphacool Monsta compared with Alphacool UT 60
b) Alphacool Monsta 360 with different fans
c) Coolgate 360 Cu compared with Alphacool UT 60
d) Phobya Extreme 200 compared with Collgate 360 Cu
e) Fintube radiator Admiral 360 compared with Coolgate 360 Cu and Alphacool UT 60
f) Watercool HTSF 2 compared with Coolgate 360 Cu and Alphacool UT 60
g) Magicool Monsta 420 compared with Coolgate 360 Cu and Alphacool 360 UT 60
h) Aquacomputer AMS 360 Al compared with Watercool HTSF 2 and Alphacool 360 UT 60
i) Mora 3 9*140 /4*200/8*200
....
7) Acknowledgement
1) Introduction
In 2011, I used an Intel i5 2500K / GTX 285 SLI PC system as a heat source to compare the new 420 AMS radiators with several other radiator types(1). When I started with new radiators and fans in the winter 2012, I had problems reproducing the values of a well measured radiator.
The deviation was so large that a comparison of eLoop fans with normal fans (Yate Loon) on a 360 radiator gave no clear results. Hence, there was no winner. I therefore decided to change the heat source.
2) Construction of the testing environment
With some help and discussions of experienced HardwareLuxx users and some old information from Dexgo, I constructed a system in which the heat source could be kept within a constant, measurable dimension and the (lost) transfer of heat out of tubes/hoses could be minimized.
There is a clear advantagehere; I could scale the measured value (Delta cw/rt [K] = Delta cold water/room air temperature in Kelvin) to a fixed heat output (500 W). With this fixed heat output value, it is possible to calculate (Delta cw/rt [K]) to any heat output you have on a real PC system.
I will integrate one measurement with reduced heat output and compare this to the "Standard" 500 W output.
This system and the achieved values are easily reproducible and less expensive than a PC system.
I used two external passage heaters (Hydor ETH - External Thermal Heater) each with 300 Watts of heat as output (datasheet) out of aquaria supply. In reality, each heater consumed 230-252 W of electrical energy. This power was mainly transformed into heat to increase the water temperature.
The radiators had at equilibrium (30-45 minutes after fan speed change) a constant heat quantity (I assume) and a constant flow. Both heaters were used for most of the measurements; otherwise, it will be mentioned separately. This system simulates modern high-end systems with one CPU (130 W heat output) and more than one GPU (SLI/Crosslink).
The test system was build out of a 60 mm cylindrical reservoir (Alphacool HF 38 Cape Cyclone 250 V.2), one Aquacomputer (AC) Aquastream Ultra (at 50 Hz pump frequency) with fitted AC Highflow DFM (flow meter).
Data Management:
Flow data and fan speed were collected with one or two AC Aquaeros 4.0/5.0. The fan speed was adjusted for each fan separate for up to 6 fans with the Aquaero and collected for every point of measurement. The final fan speed was the average of each fan collected by the Aquaero.
Temperature data was collected with three Greisinger Digitalthermometers. These thermometers were calibrated at 0?C and 100?C with an average deviation of +/- 0.1K inside this range.
The warm water thermometer was inserted into the 16/10 mm PVC-hose in front of the radiator inlet.
The cold water thermometer was inserted into the reservoir just behind the cold water outlet.
The air temperature was measured in front of the middle position fan (beneath the center of the rotor) in case of push mode. In case of pull mode, the air temperature was measured one to two cm in front of the middle of the radiator.
Energy consumption was measured with a Voltcraft Energy Monitor 3000.
I collected ten data points for air temperature (?C), cold water temperature (?C), warm water temperature (?C), consumed energy (W) and flow (L/h) for every fan speed inside a diagram. Every column of data was used to calculate the arithmetic mean and a root mean square deviation to get an impression of the system?s balance and quality of the measurement.
Sequence of components:
Reservoir (cold water digital thermometer) -> pump -> flow meter -> heater 1 -> heater 2 -> (warm water digital thermometer) ->VL3N -> radiator -> VL3N -> reservoir (circuit closed).
The PVC-hose was insulated with Armaflex. Armaflex junctions were covered with clear tape.
The heaters were insulated with three layers of blister film.
I used demineralized water with 15% G48 as fluid. The circuit had up to 1.5 L of fluid (regarding used radiator size)
Picture 1: Test system and test environment
Since I moved into a place with more space I have a similar setup:
Picture 2/3: New test system and test environment
3) Data collection and interpretation
I adjusted the fan speed for every collection of data (fan speed down to a critical cold water temperature of approx. 50?C) with a tolerance of +/- 20 rpm. The fan speed was logged inside the Aquaero and the average value of each fan was used at the end of a measurement to calculate the fan speed.
