Should I Use Heat Pipes or Vapor Chambers to Cool My Electronics Application

Should I Use Heat Pipes or Vapor Chambers to Cool My Electronics Application

I’m often asked which two-phase cooling technology, heat pipes or vapor chambers, is best suited for a particular electronics application. While there are some good rules of thumb the answer isn’t always simple. Nonetheless, this blog post will cover the basics with additional information covered in ann online webinar found in the Resources section of this site.

Let’s start by saying that a transition from traditional heat sinks to either type of two-phase technologies should only be considered when the design is conduction limited and/or when non-thermal goals such as weight or size can’t be achieved with other materials such as solid aluminum or copper.

Once it’s determined that traditional heat sinks just won’t meet thermal requirements, how do mechanical engineers begin to determine the next best option?

Are You Moving or Spreading the Heat?

While there’s no hard line of distinction, think of the difference like this.

moving heat use heat pipe spreading heat use vapor chamber

 

Space constrained applications and/or those with two or more heat sources sharing a common condenser (fin stack) are good candidates for moving the heat to a remote location of more than around 40mm distance. In these cases, the design flexibility of heat pipes, by virtue of their bendability in all directions, are the default choice. And this is especially true if only a single heat pipe is needed. For reference, a 3mm sintered metal heat pipe using water as the working fluid is adequate for cooling 5-20 watts of power while an 8mm pipe can carry 20-90 watts. The range is affected by the porosity and thickness of the wick structure, the amount of working fluid, the degree of bending required and the amount of flattening done to the round pipe.

In my experience heat pipes are used in 95%+ of applications where heat needs to be moved to a remote condenser. However, when conduction loss needs to be reduced further, such as in the laser diode example below, vapor chambers become a viable option. Here three heat sources share a common condenser, reducing temperature rise by roughly 15% over a similar heat pipe alternative.

vapor chambers cool high power laser

450W RGB Laser Diode Solution for a 3D Projector

 

 

When it comes to spreading heat to a local condenser, the choice between heat pipes and vapor chambers becomes a lot more complicated….

Heat pipes are probably a best choice if:

  • There’s plenty of air flow
  • You’ve got lots of room for fins
  • Nominal power densities are <25 w/cm2
  • Ambient temperatures are normal – let’s call this below 45oC
  • Cost is a key consideration – every penny counts!

A vapor chamber should be considered if:

  • Power densities are high – certainly by the time they hit 50 w/cm2
  • Reducing hot spot across the die is a key concern
  • Z-direction (height) is constrained, yet fin area needs to be increased
  • Atypical ambient temperatures and/or low air flow
  • Performance is a key consideration – every degree counts!

Before I review a couple of examples, it’s important to note some of the reasons vapor chambers can perform better than heat pipes for spreading heat. First, vapor chambers make direct contact with the heat source, whereas heat pipes usually require a solid base plate increasing conduction loss –from the base plate itself and from an additional TIM layer between the base plate and the heat pipes. Second, heat spreading via vapor chambers is multi-directional while with heat pipes it’s linear. Third, a vapor chamber solution often allows for additional fin area. These advantages typically allow vapor chamber solutions to have between 10-30% better performance than their heat pipe counterparts. This translates to around 3-9oC for most applications.

Despite the performance difference, vapor chambers were regarded as a very niche solution due in large part to their cost premium over heat pipes, especially for non-consumer (low volume) applications. But manufacturing innovations from an increasing number of suppliers helped narrow the gap.

Let’s take a look at a couple of examples. The first is for a telecom infrastructure manufacturer that wanted to understand both the cost and performance difference between competing heat pipe and vapor chamber designs. The heat source was in the 85 watt range with ambient temperature of 55 degrees Celsius. We compared a four 8mm heat pipe solution to a single vapor chamber design. As the image below shows, the former required a base plate as secondary machining of the heat pipes was not possible due to the wall thickness of the pipes. Testing showed the vapor chamber design to outperform the alternative by a full 4oC, allowing the assembly to meet specifications.

heat pipes vs vapor chamber

 

The second example is from one of the first two graphic card solutions to use vapor chambers. Increasing GPU power required that the current heat pipe based solution be redesigned while maintaining the heat sink stack height. Here two 8mm pipes were replaced with a vapor chamber. Elimination of the base plate allowed us to increase fin height and overall fin area while the better heat spreading of the vapor chamber all contributed to a 6oC better solution.

vapor chamber vs heat pipes

How Does Bending Affect Heat Pipe & Vapor Chamber Performance?

