The Future of AI: Powered by Efficient Cooling

Artificial intelligence large language models are now surpassing 1 trillion parameters, making next-generation technology essential. This new era of AI-driven computing prioritizes power efficiency, acceleration, networking, and storage. NVIDIA’s GB200 NVL72 addresses these challenges, but unlocking its full potential requires reliable and efficient cooling that is easy to deploy. Liquid-cooled GB200 NVL72 racks significantly lower a data center’s carbon footprint and energy use. These systems boost compute density, optimize floor space, and support high-bandwidth, low-latency GPU communication within expansive NVLink domain architectures. The GB200 delivers 25x more performance at the same power level compared to H100 air-cooled infrastructure, while also conserving water. Boyd’s modular innovative cooling solutions ensure AI-driven data centers operate at peak performance and energy efficiency, making them essential to harness the full capabilities of NVIDIA’s GB200 in an easy-to-deploy way.

Boyd: Cooling AI with Precision and Efficiency

Streamlined GB200 Integration with Boyd’s Cooling

Precision Cooling for Peak GB200 Performance

Boyd precisely tunes cooling solutions to maximize the performance of NVIDIA’s GB200. We engineer cold plates, liquid cooling loops, rack manifolds, and CDU systems to enhance the GB200 ecosystem. By optimizing performance, we ensure your data center runs at peak efficiency.

Expanding AI Capabilities: Efficient Cooling for Higher Density

Boyd’s innovative cooling solutions increase compute density without expanding a data center’s footprint. Scale up operations while minimizing space and power consumption with our energy-efficient cooling systems to boost your AI capabilities without compromising sustainability.

Pioneering AI Cooling with Boyd

Boyd and NVIDIA lead the charge in making the new era of computing cooler and more efficient. Our partnership demonstrates the commitment to deliver sustainable, high-performance cooling for next generation computing applications. Collaborate with us to leverage our rich heritage, expertise, and capabilities in developing cutting-edge systems that deliver energy-efficient and high-performance cooling solutions tailored to your needs. To learn more about our innovative cooling solutions or to discuss your project needs, schedule a consultation with our experts to see how our expertise can elevate your data center’s efficiency and performance.

Power Surge: AI Pushes Data Center Cooling to the Limit

A Multi-Pronged Approach: Cooling Solutions for Every Component

The COOLERCHIPS concept addresses all deployment levels and program objectives with innovative cold plate technology using a green refrigerant. Two-phase flow visualization techniques optimize cold plate architecture and operating conditions. CFD simulation refines flow and temperature distribution inside the immersion tray. At the rack level, an in-rack distributed pumping and flow separation system replaces the conventional cooling distribution Units (CDU). This system separates vapor from liquid and directs spent vapor back to the condensing unit to improve efficiency. Immersion manifolds connect directly to the heat rejection unit. Multiple identical racks connect to emulate an IT cluster, all linked to external heat rejection units with cool array coolers, potentially reducing the cooling tower footprint by four times. This integration reduces total power consumption to only 5% of the IT load.

Green & Efficient: Next-Gen Data Centers with Zero Water

Beyond the Benchmark: Cooler Tech Chases Evolving Needs

The program continues to advance data center cooling technology as AI-driven computing and high-performance computing (HPC) evolve, which increases the demand for more efficient cooling solutions. It refines advanced cooling technologies to optimize efficiency and responsiveness, leveraging achievements in high power density, low PUE, and environmental sustainability. This ongoing refinement ensures that the program remains at the forefront of innovation in data center cooling, meeting evolving industry needs and setting new benchmarks for efficiency and environmental responsibility.

Boyd Powers COOLERCHIPS: Cold Plate Tech Drives Efficiency

Boyd’s extensive expertise in two-phase cooling and liquid cooling systems is fundamental to the success of the COOLERCHIPS program. Our advanced cold plate technology plays a pivotal role in achieving the program’s ambitious objectives, effectively managing the substantial thermal demands of next generation data centers. Boyd’s cold plates, engineered for optimal efficiency and performance, are essential components in this effort. By incorporating Boyd’s innovative cooling solutions, the COOLERCHIPS program ensures that data centers operate not only with enhanced efficiency but also with a sustainable approach, paving the path for future innovations in data center cooling.

Collaborate with us to leverage our rich heritage, expertise, and capabilities in developing innovative liquid cold plates for advanced two-phase pumping systems, enabling energy-efficient and high-performance cooling solutions for your specific applications.