I tried to measure one dataset for every 200 rpm starting with the 12 V fan speed.
One experiment with fresh air ventilation (open windows) gave large differences in room temperature (+/- 1-3 ?C) and the measurement had a large root mean square deviation. I decided to measure (if possible) the complete fan speed range in one day. The room temperature went up during this measurement. The Delta cw/rt [K] was constant at a defined fan speed over a large range of room temperature if the system was in equilibrium (as expected). If a complete fan speed range could not be done in one day, the new measurement started the next day with the last fan speed. The Delta cw/lt was often not larger than 0.2 K.
To collect rough temperatures inside the Aquaero, three film sensors were applied (two inside the water, one in front of the radiator) to create a picture of the measurement with all data for later analysis. If the measurement showed strange, this data were used to find the reason. The film sensors had a large deviation to the digital thermometers in a nonlinear way. They could not be used to get exact data.
At the end of a data collection, all Delta cw/rt were correlated to the regarding fan speed in a curve diagram and compared to other data. Bar diagrams do not give the impression compared to curve diagrams if the measurement has inconsistencies.
All single pieces of data are available, upon requested.
The display of the radiator performance in bar charts are attached below (Overview bar graphs)
The flow is not displayed in detail. The flow was between 142 and 165 L/h at a 50 Hz pump frequency. The one exception will be discussed at the specific radiator. The flow for room temperature is attached in the radiator overview.
Without radiators the flow was 170 L/h at room temperature. I did not heat the system without radiator.
Every radiator received a rubber, silicon or polyurethane seal between the radiator and fan to minimize pressure lost.
4) Fans
I used one fan for all 120 mm radiators since it could be used for a fan sandwich too (I know, not the best one):
Yate Loon D12SH-12 (up to 2200 rpm without Radiator)
The following fans were used for fan comparisons:
Noiseblocker NB-eLoop B12-4 (up to 2400 rpm without Radiator)
Noiseblocker NB-eLoop B12-3 (up to 1900 rpm without Radiator)
Noiseblocker NB-eLoop B12-2 (up to 1300 rpm without Radiator)
Yate Loon D12SM-12D (38mm thick; up to 1350rpm without Radiator)
Scythe Gentle Typhoon Fan (up to 1850 rpm without Radiator)
Scythe Gentle Typhoon Fan (up to 5400 rpm without Radiator)
TFC Triebwerk 123 (55 mm thick; up to 2800 rpm without Radiator)
The following 180 mm fans were used, or are now available:
Phobya G-Silent 18 700rpm Red LED
Silverstone 180mm SST-FM181
Silverstone 180mm SST-AP181 (for pull mode)
Silverstone 180mm SST-AP182 (for pull mode)
The following 140mm fans were used, or are available:
Yate Loon D14SH-12 (up to 2300 rpm without Radiator)
Noiseblocker NB-BlackSilentPRO PK3 (up to 1700 rpm without Radiator)
At this point I would like to say: that I am waiting for the large (140mm/180mm) eLoops.
5) Radiators
Every radiator was analyzed with a ruler, calipers and a magnifier. The number of fins were counted and the fin surface was determined by a rough calculation (width*effective height*number of rows*number of fin packages). The volume of fluid, weight and number of tubes were measured/counted. In the attached table you find the above mentioned values. The table will be constantly updated.
The following radiators were analyzed and measured by this point:
360 mm Radiators
Alphacool NexXxoS Monsta
Alphacool NexXxoS UT60 Full Copper
Coolgate Radiator Cu
TFC Admiral (special edition??)
Watercool HTSF 2 3*120 LT
180/200 mm Radiators
Phobya Xtreme 200 - V.2 - Full Copper
420 mm Radiators
Magicool 420 Monsta
The following radiators are still waiting for measurement:
360 mm Radiators
Alphacool NexXxoS ST30 Full Copper
Alphacool NexXxoS XT45 Full Copper
Aquacomputer AMS 360 (one circuit) Al fins
MagiCool SLIM TRIPLE 360 Rev.
