• Home
  • blog
  • Measuring Heatpipe Efficiency

Measuring Heatpipe Efficiency

blog image

A new contributor reports on his efforts to determine the efficacy of the Borg 6mm HES (Heatpipe Extension Set) by mCubed, designed for use in their passive home theather PC cases. The author’s aim is to design a secondary backup system to aid a 1U HSF in cooling the CPU in an extremely low profile media PC case. He created a testing rig that’s worthy of the SPCR lab, one that allows the thermal loss factor of this heatpipe system to be easily seen. The results are applicable mostly to the mCube heatpipes and mounting system but also gives insight into the kind of results that can be expected from DIY heatpipe setups.

August 8, 2006 by Brendan Wynn

A new contributor to SPCR reports from New Zealand on his efforts to determine the efficacy of the Borg 6mm HES (Heatpipe Extension Set) by mCubed, which is designed for use in their passive home theater PC cases. Brendan’s aim is to design a secondary backup system to aid a 1U HSF in cooling the CPU in an upcoming extremely low profile Hiper Group media case. In the process, he created a testing rig that’s worthy of the SPCR lab, one that allows the thermal loss factor of this heatpipe system to be easily seen. The results are applicable mostly to the mCube heatpipes and mounting system. We simply do not have experience with a big enough sampling of heatpipes and heatpipe mounting systems to generalize. However, the article gives insights about the kind of results that can be expected from DIY heatpipe setups.

– Mike Chin, Editor

INTRODUCTION

This article was written after I became interested in ways to transport heat around SFF (small form factor) and slim cases.

My current project is a HTPC using the slimmest case I can find. At this time, the lowest profile case on the market is the Hiper Group media case, which is a mere 55mm in height! Hiper will be releasing a new version soon, and it is this media case that I have my eye on. While waiting, and after reading reviews of the current Hiper case, I decided to look into ways to augment the cooling of the 1U CPU HSF (heatsink fan) supplied as an accessory with the Hiper Media case.


Editor’s Note: The author awaits the release of the Hiper HMC-2x53x.

Today’s CPU air-cooling systems almost always use heatpipes to transport heat away from the CPU, normally into the case where it can be vented outside. Deciding this would be an effective mechanism to investigate further, I sourced some heatpipes to try them in various configurations. In scouring the Net for a source of heatpipes, I came across the mCubed Borg passive case heatsink system. Figuring I had time before the new Hiper case became available, I embarked on testing the effectiveness of the Borg system and perhaps making modifications/additions to incorporate it into a slim HTPC system.

The following is a pictorial rambling of my efforts to test and design a slim HSF heatpipe CPU cooling system.

Editor’s Note: What is a Heatpipe? (explanation from wikipedia)

A typical heat pipe consists of a sealed hollow tube. A thermoconductive metal such as copper or aluminium is used to make the tube. The pipe contains a relatively small quantity of a "working fluid" or coolant (such as water, ethanol or mercury) with the remainder of the pipe being filled with vapour phase of the working fluid, all other gases being excluded.

On the internal side of the tube’s side-walls a wick structure exerts a capillary force on the liquid phase of the working fluid. This is typically a metal powder sintered or a series of grooves parallel to the tube axis, but it may in principle be any material capable of soaking up the coolant.

Heat pipes contain no moving parts and typically require no maintenance, though non-condensing gases that diffuse through the pipe’s walls may eventually reduce the effectiveness, particularly when the working fluid’s vapour pressure is low.

The materials and coolant chosen depends on the temperature conditions in which the heat pipe must operate, with coolants ranging from liquid helium for extremely low temperature applications to mercury for high temperature conditions. However, the vast majority of heat pipes uses either ammonia or water as working fluid.

The advantage of heat pipes is their great efficiency in transferring heat. They are actually a better heat conductor than an equivalent cross-section of solid copper. The general principle of heat pipes using gravity dates back to the steam age.