A $5 Million grant from the U.S. Department of Energy (DOE) supports NVIDIA and seven partners, including Boyd, to build an advanced liquid cooling system that enables a future class of efficient power-dense data centers. The DOE program, called COOLERCHIPS, focuses on innovative technologies to reduce data center energy consumption and environmental impact. Boyd is excited to be part of this program alongside other industry leaders in a collaborative effort to develop a new generation of energy efficient cooling solutions.

Need for Change: Advanced Cooling Solutions for Data Centers

Extreme air-cooling using two-phase technologies and current liquid-cooling systems face limitations in cooling intense high-power computing environments. As data centers continue to evolve with more computing power, system architects must innovate cooling solutions to meet increasing demands. These next generation data centers require exceptional performance, reliability, and energy efficiency on top of increased heat dissipation.

Innovative Cooling Technology: Two-phase Cooling and Liquid Cooling Systems Combination

Combining two-phase cooling and liquid cooling systems into pumped two-phase systems, or evaporative liquid cooling, leverages the benefits of both cooling technologies. First, two-phase cooling utilizes cold plates to cool chips and facilitate evaporation and re-formation of the coolant. Next, entire servers, including their lower-power components, are encased in hermetically sealed containers filled with coolant to provide a comprehensive cooling solution. With an optimized pumped two-phase cooling system, data center engineers can overcome anticipated thermal challenges for future data centers.

Innovate with Boyd: Two-phase Cooling and Liquid Cooling Systems

Boyd’s strong heritage of two-phase cooling and liquid cooling systems technologies provides a solid foundation for collaborative development. Our ability to innovate with technologies like coolant distribution units, 3D vapor chambers, liquid loops and cold plates, remote heat pipe assemblies, and chillers enables us to explore new possibilities and develop advanced cooling solutions that best fit your applications. Partner with us to tap into our strong heritage, expertise, and capabilities to develop innovative liquid cold plates for advanced two-phase pumping systems, enabling energy efficient and high performance cooling for your applications.

Although many local shops are able to manufacture a simple tubed cold plate, a properly designed and manufactured cold plate based on application will result in significantly better performance and reliability. However, assessing the quality of a cold plate can be difficult if you do not know what to look for.

How Construction Affects Tube Cold Plate Reliability and Performance

Next look at how the tubes are attached to the plate. For maximum thermal transfer, engineers should be careful designing cold plates utilizing epoxy between the tubing and the plate. Although epoxy can be used to keep the tube in place and keep thermal contact between the two components, it can also act as a thermal insulator when used incorrectly, such as being applied too thickly or not utilizing the correct epoxy.

Finally, examine the cooling surface of the plate. For demanding applications, direct contact between the tube and the component is best. This means that the tube surface must be flush with the aluminum plate. A common manufacturing technique is to put an over-sized tube into a channel and machine off or fly-cut the top. However, this costly technique, known as skim cutting, creates a tube section of varying thickness that affects structural integrity and performance.

Another alternative is to embed the tube below the surface of the cold plate and add copper inserts to level the surface. This technique is both expensive and limits performance. The cooling tube is further from the components being cooled, and the additional layer of epoxy required between the tube and the insert further decreases performance.

Most applications use a thermal interface material (TIM) between the component or board and the cold plate to help to minimize the gaps. A TIM should be as thin as possible, as the relatively high thermal resistance of the TIM greatly overshadows any conductivity improvements from having a smoother surface. Increasing the clamping force of the component or board to the cold plate can also help to offset a higher roughness, but it may increase the stress on the board or component. Clamping stress can also increase the impact of coefficient of thermal expansion (CTE) mismatches as the cold plate and component or board heat up.

To avoid galvanic corrosion, utilize the same materials, or materials with similar electrical potential, throughout your cooling loop. Ensure that the plumbing, connectors and other components do not introduce a reactive metal into the system.

In addition to these considerations, other factors of cold plate performance include: fittings, connectors, materials, fluids, channel patterns or number of passes, and the associated heat exchanger. Boyd offers a variety of options in standard sizes and custom and semi-custom designs.

The Impact of Using Different Technologies in Liquid Cold Plate Fabrication

The two biggest cost drivers in cold plate manufacturing are thermal performance requirements and annual demand, which generally thermal engineers and manufacturing engineers have little or no control over. However, you can reduce costs by understanding how roughness, flatness, hardness, surface topography, mounting features, and liquid connections specifications can all affect the cost of a cold plate. By involving your cold plate manufacturer early in the design process, you’ll be able to identify the manufacturing cost drivers and select the most cost effective design.