How Does Bending Affect Heat Pipe & Vapor Chamber Performance?

The majority of applications for heat pipes, and to a lesser extent vapor chambers, require these products to be bent. Below are respective examples for high-performance graphics card, semiconductor equipment and networking applications.

Bent Heat Pipes and Vapor Chambers

Figure 1: Heat Sinks Using Bent Two-Phase Devices

 

But what effect does bending a heat pipe or vapor chamber have on their performance? To answer this, let’s first talk about bend parameters then we’ll move into the meat of it and discuss both evaporator and vapor resistances.

Although smaller bend radii are possible, heat pipe guidelines almost universally put their c/l (centerline) bend radius at 3X the diameter of the heat pipe being bent. In other words, for a 5mm round heat pipe bent into a U shape, the resulting OD would be 35mm.

Bend radius for a 1.5-3.5mm thick vapor chambers are about 10mm. For instance, for a 2mm thick vapor chamber the OD of a 180 degree bend would be 22mm. While these would have to stamped into place for a traditional two-piece vapor chamber, one-piece designs can be bent post production, even into a U-shape.

Figure 2: Bending Examples

Figure 2: Bending Examples

 

Two areas of thermal resistance in these devices must be closely examined. The first is evaporation resistance which is the deta-t due to conduction through the device wall and wick structure as well as phase change of the working fluid into vapor.

In the vast majority of applications, the evaporation resistance is the dominate resistance; therefore, making these devices somewhat length independent. This means that a two-phase device with a transport distance of 75 mm will have almost the same Tsource -Tsink as one with a 150 mm transport distance. This, in effect, doubles the effective thermal conductivity for the longer devices. For a typical water/copper heat pipe with 0.5mm wall and 0.4mm sintered wick, thermal resistance has a nominal value of about 0.1 oC/w/cm2

Figure 3: Heat Pipe

Figure 3: Heat Pipe

 

The second is vapor transport resistance which is the temperature loss in the transport region due to pressure drop and condensation delta-t’s. This resistance is typically a function of the power density in the vapor space – nominal power densities are 300-400 w/cm2 with a typical maximum of 800-1,000 w/cm2. The nominal thermal resistance of vapor transport is about 0.01o C/w/cm2. Because of the correlation between cross sectional area of the vapor space and vapor resistance, smaller diameter heat pipes or those that have been flattened increases vapor flow thermal resistance.

So how are these two key resistances affected by bending a two-phase device?

Because the evaporator is almost never placed at the arch of a bent two phase device, we would expect relatively little to no increase in its nominal value when comparing straight and bend two-phase devices. When testing a 3mm thick U-shaped vapor chamber with a 10mm bend radius we see this to be true. Test data showing the evaporation resistance before and after bending is shown in Figure 4. The results are identical within measurement error.

FIgure 4: Evaporator Resistance

Figure 4: Evaporator Resistance

 

Now let’s look at vapor transport resistance. Due to the pressure drop induced by bending the vapor chamber, we would expect vapor transport resistance to increase due to an increase in pressure drop, thus decreasing the thermal conductivity of the device. Again, testing supports this claim and is relatively consistent between heat pipes and vapor chambers. Bending a two-phase device 180 degrees increases vapor flow resistance by around 50%.

Figure 5: Vapor Resistance

Figure 5: Vapor Resistance

 

It’s important to note that while vapor transport resistance is significantly affected by bending, its relative contribution to overall thermal resistance is often small. Remember that for an un-bent part, evaporator resistance is a full order of magnitude greater than that of vapor flow resistance.

So, when we examine the overall thermal resistance, we see that the net effect is between 18-40%. You will also note a decrease in the total power. When relying on capillary pumping in the wick structure to return the fluid the bending affects the pore radius and porosity in the bend region. This has about a 10% effect on the Qmax post bending.

