Thermal Testing
I’m using my new V2 radiator test bench for testing which nets significantly lower watts dissipated numbers than my old open air testing. This new bench is both insulated and shielded from getting any external cooling help and likely more accurately simulates a closed case condition. The tested parts are as follows:
Test Specifications:
- Thermal Sensor logging: Crystalfontz CFA-633, logs up to 32 temperature probes and up to 4 fan channels: http://www.crystalfontz.com/products/633/index.html
- Temperature Probes: Dallas Digital One Wire DS18B20 probes. These are good to about .2C absolute accuracy in normal water cooling temp ranges and have a nice fine .0625C resolution. Also since they communicate digitally, you can string the power, ground and Vdd wire in series between all the sensors limiting the amount of wires significantly.
- Pump: Swiftech MCP-35X2 at 40% PWM. Simulates medium pumping power and results in roughly 1.5GPM +- depending on radiator restriction.
- Block: Danger Den MC-TDX block
- Heater: Standard Aquarium 300W Heater with safety switch soldered in the closed position so heat remains on regardless of temperature. This requires removing the heater element from the glass tube, soldering the two tabs together and putting it all back together.
- Heater/Reservoir Bath: I fabricated this from 1″ schedule 40 PVC with a T and elbow to 1/2″NPT threaded fittings and then used NPT nylon barbs to connect tubing. The cap I had to use a 1-1/4″ threaded cap and turned it in my lathe to fit the exterior of the 1″ T fitting.
- Insulation was a combination of 1″ and 1/2″ pipe insulation cut to fit.
- Tubing is 3/8″ ID x 1/2″ OD tubing. Koolance 3/8″ barbs are also used as the test standard for thermal testing.
- Case Material – 1/2″ x 8″ Pine it’s a little over 24″ wide to barely fit 4x140mm rads. The bench overall height is about 18″.
- Inlet port was fabricated from a 4″ flange material making an ID opening of 4.540″ Diameter. I shaved the threads out of it and tapered the inlet on the lathe for a smoother inlet.
- Acrylic panels are all .100″ thickness and I dado cut slots into the shelves for them to fit into.
- Current Fan Controller is a Scythe Kaze Master, fed by a Koolance SPD-24 to slightly overvolt fans to 2200RPM.
- Heater Control is via a 3A generic Variac, although 5A would be better for higher heat loads. I am using the variac to dial in the Watts into the variac to 100W, 200W, and 300W for each RPM. 5Watts is then added for the pump heat minus variac heat, so each test is targeting approximately 105W, 205W, and 305W.
- For Watt Metering – P3 Kill-A-Watt and just manually observing and correcting wattage is used. Wattage normally does not vary by more than 2-3 Watts.
- Fans- Titan Kukri H PWM fans – I picked these fans because they had a good RPM range and I thought would better represent 25mm fan performance over using 38mm fans. They do have a more dense 9 blade fan design similar to the Gentle Typhoons. I’m not using them because I think they are superior in noise over other fans, I am simply using them because I can run them from about 650RPM clear up to 2200RPM to get a good broad range of RPM levels tested with a single fan type. They seem like a pretty good fan, but I wouldn’t suggest they are superior over other 25mm fans.

Wow, that’s impressive, while most of the rads tend to favor performance in one specific area, the UT60 is extremely strong across the range and more of an all around extreme performer. While differences between many are very small, the SR-1 was the only one to show a slight lead over the UT60 and that is only for extremely slow RPM ranges. The UT60’s thermal performance is nothing short of amazing!
Summary Bar Chart

Even after testing 12 radiators, the UT60 holds the lead most of the mid to upper fan speed area, excellent!


As a follow up …..
http://www.madsci.org/posts/archives/dec2001/1009665868.Cs.r.html
“Lets say that you have three servers, each consuming 300 watts of
electricity, and a monitor producing 100 watts of heat. So the heat load
from the electrical equipment is 1000 watts total. Now it is true that a
few watts of power exits the room in the wires, but essentially all of the
electric power that the equipment uses is converted to heat.”
See also http://www.overclock.net/t/126350/heat-from-cpu-w
Finally, you’ll also note that plugging anything into the PSU calculator
http://extreme.outervision.com/psucalculatorlite.jsp
and for determining ya heat sink size
http://www.extreme.outervision.com/tools.jsp
give the exact same number
It is true that there are heat losses through the motherboard, but there are also significant cooling losses with heat recirculation, dust, and air flow restrictions. Even more important is the actual loading in real world use vs. synthetic testing. Any amount of heat/cooling planning is good enough, don’t get too caught up in the being exact.
Certainly true …. I make that point often over on OCN….. I always calculate the numbers and find out what the math says should be the number. That’s the idealized goal :)…… then we get to reality. As I remember the radiator sides / shroud is not accounted for in your graphs….there’s also the res, tubing, fittings all of the components, backplates, etc.
So once I get the number for the 10C “ideal”, I look at what can realistically fit and look for the most cost effective / realistic alternative. In my current build, at 1250 rpm, I can exhaust 62% of the idealized heat output thru my rads…. this creeps up to 87% at 1800 rpm. At real loads….I think I’ll ever tax even the 62% loading.
Hi Martin,
I am going to build a liquid cooled rig, let’s say with several 7990.. It is better to rely on a single massive radiator like the mora 3 pro (9-18×140) or several smaller like this one, the ut60 360?
I guess the first solution is the most efficient, but the second one has the advantage to insert a radiator between some gpus..
What do you think?
Hi Martin.
Considering the very linear curve of the UT60 performance, would it be fair to assume 600 w ofcooling at 4000 rpm (twice that at 2000 rpm and 4 times that at 1000 rpm) ?
Hi Martin.
Considering the very linear curve of the UT60 performance, would it be fair to assume 600 w ofcooling at 4000 rpm (twice that at 2000 rpm and 4 times that at 1000 rpm) ?
And what cooling power would you estimate at 4000 rpm in push/pull?
I can’t imagine being in the same room w/ 4000 rpm fans, I can’t be in the same room with a H100 and it’s 2,600 rpm fans due to the 60 dBA vacuum cleaner level equivalent sound.
As for cooling, martin would know better but from my experience with building ventilation, after a certain point, the air resistance and turbulence brings a point of diminishing returns.