Additional sources of information about heatpipes:
http://www.heatpipe.com/HomePage/abouthpt/heatpipes.htm
http://www.thermacore.com/thermal-basics/heat-pipe-technology.aspx

TEST RIG

The first thing I needed was a test-bed. It had to include a heat-source (CPU Simulator) and method of measuring "CPU" temperature. So I made a CPU Simulator around a 100Watt, 2.2 ohm aluminium-encased resistor. I mounted it in a length of aluminium channel and encapsulated it in an epoxy potting-resin (photo is pre-potting).


CPU Simulator: 100Watt, 2.2 ohm aluminium-encased resistor


M3 screws to attach the CPU-Simulator to a HSF, in this case a Thermaltake Pipe101.

I also encapsulated a NTC-resistor to enable temperatures to be monitored. This NTC-resistor was mounted in the centre of the rig, which meant it measures the highest possible temps. I also soldered in a feedback cable so the voltage across the resistor could be measured here rather than at the PSU to avoid any potential voltage drop. [Editor’s Note: A Negative Temperature Coefficient (NTC) resistor is a temperature-dependent resistor, commonly called thermistor, with a negative temperature coefficient. When the temperature rises, the resistance of the NTC resistor drops. They are often used in temperature detectors and measuring instruments.]

To make the CPU-Simulator useful I needed to make a test-bed. This was done using a laminated board, multimeter, PC PSU and cabling.


The test bed.

I wired the HS fan to a switch, enabling easy selection of 5V or 12V fan voltage. This meant I could test various configurations with different cooling capacities.

With the resistor connected to the PSU 5V rail, the 2.2 ohm resistor becomes an 11W heat source. With it connected to the 12V rail, it turns into a 59W heat source. The voltages were actually slightly less than 5 and 12 volts, as there was voltage drop across the supply cable — hence the odd wattage values.

ESTABLISHING REFERENCE VALUES

Now that I had a test-bed, I set about making measurements using the Thermaltake Pipe101 as the reference HSF. These would become the reference values for comparing the heatpipe results against.

















CPU Simulator Reference Data: Thermaltake Pipe 101



Fan



Sound



Heat



Max Temp.



Ambient



Temp. Rise


°C / W


12V

Loud

59W

32.5°C

19°C

13.5°C

0.2288

5V

Very quiet

59W

43.5°C

19°C

24.5°C

0.4153

Notes: Arctic Silver 5 TIM was used at the heatsink / simulator interface. Temperatures were recorded after the simulations had run for 48 hours

The values in the right columns are the most important, as they are the temperature rise above ambient and the resulting calculated cooling capacity (°C/Watt). As can be seen, with the fan at 12V we have a temperature rise of 13.5 °C, and a rise of 24.5°C with the fan at 5V. So with these reference temps I now had something to compare the effectiveness of the Borg heatpipes against.

BORG HES 4-PIPE SYSTEM

Previously I had ordered a Borg 6mm HES (Heatpipe Extension Set) kit:


mCubed Borg 6mm heatpipe extension set

THE REAL TEST

It was now time to configure the test-bed to utilise these mCubed Borg heatpipes. I added another mount system using aluminium channel. The channel was cut and drilled to function as a bracket to hold the plates against the heat/cooling sources. To insulate the plates from the channel I used cardboard squares, quite remarkably these proved to be very effective insulators (the brackets did not heat-up at all).

I used the supplied mCubed heat-transfer compound for coupling the pipes to the heat-plates, and Arctic-5 for coupling the plates to the CPU Simulator and Pipe-101 HSF.


mCubed heatpipes set up with heating and cooling blocks (evaporator and condensor).

And then the HSF was bolted on tightly, completing this configuration:


The Thermaltake HSF was used again.

TEST RESULTS

I again ran the same tests as before. This time the heat would be removed from the CPU Simulator via the heatpipes, then dissipated into the air by the Thermaltake HSF as before. In essence, the only real change is the addition of the heatpipes in the heat transfer path, although it is far from perfect; there are small unavoidable losses in the interfaces at both the evaporator (where the CPU heat evaporates the coolant in the heatpipes) and at the condenser (where the HSF condenses the coolant back into liquid form).


mCubed 4-pipe HES w/ Thermaltake Pipe 101 fan on CPU Simulator


Fan



Sound



Heat



Max Temp.