Figure 6: Total Thermal Resistance

Figure 6: Total Thermal Resistance

 

Given that one should generally design a two-phase device to operate at 70% of its Qmax (we’ll use total power of 70W), we see that the thermal resistance of an un-bent device from 0.047 degrees c/w (3.3 °C) to 0.063 (4.4 °C) for a device with a U-shaped bend. This translates to a delta-T of only 0.9 degrees Celsius.

Based on this we can extrapolate some rules of thumb.

  • Spec a straight heat pipe or vapor chamber with 30% thermal safety margin.
    • Example: A lead load of 70w should use a heat pipe designed with a Qmax of no less than 91w.
  • Add total bend radius of the heat pipe/VC. While not perfect this will get you very close to actual.
    • Example: one 90 degree bend and another 45 degree bend = 135 degrees of bend
  • For each 10 degrees of bend Qmax will decline by .56%.
    • In our example from above: 135 degree total bend divided by 10 multiplied by 0.56% = 7.6% decrease in Qmax.
  • So for our 70 w heat source with two bends totaling 135 degrees we’ll need a heat pipe with a Qmax of 70*(1+(.3+.076)) = 96.3w.
Heat Pipes and Vapor Chambers – What’s the Difference?

Heat Pipes and Vapor Chambers – What’s the Difference?

At some point in a thermal system design project it may become apparent that the tried and true methods of increasing thermal efficiency – solid base, fin & fan – just aren’t sufficient. Reasons include:

  • Keep out zones prohibit a larger heat sink (thicker base, added fin area, etc.)
  • Enclosure size and/or airflow can’t be increased.
  • Transitioning to a solid copper heat sink, in whole or in part, adds too much weight and in some cases too much cost.
  • Component power /density necessitates heat be moved to a remote location more than 40-50mm away from the heat source.

Regardless of the reason, most thermal engineers are going to need a two-phase cooling solution using either heat pipes or vapor chambers on numerous projects with which they are involved. But, which one is likely the best choice? In today’s blog I’d like to do a topline overview of structural differences and thermal design considerations between these very similar yet somewhat unique two-phase devices.

How does a heat pipe work

 

It probably goes without saying, but the operating principles of all two-phase devices are identical. A wick structure (sintered powder, mesh screens, and/or grooves) are applied to the inside walls of an enclosure (tube or planar shape). Liquid (usually water) is added to the device and vacuum sealed at which point the wick distributes the liquid throughout the device. As heat is applied to one area, the liquid turns to vapor and moves to an area of lower pressure where it cools and returns to liquid form whereupon it moves back to the heat source by virtue of capillary action. In this sense, heat pipes and vapor chambers are the same thing.

For simplicity’s sake, I’ll be focusing on the most common type of two-phase device: an all copper vessel, using a sintered copper wick structure with water as the working fluid.

Structural Design & Cost

 

Difference between heat pipes and vapor chambers

 

Thermal Design Considerations

Heat Pipes

For decades, heat pipes have been the default two-phase device of choice for thermal engineers due largely to the cost difference relative to vapor chambers. They were used both for heat transport, for which they still have an advantage, and for heat spreading, typically using multiple pipes in close proximity to one another. For lower power applications, perhaps requiring only a single, small heat pipe, or those where heat must be effectively transported, heat pipes still maintain dominance due to their low cost and design flexibility.

Heat pipes cooling notebook pc

Heat Pipes Cooling Notebook Computer (Wikipedia)

 

Greater total power and power densities eventually led to multiple heat pipes being used to solve the thermal challenge. Both images below are heat sinks for a small form factor, high performance desktop PC. The one on the left uses a copper base plate in between the heat source and the heat pipes, as is common with heat pipe applications (indirect contact). As processor heat increased in the subsequent generation of this product, the company encountered thermal issues, but did not want to radically redesign the thermal solution as can be seen on the image to the right. Here a vapor chamber replaced the copper base plate, spreading heat more evenly across the heat source and transferring it more effectively to the heat pipes. This is a great example of how both types of two-phase devices can be used together.

