2021 Data Center Cooling Solutions: Microchannel, Microconvective & Immersion Cooling

2021 Data Center Changes and Trends

Read our latest data center case study now

Data center infrastructure is undergoing massive changes to keep pace with public cloud computing companies. During the pandemic, public cloud providers gained momentum in the market as consumers turned to cloud-based technology to work, play, and stay connected with friends and family. Public cloud giants like Amazon, Google, and Microsoft went on a spending spree resulting in them collectively owning 50 percent of the world’s largest data centers [1].

Combating the market takeover, data center providers are redesigning their sales operations to better position themselves in the market. As a result, “as-a-service” data consumption has arrived. Data center “as-a-service” refers to a hosting service in which physical data center infrastructure and facilities are provisioned to clients, according to Insight [2]. Data center providers are increasingly providing consumption-based, pay-as-you-go packages and solutions to better suit increasingly diverse user needs. However, this doesn’t mean cloud computing is going anywhere. Edge computing and other cloud computing software are quickly becoming the backbone of our digital economy and society. In fact, the number of data centers operated by hyperscale cloud providers has increased to nearly 600 [3].

As high-density cloud services and data center providers continue to grow, energy consumption costs will skyrocket. Currently, U.S. data centers consume approximately 2% of all U.S. energy, according to Mordor Intelligence data center liquid cooling market report [4]. Global adoption of cloud-native services will be directly tied to sustainability efforts. As a result, data center and cloud providers are pursuing sustainable, environment-friendly solutions to decrease energy costs, as well as operational costs. Major market players like Amazon Web Services, Google, and Microsoft have spearheaded the sustainability shift, raising the overall market standards [5]. This positive correlation between sustainability and technological progress has led to innovative solutions.

How Liquid Cooling Can Help

Innovative liquid cooling technologies provide powerful solutions to support the ongoing data center infrastructure changes. Liquid cooling solutions have been exploited in the market to increase performance and decrease costs. However, as compute loads continue to densify, traditional liquid cooling technologies are reaching their limit. New liquid cooling technologies are needed to bridge the heat transfer gap enabling performance. Luckily, there are several innovative liquid cooling technologies circulating the market that address these performance and sustainability demands.

Direct-to-Chip Cooling Solutions

There are many different direct-to-chip liquid cooling technologies, however, all operate on the same premise – they target the chip directly with a heat transfer fluid. In comparison to legacy liquid cooling technologies that use liquid as an intermediary to provide cooling to air entering the rack, direct-to-chip cooling technologies use a heat transfer module (cold plate or heat sink) placed on the server board to directly cool the high-power rack components. Specifically, the cold plate or heat sink is attached directly to heat-intensive components such as CPUs or GPUs to distribute the heat transfer fluid or coolant to high heat emitting sections.

Generally, direct-to-chip liquid cooling methods allow enterprises to increase their overall cooling temperature. By increasing the coolant operating temperature, enterprises can potentially eliminate the need for cooling towers or chillers. The elimination of chillers and cooling towers decreases water waste and energy consumption and, by extension, operating costs. This traditional form of liquid cooling has been advanced to address hyperscale computing and dense workloads. Let’s learn more about advanced direct-to-chip liquid cooling solutions.

Microchannel Liquid Cooling

One innovative liquid cooling solution is microchannel liquid cooling. Microchannel cooling implements a form of direct-to-chip liquid cooling that uses cold plates to directly target CPUs, GPUs, or other memory modules. Simply put, microchannel cooling operates on a heat-spreading premise. Specifically, microchannel cold plates (sealed metal plates) spread the heat generated in the device into small internal fluid channels that can be removed by a flowing coolant. The heat spreading technique provides a high cooling surface area, while the small fluid channels promote intimate interaction between the coolant and the heated surface. Channels may be skived, pin fin, or machined channels in nature. Because they are closed cold plates, most microchannel solutions require a thermal interface material (TIM2) to be applied when attaching to a heat-generating component.