Ambient



Temp. Rise


°C / W


12V

Loud

59W

48°C

21°C

27°C

0.4576

5V

Very quiet

59W

58°C

21°C

37°C

0.6271

12V

Loud

11W

25°C

21°C

4°C

0.3636

5V

Very quiet

11W

27.5°C

21°C

6.5°C

0.5909

Off

Silent

11W

37°C

21°C

16°C

1.4545

Reference Results: Thermaltake Pipe 101 directly on CPU
Simulator

12V

Loud

59W

32.5°C

19°C

13.5°C

0.2288

5V

Very quiet

59W

43.5°C

19°C

24.5°C

0.4153

In comparing the results between direct HSF cooling and heatpipe-coupled HSF cooling, there is degradation of 13.5 °C for the HSF in high cooling mode (fan at 12V) and 12.5°C with the HSF in silent mode (fan at 5V). This drop in performance was greater than I expected. Perhaps there was too much thermal loss in the mechanical interfaces between the pipes and the blocks? Just to double check, I reassembled the plates, heatpipes and HSF and ran the test again, but there was no change.

At this point, I wondered what the effect of less power would be. As can be seen in the table above, with the CPU simulator at 11W, the heatpipes easily transport almost all the heat to the HSF.

mCubed claims that with the Borg system, "4 heatpipes can transport up to 120W to the heatsink". Although my four-heatpipe test results suggested the heatpipes were near capacity, this spec indicated that two heatpipes could actually carry the test-rig’s 59W. To test this, I removed the two outer heatpipes and retested. Here is a photo of the new configuration:


Just two pipes this time.


mCubed 2-pipe HES w/ Thermaltake Pipe 101 HSF on CPU Simulator


Fan



Sound



Heat



Max Temp.



Ambient



Temp. Rise


°C / W


12V

Loud

59W

>80°C

17°C

>63°C

n/a

5V

Very quiet

59W

n/a

17°C

n/a

n/a

12V

Loud

11W

25°C

21°C

4°C

0.3636

5V

Very quiet

11W

28°C

21°C

7°C

0.5909

Off

Silent

11W

37°C

21°C

16°C

1.4545

Notes: Arctic Silver 5 TIM was used at the heatsink / simulator interface. Temperatures were recorded after the simulations had run for over 36 hours

The temperature kep climbing past 80°C
with the HSF at 12V at 59W heat load. There was no point in testing cooling performance with the HSF at 5V. I also tested with reduced CPU-Simulator power (by reducing the source voltage to 5V). It’s pretty obvious that a two-heatpipe configuration is not up to moving ~60Watts efficiently. I would estimate from these results that the two-heatpipe configuration is good for 40W maximum. *(See Final Editor’s Note below.)

CONCLUSIONS

From my experience in working with and testing the Borg Heatpipes, I came to the following conclusions:

  • The heatpipes are very soft and easily bent into position (see photo below).
  • Adding an additional right-angle bend to each of these two heatpipes had absolutely no effect on the efficiency of their heat transport.
  • These heatpipes work! I would not however try to move 30W of energy per pipe that the specs indicate as their capability. Rather, I would suggest halving the heat to 15W per pipe in order to keep the CPU temperature to a reasonable level. The heatpipes and blocks must be well "connected" mechaniically to both the CPU and cooling mechanism (eg: HSF or Radiator Block).
  • Given their form and flexibility, these heatpipes are great for redistributing CPU heat.
  • For a slim low-profile case, I expect the best configuration would be the standard 1U HSF, but with cooling assistance via heatpipes and a secondary cooling system – this will be my next project!


The extra bend had absolutely no effect on the efficiency of either heatpipe’s heat transport.