Most cold plates are made of aluminum but some new technologies use copper. Although copper has better thermal conductivity, aluminum is used more often because it is usually cheaper, lighter, and easier to work with. Machining copper is very difficult and expensive. If aluminum meets the thermal performance specifications, it is generally the best material to use.

Two of the most popular aluminum cold plate technologies are tubed and vacuum-brazed (See Figure 1). Tubed cold plates are usually copper or stainless steel tubes pressed into a channeled aluminum extrusion. They are cost-effective and offer good bulk heat removal for low-to-medium watt densities. Vacuum-brazed cold plates consist of two plates metallurgically bonded together with internal fin. They are available in all sizes and offer extremely high performance, making them ideal for applications where heat loads are concentrated. Labor time is limited with both tubed and vacuum-brazed cold plate technologies. For that reason, U.S. cold plate manufacturers tend to be competitive with offshore manufacturers for moderate volumes. The lower labor cost reduction from buying offshore is typically offset by shipping and customs costs, and additional inventory associated with long transport times. The threshold quantity for offshore savings is usually about 10,000 cold plates or more per year.

The biggest cost drivers for aluminum cold plates, after those mentioned above, are machining time and additional processing steps. Cold plate manufacturers typically have a cost associated with machining time which covers depreciation costs of the machine, power, supplies, and maintenance. Therefore, the longer the cold plate sits in the machine the more costly it is. Each additional processing step continues to drive the cost up.

Extrusions and Castings

To minimize machining time and drive down cost, it is best to use extrusions and casting as much as possible. An extrusion is produced by pushing metal through a die to create an object with a fixed cross-section. Dies for a new extrusion are relatively inexpensive and extrusion size is limited to about 9 inches (22.86 cm) wide. Extrusion wall thickness needs to be relatively consistent and any channels or features need to be straight.

Your manufacturer can also use a combination of extrusions and machining to reduce costs. An extrusion can be made for some of the features and then the more complicated features can be machined. Another option for prototyping purposes is to machine the cold plate for lower quantities and then, once the design is proven and fixed, make the die for the extrusion. This will help to keep extrusion costs down, provided you design the plate with the extruded features in mind.

Another option is to combine casting and machining to make cold plates. For instance, if the casting is not flat enough, a secondary operation to get the cold plate to the required flatness specification will be necessary. It is important to note that sand castings are not an option for vacuum-brazed cold plates because most alloys used have a melting temperature below the vacuum-brazing temperature. Their use is strictly limited to tubed cold plates. Obtaining quotes on the two production processes and weighing the pros and cons is recommended.

Typically, the minimum purchase quantity for an extrusion or a casting is high, so you need the right application in order to justify using these processes. Both extrusions and castings can provide significant cost savings overall.

A typical machined cold plate will have a surface finish of 32-64 µin (81-163 µcm), which is sufficient for most applications. Roughness can be reduced to 16 µin (41 µcm) using a standard machining center, however this requires more rigid fixturing to reduce any potential chatter and slower speeds and feeds of the machining head. (Speed is the rate at which the cutting tool head spins and feed is the rate at which the machine head moves across the cold plate.) Reducing both speed and feed translates to longer machining center time, thereby increasing cost.

Most applications use a thermal interface material (TIM) between the component or board and the cold plate to help to minimize the gaps. A TIM should be as thin as possible, as the relatively high thermal resistance of the TIM greatly overshadows any conductivity improvements from having a smoother surface. Increasing the clamping force of the component or board to the cold plate can also help to offset a higher roughness, but it may increase the stress on the board or component. Clamping stress can also increase the impact of coefficient of thermal expansion (CTE) mismatches as the cold plate and component or board heat up.

Surface Flatness

Surface flatness has more impact on a cold plate’s thermal performance than surface roughness, as the contact area is greatly reduced if the cold plate is not flat (See Figure 2). The standard flatness specification is 0.001 inch/inch (0.003 cm/cm). Therefore, within an inch of your measurement point, the cold plate’s lowest point will not be more than 0.001 inch (0.003 cm) lower than the highest point. If your specification requires flatness better than 0.001 inch/inch (0.003 cm/cm), one way to save money is to specify a local flatness rather than a tight flatness across the whole plate. For instance, if you are mounting multiple insulated gate bipolar transistors (IGBT) on a cold plate and each IGBT requires 0.001 inch/inch (0.003 cm/cm) across the whole base plate, specify the local flatness for a single IGBT rather than requiring the whole plate to be very flat.