180/200 Radiators
Alphacool NexXxoS XT45 Full Copper 180mm Single
420 Radiators
Hardware Labs Black ICE Radiator GT Xtreme 420 - black
Magicool 420 Slim
TFC Monsta Lite
Alphacool NexXxoS UT60 Full Copper 420mm
Alphacool NexXxoS XT45 Full Copper 420mm
Alphacool NexXxoS XT30 Full Copper 420mm
XSPC EX420
Aquacomputer AMS 420 (Cu /Al fins)
Aquacomputer Evolution 360/420
Monsta Radiators:
MoRa 3 4*180mm
MoRa 3 9*140mm
[/spoiler]
Radiator Overview:
I have neither the hand nor the passion to make pictures of the unused radiators. Please excuse this disadvantage. The radiators pose much better if Bundymania has a camera in his hands and he has made pictures of all the radiators (most of them..) I measured.
6) Test results
I would prefer to display of test results in curve diagrams. Curve diagrams show very clearly if something during the test went wrong and where the limits of the radiator are.
The comparison of different radiator sizes in one diagram shows which one is better. If the curves are near each other, they show that the properties are similar at selected power outputs.
Overview bar graphs:
Data will be updated.
a) The 80mm thick Alphacool Monsta 360 radiator compared with the 60mm thick Alphacool UT 60 360er Radiator
I will start my measurements with two well-known Alphacool 360mm radiators:
The Alphacool NexXxoS Monsta 360mm and the Alphacool NexXxoS UT60 Full Copper 360mm.
The Alphacool NexXxoS Monsta 360mm is the thickest in its class and only a few cases will allow him to be under the case top. This is a front radiator which could be mounted in the drive compartment or in the ground floor of a Big Tower case (here maybe as the 480 mm or 420 mm version).
The 20mm thinner version, like UT60, has the same problem.
Both radiators have the ?Alphacool? fins which have a broken surface to increase turbulence and heat transport to the air. This surface gives some resistance to the air flow. The thicker the radiator is, the larger this effect will be.
How will they behave with a heat treatment of 500 W?
The comparison of fin surface shows that the Monsta has (if I did not make an error) a larger surface by 1/3 than the UT 60 (800.000 mm2 to 610.000 mm2 =12mm*48mm*8*132)
The cooling properties do not show this relation at 500 W.
The Monsta shows (3 Yate Loon D12SH-12D ) a 1K larger Delta cw/rt then the UT 60 at high fan speed.
At slow fan speed, the values of both Alphacool radiators approach each other (inside precision of measurements). At and below 1000 rpm, the properties of UT 60 and Monsta show no differentiation.
A fan sandwich with push/pull reduces the Delta cw/rt by 2 K. However, the UT 60 is still 1 K better than the Monsta.
This picture shows the deviation in consumed Watts during one measurement cycle. Using the real Watt value would give irritating curves. Scaling to a fixed Watt value gives us better comparable values.
Flow rate:
Alphacool Monsta - 160 l/h (+/- 3 l/h)
Alphacool NexXxoS UT60 ? 159 l/h (+/- 2 l/h)
b) Alphacool Monsta 360 with different fans
Here you find an overview of how the Monsta behaves with different fan types:
[img][/img]
This picture shows that the Monster needs a lot of pressure for ventilation, with two fan layers in push/pull configuration if possible.
The available fans show in push mode that the Gentle Typhon D1225C (max. 1850 rpm; 25mm), the Noiseblocker B12-4 eLoops (max. 2400 rpm; 25mm) and the thick Yate Loon D12SM-12D (max. 1350 rpm; 38 mm) are superior ( > 1 K lower Delta) to the used ?standard fan YL D12SH-12? because they create more pressure.
The old Feser Triebwerke 123 (max. 2800 rpm; 55mm) is better, but not with regard to the additional thickness. Two fan layers YL D12SH-12 give a better performance than every single fan.
c) Coolgate 360 Cu compared with Alphacool UT 60
Flow rate Coolgate 360 Cu: 158 l/h (+/- 3 l/h)
The Coolgate 360 CU is as thick as the Alphacool 360 UT 60, but has less fin surface and less volume. The properties at 500 W heat treatment are similar; the Alphacool is a little bit (0.3 to 1.0 K) better ? but this is in scale of root mean square deviation(+/- 0.1 to 0.2 K). Only at 1200/1400 rpm does there seem to be a significant differentiation ( 1K) due to higher Coolgate values.
These higher Coolgate results do not justify an additional run since they represent the overall picture. The curve graphs show even smaller deviations.
d) Phobya Extreme 200
Flow rate PX200 : 160 l/h (+/- 2 l/h)
The Phobya Extreme 200 (PX200) is one of the first 180/200 mm radiator that I have ever had. The used fans can be described as silent (Phobya) or loud (SST_AP). The Phobya 900 or the AP 182 are available now. The PX 200 become better with higher fan speed. There will be a new testing in the winter together with the Alphacool 180 radiator. At low fan speed temperatures of the fluid went up to 55?C.