In short, the enhanced area and intimate fluid contact with the cold plate structure allow for high heat removal: however, the thermal interface material can sometimes be a limiting thermal resistance layer as thermal design power (TDP) reaches peak performance levels.

Microconvective Liquid Cooling

Another new innovative form of direct-to-chip liquid cooling is microconvective liquid cooling. Created and coined by JetCool, microconvective liquid cooling technology aims to further enhance current liquid cooling systems to enable improved performance for applications with the densest compute profiles.

Microconvective cooling uses concentrated arrays of small fluid jets to bring coolant directly to the surface of the heat-generating device. This creates high effectiveness turbulent flow where the heat is being generated. In contrast to typical parallel flow configurations found in microchannel solutions, microconvective cooling facilitates perpendicular flow onto the device, which increases the heat transfer coefficient for heat removal. Because direct coolant contact occurs on the device’s surface, thermal interface materials are eliminated. As a result, high heat fluxes can be managed easier due to the minimized thermal resistance. This technique can handle the highest TDP processors on the market, allowing data center integrators to future-proof their data center systems.

Learn more about JetCool’s microconvective liquid cooling solutions for data centers here.

Immersion Cooling Solutions

Immersion cooling, sometimes called liquid submersion cooling, is the practice of submerging data center hardware in a thermally conductive liquid bath. As a result, the liquid bath covers all components on a server board directly transferring the heat via the fluid. This technique is efficient because the heat transfer fluids have a higher heat capacity than typical ambient air surroundings. Of course, it’s important that the heat transfer liquid has low electrical conductivity to ensure there are no electrical failures or short circuits. Consequently, heat transfer fluids are typically a form of mineral oil, fluorocarbon-based fluids, or synthetic oil. Although these heat transfer fluids are an improvement over traditional air-cooled systems, immersion coolants are unable to match the thermal performance of water-based coolants used in direct-to-chip liquid cooling, which can limit the processor TDP in immersion systems. To learn more about the properties of heat transfer fluids, check out our blog post here.

Two-Phase Immersion Cooling

This cooling solution is an extension of single-phase immersion cooling described above. In single-phase liquid cooling, the fluids remain in its liquid form during the entire cooling process, elevating in temperature as they remove heat from heat-generating devices. Meanwhile, in two-phase immersion cooling, the coolant changes its state of matter. As the heat transfer fluid heats up to boiling point, heat is removed by latent heat transfer; liquid converts into gas instead of raising the surrounding coolant temperature. Simply put, this method utilizes phase change as a form of heat extraction, which can offer higher local heat fluxes and lessen the pumping infrastructure requirements. While fairly new, this cooling technology is gaining some market traction. As of April 2021, Microsoft is testing two-phase immersion liquid cooling on a hyperscale Azure data center.

Considerations for Immersion Cooling

Many industries that need hyperscale data centers or computing are considering a large-scale infrastructure transition to immersion cooling. However, there are some practical considerations to keep in mind when considering immersion cooling methods.

Modified Hardware. Since immersion cooling fully submerges data center hardware into a heat transfer fluid, the hardware often requires modifications to allow for high reliability, long lifetime operation. For example, many immersion systems move away from the vertically stacked rack configuration, flipping the rack onto its side before submerging in a bath with a horizontal form factor.

Complex Installation. While immersion cooling has its benefits for large-scale implementation, there are many parts that must be installed for successful cooling. Depending on the company, installation of immersion cooling can include tanks, fluid, pumps, chillers, filters, piping, sensors and more.

Conclusion

Overall, data centers are undergoing many changes as they prepare for the increase in compute loads and demands. As processor performance continues to advance, devices and servers will need more efficient, intuitive systems to fulfill projected growth rates. Luckily, several liquid cooling solutions can help growing enterprises bridge this gap between supply and demand. As the sustainability trend sweeps through the compute market, liquid cooling solutions offer a helping hand to high-waste enterprises. In fact, most liquid cooling solutions empower enterprises to meet market demands while remaining environmental conscious and responsible. All in all, liquid cooling solutions provide promising solutions to address the changes in the data center market.