The flattening process of a cold plate typically involves a hydraulic press. A skim cut can be used to improve flatness. With a skim cut, the machine tool determines the lowest point of your cold plate and skims off very little metal at the lowest point and as much metal as necessary at the higher areas to achieve a flat surface. While skim cutting a block of aluminum is very easy, skim cutting a vacuum-brazed cold plate or the tubed side of a tubed cold plate is more difficult. The cooling surfaces of vacuum-brazed cold plates and tubes in tubed cold plates are typically thin in order to optimize thermal performance. If the cold plate is not flat the skim cut could be too deep and the walls will be thinned, potentially becoming too weak to hold pressure or even breaching. Alternatively, you can start with a thicker cold plate to eliminate the potential for leaks, but you’ll sacrifice some performance.

Surface Topography

Minimizing surface topography is also important in keeping costs down, particularly for board applications. Complex surface topography generally requires starting with a thick block of aluminum and machining off aluminum that is not needed. This results in high raw materials costs and excessive machining time. If topography cannot be eliminated, bundling components with similar heights on the board can reduce the machining requirements.

Hardness

Mounting Features / Holes

Another cost adder in cold plate manufacturing is the addition of holes. One hole may add as much as $3 to the cost of a cold plate. One of the main reasons that holes add costs is that holes cannot be made in the fluid path. Therefore, for a tubed cold plate, a bend in the tube needs to be made to accommodate the hole, and each bend adds cost. For a vacuum-brazed cold plate, an island must be created in the fluid path, which also means electrical discharge machining (EDM) the internal fin. This can add quite a bit of machining time and therefore cost.

There are several types of holes. One type is a through hole, which passes from one side of the cold plate to the other. A second type is a tapped hole, which has screw threads. Since aluminum is relatively soft, tapped holes have a limited life if the components or boards are frequently changed. With tapped holes, helicoils are often used. A helicoil is a sturdy steel insert that adds strength to the threads for applications in which there is likely to be frequent component change out. Through holes are produced by a single drilling process, while tapped holes require an additional tool on the same machine set up. Helicoils require a tapped hole in order to be installed, and the helicoil installation itself is completed outside of the machining center. In summary, through holes are the least expensive and helicoils are the most expensive.

Tight tolerance of the location and spacing of holes can also drive up costs. A reasonable tolerance specification is ±0.005 inch (±0.013 cm). As with flatness, specifying local tolerances when possible will reduce cost. With big cold plates where holes can be relatively far away from each other, the tolerance becomes harder to maintain. One reason is that machine tool tolerances increase as the head has a farther distance to travel. Another reason is that there may be thermal gradients of as much as 18°F (10°C) in the machine shop, which can expand or contract the cold plate by as much as 0.005 inches (±0.013 cm). Through holes are the easiest to specify a tighter tolerance for because creation of a through hole is accomplished with a single tool operation, while tapped holes are not as easy to tolerance because making them involves two tools. Helicoils are the hardest to tolerance because the process requires a tapped hole and the helicoil itself has a tolerance. All the tolerances add up, making it harder and more expensive to manufacture. Avoiding small tapped holes will also help to reduce cost. Hole sizes of 4-40 or smaller become difficult to tap as the taps can break while drilling. In order to minimize this problem, the machine must run much slower. One way to counter tight tolerances requirements on a cold plate is to increase the size of the mounting holes in the component or board.

Liquid Connections

For liquid connections, straight threaded O-ring female ports generally work best. Other than a welded system, it provides the best sealing at the lowest cost. Plumbing connections, such as a NPT fitting, do not provide the precision needed for components such as cold plates. On a vacuum-brazed cold plate, a male fitting, such as a barbed or beaded fitting, should be avoided because it requires another operation such as welding to attach the fitting. In addition, fittings that extend beyond the cold plate need to be protected during shipment, potentially adding packaging costs. Quick disconnects should only be used when necessary because they can cost as much as $750 per pair. Quick disconnects are required on cold plates or electronics that need to be frequently replaced. They are also required for cold plates that are shipped already charged with cooling fluid. With liquid connections another consideration is the port tolerance. Usually incoming plumbing has some flex to it. A reasonable tolerance is between ±0.030 inch (0.076 cm) and ±0.060 inch (0.152 cm).