Heatsink using heat pipes and vapor chamber

Heat Pipe Heat Sink – (L) Using Solid Copper Baseplate, (R) Using Vapor Chamber (ixbtlabs.com & Celsia)

 

A potential alternative to this problem might have been to implement what some call ‘direct contact’ heat pipes. But, this solution has its drawbacks as well. As seen below, this design option uses slightly flattened and machined heat pipes cradled in an aluminum mounting bracket to make direct contact with the heat source. While eliminating the base plate and additional TIM layer – decreasing thermal resistance – it doesn’t spread the heat as effectively as a vapor chamber solution.

Heat sink with flat heat pipes

Direct Contact Heat Pipes (Silverstonetek)

 

Traditional (2-Piece) Vapor Chamber

Most manufacturers of vapor chambers use a traditional two piece design. While studies and practical application shows that the performance of heat sinks using vapor chambers can be enhanced by 20-30% over their heat pipe counterparts, a two piece design has cost implications of roughly the same magnitude versus a multiple heat pipe configuration. Nonetheless, vapor chamber usage has grown with the increasing power and power densities of today’s devices.

Because they do an incredible job of spreading heat, allow for low profile heat sinks, can be made into virtually any shape, embossed, and make direct contact with heat source, these devices are used in a wide variety of higher power applications. Below are two examples of heat sinks using two-piece vapor chambers.

vapor chamber heat sinks

Traditional Two-Piece Vapor Chamber (anandtech.com & Celsia)

As mentioned earlier, the increased cost of this design sometimes limits its incorporation into thermal solutions. Another potential drawback is that there’s little design flexibility in the z-direction. Making a U-shape for instance, while conceivably possible, would be impractical from a manufacturability/cost perspective.

Hybrid (1-Piece) Vapor Chamber

Available from a growing number of manufacturers, one-piece vapor chambers are a cost reduced version of their two-piece counterparts, yet maintain the thermal performance characteristics while adding some unique capabilities (e.g. U-shape bending). Like heat pipes, a one-piece product begins its life as a single copper tube, hence the 1-piece moniker. Like traditional two-piece designs, one piece vapor chambers make direct contact with the heat source, have a multi-directional heat flow, and can support clamping forces of up to 90 PSI. But they’re less expensive to produce because they require less tooling, don’t use individual support posts, and don’t have to be welded on all four sides. Below are a few examples of one-piece vapor chambers.

Inexpensive vapor chamber heat sinks

Hybrid One-Piece Vapor Chambers

 

The thing to remember about two phase devices is that heat pipes favor moving heat over spreading it, while the reverse is true of vapor chambers. For sure, there are numerous thermal challenges where either could be used with good results so it’s important to do a thorough review process of both designs for settling on one.

Two-Phase Thermal Review Process

(I’ll be presenting a detailed version at Semi-Therm next month)

A typical, although not ideal, thermal design scenario is one in which several key variables are already defined. These include enclosure shape / size, component layout (with associated keep out zones) and airflow, as well as the total power, power density, and size of the heat source(s).

Step #1 – Start by looking at fin location

Given these constraints, the focus is on the heat sink itself and one should start by understanding if the condenser is remote or local to the heat source.

  • If remote by more than 40-50mm begin the investigation with heat pipes. Design flexibility is high allowing bending and flattening to conform to almost any shape in all three dimensions.
  • If local and copper alternatives have been ruled out then a vapor chamber solution that spreads the heat is the best starting point provided the perimeter ratio of the heat sink to the heat source is greater than 30:1.

Step #2: Run an Excel model to determine heat exchanger performance

Based on this input a simple excel model should be run to determine the performance of the heat exchanger. This tells us how much of the thermal budget – fin to air and air temperature rise – are being used. This provides information on how much is left for conduction and interfaces. Since these two components, fin to air and air temperature rise, are the largest resistances in the system, the tail that wags the dog, a design review at this point is normal to optimize the area for the fins and the air flow and pressure drops.