It is interesting that the PX200 performs at low speed better than a 60 mm 360 radiator.
The PX200 has a mounting plate for a 200 mm fan, but I have no extra 200 mm fans in my collection now; this might change in winter.
An interesting detail is that the free space between fan and fins is small. I expect that there is some ?dead space? between the fan center and edges which do not get the required ventilation. Maybe thick shrouds can help. I plan to test this.
e) Fintube radiator Admiral 360 compared with Coolgate 360 Cu and Alphacool UT 60
The properties of the Admiral 360 need no additional comment if it is compared with two radiators of the same size. The Admiral?s flow rate was 112 L/h in cold condition and during the measurement of 124-127 L/h. A view into the in- and outlet of the radiator explains this; there seems only be one tube instead of a POM labyrinth as in the AMS radiators or a collector as in HTSF 2.
The system has a flow rate of 170L/h; with Alphacool or Coolgate the average flow rate was 160L/h (+/- 3L/h during the test). The Admiral is advertised as high flow and high performance ? make your own opinion.
f) Coolgate 360 Cu and Alphacool UT 60 compared with Watercool HTSF 2
At high fan speed, the difference of this nice tube radiator to classical radiator types is significant. The lower the fan speed becomes, the closer the curves are and the HTSF2 shows its quality. The 50 mm thick HTSF 2 was constructed for low fan speed as we can see here quite well (silent systems).
At a heat load of 500 W, the end is reached with 750 rpm. Nevertheless, the distance to 60 mm radiators is only 0.5 K and this a very good result for the 50 mm thick HTSF 2 radiator. Maybe a comparison to a 45 mm classical radiator with 250 W heat load can give additional information.
The improvement with a fan sandwich is with 3 to 4 K significant. The improvement with eLoops is only small (< 1K).
What happens if you add a second HTSF 2? I had the chance to prove this:
4 K coller at high fan speed (2100 rpm) and appr. 14 K cooler at low fan speed(700 rpm) compared to sandwich fan configuration. I think this is an argument for a second radiator.
g) Magicool Monsta 420 compared with Coolgate 360 Cu
Flow rate Magicool Monsta 420: 160 l/h (+/- 3 l)
The Magicool Monsta has a large fin surface. However due to his thickness, it has the requirement of high air pressure to perform. The fins are simple. Compared to a 360 mm radiator, it seems to be much better. As soon as other 420 mm radiators have been tested, the Monsta?s properties will be compared against the data.
h) Aquacomputer AMS 360 Al compared with Watercool HTSF 2 and Alphacool 360 UT 60
The AMS is as expected not so good as the Alphacool 360UT 60. Even the HTSF 2 is a bit better. I expected better results. The construction is more for low fan Speed optimized. Later I will try low speed fans at 300 W too.
Double sided fan configuration is completed. At low fan speed (for my asset) the graph is approaching the double sides HTFS 2, but does not reach it. For low speed usage the Aquacomputer AMS 360 should have two fan layers.
i) Watercool Mora 3 9*140 with different fan settings compared with one and two HTSF 2 with different fan settings
I have the impression that the setting with four 200 mm fans create a 2 K better cooling than the 9*140 mm fan setting at the same speed.
I can not measure pressure or airflow differences between the fans, but there is a strong indication that these 200 mm fans are more efficient. There seem to be a wide quality range (same voltage, up to 100rpm different speed). Silent (for me) from 400 rpm on.
The switch from one HTSF 2 (360mm radiator) with one layer Yate Loon fans to two HTSF 2 with one Layer of Yate Loon fans gave a real improvement in cooling. 5 K better at high speed (2100 rpm), 12 K better at low speed (700 rpm). Dual fan layer on one HTSF 2 improved cooling only by 3 to 4 K.
7) Acknowledgement
I would like to thank VDC and bundymania for many discussions in advance of the testing. They helped me avoid simple errors and start in the right direction.
Since I buy the equipment by myself, I have to thank my wife, in particular. She gives me her patience and frees up the space to let me do what I like in my "man cave".
I would like to thank Jon, my Canadian English teacher too. He corrected my awful school English.
Please feel free to criticize and give input or to discuss the results with me.
Please excuse me if my English wording or grammar is not in proper shape, corrections are welcome. I hope that this report contains no hard language, only hard results (sometimes I go like a bull at a gate ? not everybody likes this ).
Links to the German original reports:
(1) http://www.hardwareluxx.de/community...en-848542.html
(2) http://www.hardwareluxx.de/community...l#post20956104
Bookmarks