Design and Manufacturing Partnerships

Working with a cold plate manufacturer early in a cold plate’s design or being flexible on a build to print design will allow for the greatest amount of cost savings. Although the two biggest cost drivers in cold plate manufacturing are thermal performance requirements and annual demand, there are many other factors over which thermal and/or component engineers have some control. Ensuring that there is a reason for every specification, as every specification may drive up cost, will help to keep costs down. It’s important to determine when roughness, flatness, surface topography, hardness, mounting features and holes, and liquid connections specifications are necessary. In addition, it’s important to realize that there are many alternatives in not only the design, but also the manufacturing processes used, which can save hundreds if not thousands of dollars in manufacturing.

Visit our Liquid Cold Plate Section to learn more about our solutions.

Considerations for Making a Cold Plate Specifically for Your Application

Cold Plate Technologies

Normalized Performance Curves

Channeled Cold Plates

Key Thermal Considerations During Liquid Cold Plate Design

Cold Plate Thermal Specifications

In addition to four types of cold plate technologies, there are also four scenarios of thermal requirements, which are listed below:

• Uniform Heat Flux, Fixed Flow Rate, 1 Maximum Pressure Drop, 1 Maximum Surface Temperature – With thermal scenario one, there is uniform input heat flux, a fixed flow rate, one specified maximum pressure drop that is limited at a fixed flow rate, and one specified maximum surface temperature where the surface temperature does not need to be uniform.
• Same as 1, but Non-Uniform Heat Flux – Thermal scenario two has the same specifications as scenario one, but heat loads vary instead of being uniform. The heat loads are concentrated in several locations under components or under specific areas
• Same as 1, but Surface Temperature Maximum Varies – Thermal scenario three also has the same specifications as scenario one, but thermal scenario three has specified maximum surface temperatures that vary across the cold plate, usually at the individual components.
• Same as 1, 2, or 3, but Surface Temperature Uniformity Required – With thermal scenario four, the thermal specifications may be the same as with thermal scenarios one, two, or three, but with the additional requirement that the maximum surface temperature must be uniform across the entire cold plate or under specific components. For example, if there are two types of components mounted on the cold plate, each component type may require temperature uniformity of the common components, but the two types may have different maximum surface temperatures.

Cold plate scenarios 2 and 3 are the ones most commonly encountered in custom cold plate design. Scenarios 1 through 4 are listed in order of increasing complexity and cost.

When designing custom cold plates to any specification, the logical steps most thermal experts take are defining the thermal map, generating the liquid circuiting concept, calculating temperature rise and pressure drop, and rerouting the liquid circuit if necessary.

Defining the Thermal Map

With several possible thermal scenarios, step one in custom cold plate design is to define the thermal map in detail. To create a thermal map, an engineer needs the dimensions, locations, and heat loads of the components to be cooled. The maximum allowable cold plate surface temperature(s); the coolant composition, its flow rate, and inlet temperature; and available pressure drop are needed as well. Also, heat flux must be calculated for each component which includes thermal spreading, if necessary.

Key Thermal Considerations During Liquid Cold Plate Design

Generating the Liquid Circuit Concept

The next step is to generate the first iteration on a liquid circuit concept. The liquid circuit must provide the required performance to cool the component with the highest heat flux and each component after it on the liquid circuit. In addition, it must do so with the specified flow rate and with an acceptable pressure drop. Sometimes techniques such as uneven widths of liquid series passes, different fin densities under individual components, and varying fin heights and types can be used to satisfy the competing requirements of performance and pressure drop. The fin’s geometry and height determine the “fin efficiency”, or how well it transfers heat to the liquid.

Sometimes the shape of high heat flux components (e.g. – a large round footprint) requires a change from the natural uniform flow distribution over the pass width to force non-uniformity, which can be achieved by using different lengths of fin or different fin densities over the pass width. Before the next component, some liquid equalizing pools (i.e. – mixing pools) should be designed in. Another fluid distribution challenge is the need for islands in the fluid path to accommodate component mounting. Any complication mentioned above can increase the cost of the cold plate due to the additional number of fin pieces, multiple depths in a cavity, multiple fin-forming equipment set-ups, and EDM cutting needed.

Calculating Temperature and Pressure Drop (Detailed Design Stage)

After the liquid circuit concept is outlined, the thermal map should be verified by calculating the maximum surface temperature under each component and calculating the total pressure drop. All the critical thermal areas must be modeled. If any one of the requirements is not met, the liquid circuits must be reworked and the calculations rerun.