  • If the remaining thermal budget for conduction is less than 10oC, a look at copper or heat pipes. For small devices at low powers a single heat pipe is often sufficient. With the total power we can estimate the number of heat pipes required to carry the heat. For example, a single 6mm heat pipe can carry the power from a 45 watt device.
  • If it is less than 5oC then a vapor chamber may be required. For small devices at high powers typically a vapor chamber is the best solution. Vapor chambers, on the other hand, generally are not run to their performance limits so they are sized to cover as much of the base of the heat sink as possible. Due to their flat format there is a direct contact between the VC and the heat generating component.

Heat sink performance basic model

 

The above image is the summary page of such a model for an LED application. In this example two 8mm heat pipes were compared against a single 15mm wide vapor chamber of similar cost. Each ‘case scenario’ represents the use of different length fins. As you can see, the vapor chamber solution provides an additional 4-5 degrees Celsius of thermal headroom, but this delta is often higher.

Step #3 – Run a more sophisticated Excel model and CFD analysis to optimize the design

Available soon on Celsia’s website, but also downloadable from several other sources, a more comprehensive heat pipe and vapor chamber modeling solution will aid in refining the two-phase design as it accounts for changes to wick characteristics, vapor space, wall thickness, working fluid, case metal, and orientation. Subsequently, CDF modeling is often used to determine the performance of variations to the full heat sink assembly. However, sometimes the best use of time and money is simply to prototype and test a couple of thermal module iterations.

Heatsink performance advanced model

 

Please contact us, if you’d like to learn more about how Celsia can help with your next heat sink project. We’ve worked on everything from consumer devices to industrial test equipment that require heat sinks to cool anywhere from a few watts to a few kilowatts.

Celsia Achieves 1,000 w/cm2 with New Hybrid Two-Phase Design

Celsia Achieves 1,000 w/cm2 with New Hybrid Two-Phase Design

Celsia Inc. has successfully tested its passive two-phase technology to 1,000 w/cm2. Available in different form factors down to 2.5mm thick, these copper based devices are designed for applications using high power density laser diodes, power electronics, concentrating photovoltaics, or ultra-HB LEDs.

“Our goal was to push the performance limits of passive two-phase systems in order to provide design alternatives to thermal engineers needing to cool small heat sources with high power densities”, explains George Meyer, Celsia’s CEO. “Rather than having to migrate to a pumped liquid system, efficient heat spreading using our new designs can extend the limits of air cooled systems.”

Celsia developed several configurations of a copper based two-phase device with a hybrid wick structure in order to concentrate liquid flow directly to the heat source and achieve the most favorable results. Testing was conducted using a liquid cold plate attached to the devices, but the spreader performance is applicable to air cooled designs as well.

Two platforms were successfully tested to 1,000 and 850 w/cm2 power levels using a 1x8mm heater which was soldered to the devices. The testing showed stable operation up to the limits of the heat sources. Dry out power levels for these designs has not been reached yet and the test set is being redesigned to accommodate even higher power densities.

Figure 1: 8x8x100mm Two-Phase Device with Test Results

1,000_watt_per_cm2_heat_pipe

Chart_1,000_watt_per_cm2_heat_pipe

 

Figure 2: 2.5x20x50mm Two-Phase Device with Test Results

 

850_watt_per_cm2_vapor_chamber

Chart_850_watt_per_cm2_heat_pipe

 

For more information or to discuss your application please contact Celsia from its website.

About Celsia Inc.

Celsia specializes in two phase heat sink design and manufacturing for Fortune 500 and mid-market enterprise companies. Custom solutions have shipped in over two million products worldwide from our US and Asia based design and product facilities. Our goal is to provide fast, affordable, and reliable thermal solutions using vapor chamber, heat pipe and hybrid designs.

Vapor Chambers & Heat Pipes Cool Performance FBDIMMs

Vapor Chambers & Heat Pipes Cool Performance FBDIMMs


In a recent post, I talked about the use of fans and micro-thin heat pipes to cool smartphones. Today, I’d like to take you through a project Celsia tackled a number of years ago; cooling performance DDR3 ‘gamer’ memory modules using ultra-thin vapor chambers. This example should serve to illustrate the design advantages of vapor chambers over heat pipes as well as to comment on how consumer perception affects product success.