* * *


*Final Editor’s Note

When thinking about heat transfer, it’s useful to consider thermal pathways, points of constriction, and the speed of heat transfer. The fact is that a heatpipe or any thermally conductive path can transfer even thousands of watts — if the pathway is narrow, it will just take longer, much like dial-up versus DSL. This is a key point. If the heat is transferred too slowly, then it will build up at the source and cause temperature rise at the source, in this case, in the CPU. If the heat builds up too much, the CPU will overheat. So mCube’s assurance that their 4-heatpipe system will transfer up to 120W is rather misleading, because it’s subject to what happens outside the heatpipe and probably applicable only in the context of their own cases.

From a practical standpoint, there are several issues around the use of heatpipes to cool a CPU:

  • The heatpipes can only transfer heat from one end of the pipe to the other. You have to get the heat into one end and get it out of the other end. So there are three points of potential thermal loss: At the hot end, through the pipe itself, and at the cooling end.
  • Assuming that ALL of the heat from the source is effectively transferred to one end of the heatpipe, the speed of the heat transfer depends not only on the efficacy of the heatpipe’s internal vapor-change action, but also on what happens at the other end. How big a cooling surface area that end is connected to, how well it’s joined, and how much airflow there is — all of these factors matter.
  • Thermal transfer can be bottlenecked at the heat source, through the heatpipe, and at the condenser end. Any of these can hurt cooling performance.
  • In Brendan’s test, both the heat source and the Thermaltake heatsink/fan are intrinsic to the test. Changing the HSF could have a significant effect on the measured °C/W
    value of the system and the apparent heat transfer of the heatpipes, just like lowering the heat at the source did. Because a heatsink with greater heatsink area (say, the Scythe Ninja or Thermalright Ultra-120) was not tried, we can’t know for sure whether the heatpipes or the HSF was the greater bottleneck. It seems logical to believe that a bigger HSF would have given better results with the heatpipes. The test results, then, are most pertinent when the cooling surface area and the airflow aross it at the cooling end is about the same as the Thermaltake HSF.

The author’s conclusion about keeping the thermal load to 15W per pipe still seems prudent. DIY PC silencers should generally exercise a bit of caution about cooling.


Other SPCR articles of related interest

Fanless Heatpipe CPU Cooling System by FMAH
Boxing & Watercooling to Silence

Zalman TNN-300 Fanless PC Enclosure System
Fanless Ultra Powerhouse PC by EndPCNoise

* * *

Comment on this article in the SPCR forums.

About the Author

Location: Auckland, New Zealand

Day Job: Account Manager – Diagnostic Medical Imaging (eg: CT, MR, X-ray, etc)

Hobbies: Breakaway PC’s. By this I mean anything that is outside the current mainstream. Once I had built a few ‘normal’ PCs, I started overclocking > which lead me to cooling and silencing > which lead me to SFFs > which lead me to cooling and silencing > which lead me to HTPCs > which lead me to slim-HTPCs > which lead me to cooling & silencing (again!).

Personal Spreel: I started life as an Aviation Technician which is where I gained my quals and experience in electronics. Although I consider myself very young 😉 back when I was learning the ropes, computers were room sized for commercial/security applications and we were calling PC’s an AT or 286. After Aviation, Diagnostic Imaging (X-ray machines back then) found me and since then I have been either fixing these systems or more lately selling them. At one point I started a PC company in order to buy from PC components from distributors/agents and to fund my hobby. This went ok but at tax time I had to spend too much time at the books, so when we left for Australia (job promotion) I left the company behind with a mate.

In the last two years, I have been following HTPC’s (with all the associated hype) and somewhere along the way decided my HTPC should be no larger than my Tivo or DVD player. This brought me back to PC cooling and silencing which has always interested me. By combining slim-HTPCs with my passion for cooling and silence I have a challenge akin to shoving a model ship into a glass bottle (there are similarities: they both have impossibly small "containers", drive you crazy trying to get it right, and look great in the living room!).

Currently I am waiting for Hiper Group to release their new Media Chassis. Whilst this happens I have delved back into – yep – cooling & silencing! And this has of course brought me to the good folks of SPCR!!

Leave a Comment

Your email address will not be published. Required fields are marked *