Rerouting the Liquid Circuit

If the cold plate requires a varying maximum surface temperature (as in thermal scenario three) and normal liquid circuiting does not meet the specifications, the liquid circuit should be rerouted to deliver the coolest liquid to critical devices first or to by-pass part of the liquid directly to these components.

Temperature Uniformity

If the cold plate requirements specify maximum surface temperatures and temperature uniformity as in thermal scenario four, the design process is even more complex. The simplest solution to provide uniformity of maximum surface temperatures of identical components is to position the components on similar points of similar parallel liquid passages. The result should be a circuit that delivers liquid with a common temperature at sufficient flow rates to these components. Another technique that is used to provide a more uniform surface temperature across the entire cold plate is to use a counterflow arrangement (Figure 2). In a number of parallel channels, on a surface or on both sides of the plate, each second channel has flow in the opposite direction. For a one-side loaded or very thin cold plate, such an approach may significantly reduce surface temperature gradient. A similar effect may be delivered by organizing two separate layers of liquid.

Reducing Complexity & Costs

Certain thermal or mechanical requirements may force an illogical pass of the liquid circuit, resulting in greater complexity and a higher cost cold plate. For example, applications frequently have predetermined mounting hole locations that the liquid circuit must navigate around and/or components and fluid inlet and outlet locations that are fixed, significantly limiting the options for the liquid circuit. Generally, the more flexible the design is, the easier the cold plate will be to engineer and the more savings you’ll realize. By working closely with a printed circuit board designer or electrical engineer, the thermal engineer can provide input on the spacing and positioning of components to ensure they are designed with electrical as well as thermal requirements in mind. This may significantly simplify the cold plate design and reduce cost. For more information on cold plate costs please see our application note “Cold Plate Manufacturing Cost Drivers”.

It’s important to understand the various design techniques that allow a custom cold plate solution to meet the most challenging thermal and mechanical requirements. With thousands of permutations for a custom cold plate design, skilled engineering is key. Flexibility with the location of inlets and outlets, proper fluid circuit routing, and the use of fin or channels can help to create a thermal solution that provides great value for the application. As heat loads become more and more concentrated and the space allocated for cooling becomes smaller and smaller, custom cold plates will be used more and more to meet applications’ unique liquid cooling needs. Aavid, Thermal Division of Boyd Corporation has decades of experience designing and manufacturing custom cold plates for printed circuit boards and other electronics and ensuring their high thermal performance requirements and cost limits are met or exceeded.

Learn more about our different Liquid Cold Plate solutions in our Liquid Cold Plate Section.

Calculating the Thermal Resistance of a Liquid Cold Plate

To select the best cold plate for your application, you need to know the cooling fluid flow rate, fluid inlet temperature, heat load of the devices attached to the cold plate, and the maximum desired cold plate surface temperature, Tmax. From these you can determine the maximum allowable thermal resistance of the cold plate.

First, calculate the maximum temperature of the fluid when it leaves the cold plate, Tout. This is important because if Tout is greater than Tmax, there is no solution to the problem.

Alternatively, you can use the heat capacity graphs found in our Thermal Reference Guide in the technical library. These graphs describe the change in temperature, ΔT, that occurs along the fluid path. To find Tout, add ΔT to the inlet temperature, Tin.

Assuming Tout is less than Tmax, the next step is to determine the required normalized thermal resistance (θ) for the cold plate using this equation:

Any cold plate technology that provides a normalized thermal resistance less than or equal to the calculated value will be a suitable solution.

Cold Plate Performance Comparison

We present cold plate performance data using local thermal resistance – the surface temperature versus the local liquid temperature. This methodology enables more precise thermal analysis for high heat loads. See full details on thermal resistance calculations and how to select a cold plate technology.

Normalized Performance Curves

Notes

Thermal resistance is inversely proportional to area. To find the thermal resistance of a 25 square inch cold plate, divide the normalized performance by 25.

Our CP30 standard cold plate is designed for prototyping purposes. It has a thick surface plate for machining. We show two traces – before machining (0.5″ / 13 mm) and after machining (0.05″ / 1.3 mm). The performance of a custom vacuum-brazed cold plate is usually significantly better than this standard part.

For comparison purposes, the performance of all cold plates is shown using water as the coolant. Treated water is recommended with aluminum (CP20 & CP30) cold plates.