Chris Carreiro, CTO of Park Place Technologies, compares the most common liquid cooling technologies. With the global data center liquid cooling market set to reach nearly $5 billion by 2023 (and projected to surpass $20 billion by 2030), the rush to adopt the latest liquid cooling technologies is on.
However, while organizations understand the general benefits of liquid cooling (improved server performance, increased rack density, lower energy costs, reduced water consumption, and more), many are stuck with the seemingly countless types of liquid cooling technologies and specific products (reservoirs, cold plates, immersion-capable servers, etc.) offered by manufacturers. While all liquid cooling technologies exceed the traditional air cooling (or crash), the best option for you will probably depend on your specific use case and on the performance metrics that are most important for you. For reasons of brevity, we will keep this discussion limited to the most common liquid cooling technologies: direct immersion and cooling or directly to the chip. Immersion Cooling
With immersion cooling, servers are completely immersed in a dielectric (or insulating) fluid inside a specially designed tank.
In single-phase immersion cooling, the fluid leaves the immersion tank via a coolant pump that passes through a heat exchanger and returns to the immersion tank at a lower temperature, where this heat transfer cycle continues. Single-phase immersion cooling is widely used in data centers to reduce energy consumption and improve sustainability by reducing the carbon footprint associated with traditional air cooling methods. It is also used in high-performance computing and mining operations (such as cryptocurrency), which generate enormous heat. The immersion cooling provides the best results of the performance of any liquid cooling technology. Allows the highest frame density (up to 10 times more than air cooling), drives the best energy efficiency (up to 50% energy saving) and provides the higher noise reduction than Any other option.
In two-phase immersion cooling, heat from the immersed server components boils the special immersion fluid. The resulting vapor heats a condenser coil at the top of the sealed chamber. Refrigerant from the condenser coil exits the chamber to a heat rejection mechanism (cooling tower, etc.). Then, the coolant is sent back to the sealed chamber at a lower temperature, ready to continue the heat transfer cycle.
Two-phase immersion cooling result in an even more efficient heat removal process (2-3X) than single-phase – however they’re more costly, more difficult to maintain, and fluorinated fluids used in two-phase immersion cooling can be harder on the environment. Use cases for two-phase cooling include IoT and locations where space and power availability may be limited.
With direct-to-chip cooling (DTC), coolant flows through a cold plate applied directly to the components that generate the most heat, such as the CPU and GPU.
In the single-phase DTC process, a cooling distribution unit (or CDU) circulates coolant across the cold plate. The heated fluid then passes through a closed loop system to a heat exchanger at another location.
Although DTC cooling performance does not match the levels achieved with immersion cooling, it does provide an excellent balance of performance, a relatively lower cost of entry, and easier maintenance. In fact, DTC applications can be 50 to 1,000 times more efficient than air cooling, resulting in CPUs and GPUs running in the thousands per chip, compared to 200-300 with air cooling. The downside is that single-phase DTC can cause liquid leakage, which can lead to motherboard corrosion.
As with immersion cooling, two-phase DTC cooling involves a phase change of the coolant, resulting in more efficient cooling and allowing for higher power densities. However, these efficiencies come at the expense of more complex systems that are more difficult to manage and maintain.
DTC is increasingly used in data centers that use AI and machine learning solutions, 24/7 telecommunications operations, and industries that rely heavily on advanced computing and electronics systems. The Final Analysis
The right liquid cooling technology for you requires a careful consideration of the performance parameters that matter most to your organization, the budget you have to implement the systems, and your ability to manage and maintain them over time, either internally or externally through a partner.
Which Liquid Cooling Technology Is Right for Your Organization? Chris Carreiro, Director
Park Place Technologies CTO compares the most common liquid cooling technologies. With the global data center liquid cooling market expected to reach nearly $5 billion by 2023 (and expected to surpass $20 billion by 2030), the race to adopt the latest liquid cooling technologies is on. However, while organizations are waking up to the general benefits of liquid cooling (improved server performance, increased rack density, lower energy costs, reduced water consumption, and more), many are stuck in the seemingly endless list of specific liquid cooling technologies (reservoirs, cold plates, submersible servers, etc.) provided by manufacturers.
While all liquid cooling technologies outperform traditional air cooling (or CRAC), the best option for you will likely depend on your specific use cases and the performance metrics that are most important to you. For brevity, let’s limit this discussion to the most common liquid cooling technologies: immersion and direct-to-chip or direct-to-chip cooling.
With immersion cooling, servers are completely encased in a liquid dielectric (or insulator) inside a specially designed tank. In single-phase immersion cooling, the fluid leaves the immersion tank through a cooling pump that passes through a heat exchanger and returns to the immersion tank at a lower temperature, where this heat transfer cycle continues. Single-phase immersion cooling is widely used in data centers to reduce energy consumption and improve sustainability by reducing the carbon footprint associated with traditional air cooling methods. It is also used in high-performance computing and mining operations (such as digital currencies) that generate a lot of heat. Immersion cooling provides the best performance results of any liquid cooling technology. It enables the highest rack density (up to 10 times that of air cooling), provides the best energy efficiency (up to 50% energy savings), and provides the highest noise reduction compared to water cooling.
In two-stage immersion cooling, the heat of the immersed server components boils the special immersion fluid. The resulting vapor heats a condenser coil at the top of the sealed enclosure. The refrigerant exits the condenser coil from the chamber to the heat removal mechanism (cooling tower, etc.). The refrigerant then returns to the sealed chamber at a low temperature, ready to continue the heat transfer cycle.
Two-phase immersion cooling allows heat removal even more efficiently (2-3 times) than single-phase. However, they are more expensive, more difficult to maintain, and the fluorinated fluids used in two-phase immersion cooling can be harmful to the environment.