Whether it’s the CPU, graphics card or memory modules, PC gamers are notorious for demanding products that push performance limits. What’s better than being able to seamlessly run a taxing application in a higher resolution, fire an extra round of ammunition before your opponent, or having the bragging rights to the coolest looking gaming rig around. Nothing!

Mushkin approached us several years ago about creating a memory module cooler that helped gamers build this kind of machine. For most of us, bare-naked modules are just fine. Gamers’ speed requirements drive them to seek the fastest, most stable components which are then overclocked. The added heat generally requires more robust thermal solutions where every degree counts. In the case of performance memory, the most common solution is a simple aluminum spreader as seen below.

Memory Module with Aluminum Heat Spreader (Source: Kingston)

Memory Module with Aluminum Heat Spreader (Source: Kingston)

 

Attempting to further cool these modules, many of the top companies added one or more heat pipes. While visually aggressive, these coolers rely on the aluminum side spreaders to transport heat to the two phase device, limiting their thermal transport efficiency. Later models tried to solve this problem using flattened heat pipes that made direct contact with the heat source, although they appreciably increased the overall thickness of the module and didn’t entirely cover each FBDIMM. Moreover, both of these solutions increased the height of the heat sink to the point where they couldn’t be used with larger, more extravagant CPU coolers – memory slots are typically located very close to the CPU.

Heat Pipes for Cooling Performance Memory (Source: Apacer & OCZ)

Heat Pipes for Cooling Performance Memory (Source: Apacer & OCZ)

Performance Memory and CPU Heat Sink

Large CPU Heat Sink with FBDIMMs Installed (Source: overclock.net)

 

Celsia’s challenge was to design and manufacture a low profile solution (x,y,z dimensions) that increased both heat spreading and dissipation while still allowing the use of any CPU heat sink and the ability to populate every memory slot. Since the mid-2000’s, we had been working on perfecting the thermal efficiency and mass production yield rates of one-piece vapor chambers. These devices differ from heat pipes in that they can be made thinner while still allowing adequate vapor flow. They differ from the traditional two-piece vapor chamber designs due to both lower cost and reduced thickness. In either case, this new solution needed to outperform existing competitive offerings while hitting cost targets.

We modeled and turned around our first prototype within a few weeks. A couple of design tweaks later, we arrived at a solution which used two 1.5mm thick vapor chambers and TIM material sandwiched between ribbed aluminum spreaders with a small vertical fin stack. Mushin attached this heat sinks to their top of the line memory and named them “eVCI Coolers” (enhanced vapor chamber interface – oh those marketing folks). In addition to meeting all the design and performance criteria, it was the first time vapor chambers had been used to cool FBDIMMs.

Celsia Designed Vapor Chamber Memory Cooler

Celsia Designed Vapor Chamber Memory Cooler

Two 1.5mm Vapor Chambers per Module

Two 1.5mm Vapor Chambers per Module

 

Mushkin tested this heat sink against bare modules, in real-world gaming scenarios, in order to highlight the benefits. The temperature of modules without a heat sink was 43.7 degrees C above ambient while those using eVCI heat sinks measured 22.7 degrees C above ambient. In spite of achieving some really stellar thermal performance figures against many FBDIMM heat sink styles, these modules were not received well by the market. Perhaps because they were thermally cool but not visually cool, gamers opted for a more in-your-face design. Maybe elaborate heat pipe solutions just looked like they’d cool better. Market mysteries! But, it might serve as a lesson to thermal engineers who design for consumers whose devices are viewed as an extension of themselves and who revel in technical achievement as well as visual appeal.

This project is a good example of how ultra-thin vapor chambers can be used in space constrained environments. Power and power densities were low enough in this application to allow for a custom mesh wick to be used while a one-piece vapor chamber design kept the cost down. If you’d like to learn more about how Celsia can help with your next heat sink project, please contact us. We’ve worked on everything from consumer devices to industrial test equipment that require heat sinks to cool anywhere from a few watts to a few kilowatts.