Liquid Cooling Technologies for Data Centers and Edge Applications

Introduction to Liquid Cooling:

As IT equipment technology advances, so do the cooling solutions needed to manage heat. While liquid cooling has been used for many years in mainframes and high-performance computing (HPC), the current demands of cloud, IoT, AI, and edge applications are prompting a renewed focus on liquid cooling. This technology is also being reconsidered due to its potential for enhancing energy efficiency and sustainability in data centers.

Comparing Air Cooling and Liquid Cooling:

Air cooling, the dominant method in data centers, relies on airflow to remove heat from IT equipment. However, with the increasing use of GPUs and high-core-count CPUs, which generate substantial heat, liquid cooling is becoming more efficient. Liquids have a higher capacity to capture and transfer heat, enabling better performance and energy efficiency. Additionally, liquid cooling allows for heat to be reused in other applications, such as district heating.

History and Evolution of Liquid Cooling:

Liquid cooling has a long history, dating back to the 1960s when it was used in IBM mainframes. In the early 1990s, the introduction of CMOS technology reduced power consumption, making convective airflow cooling sufficient. Today, liquid cooling is again gaining traction, especially in gaming, blockchain mining, and HPC applications that require higher compute capacities.

Methods of Liquid Cooling:

Liquid cooling methods are categorized into two main types: direct-to-chip and immersive. These categories include five primary methods:

1. Direct-to-Chip Liquid Cooling (Single-Phase):

This method involves taking the coolant directly to the hotter components (CPUs or GPUs) with cold plates on the chips. Fans are still needed to remove residual heat, but conventional air-cooling infrastructure is reduced.

2. Direct-to-Chip Liquid Cooling (Two-Phase):

Similar to the single-phase method, but the coolant changes state (liquid to gas) to remove heat, requiring additional system controls.

3. Immersive Liquid Cooling (IT Chassis – Single-Phase):

IT components are fully or partially immersed in a dielectric liquid, eliminating the need for fans and offering near-silent operation.

4. Immersive Liquid Cooling (Tub – Single-Phase):

IT equipment is submerged in a fluid within a tub. Heat is transferred to a water loop via a heat exchanger. Servers are pulled out on a vertical plane, similar to laying a traditional rack on its back.

Key Attributes for Selecting Liquid Cooling Methods:

1. Capital Cost:

Liquid cooling can offer cost savings over air cooling, particularly in greenfield facilities optimized for it. Retrofit costs may be higher but can free up stranded power and space.

2. Energy Cost:

Liquid cooling is known for its energy efficiency and can reduce IT fan energy, potentially lowering overall electricity bills despite possible increases in Power Usage Effectiveness (PUE).

3. Serviceability:

Liquid cooling requires new procedures and familiarity. Dripless connectors and proper maintenance practices are essential, especially for immersive cooling.

4. Rack Density/Compaction:

Both direct-to-chip and immersive cooling support higher rack densities, with immersive cooling capable of achieving over 100 kW/rack.

5. Water Usage:

Liquid cooling can significantly reduce water usage compared to air cooling systems that rely on evaporative cooling.

6. Harsh Environments:

Immersive cooling is sealed and does not require airflow, making it suitable for harsh environments and edge applications.

7. Fan Noise/Air Movement:

Liquid cooling reduces or eliminates fan noise, offering quieter operation.

8. Room Layout:

Immersive cooling provides greater flexibility in data center layouts as it does not require airflow through the IT equipment.

9. Ability to Retrofit IT Equipment:

Direct-to-chip cooling is easier to adapt to existing air-cooled servers, while immersive cooling requires more significant changes.

10. Scalability:

Both methods can scale incrementally, with tub-based immersion requiring the deployment of entire tubs.

11. Fluid Tradeoffs:

Different fluids (water-based, hydrocarbon oils, engineered fluids) offer varying thermal performance, cost, and safety considerations.

Conclusion:

Liquid cooling is gaining interest in data centers and edge environments due to rising chip densities and a focus on energy efficiency. Direct-to-chip and immersive methods offer significant benefits over air cooling, with specific methods suitable for different applications. While broader adoption requires further efforts, liquid cooling is poised to play a crucial role in future data center and edge application designs. For retrofit sites, direct-to-chip and chassis immersive liquid cooling provide the easiest transition, while new sites and those in harsh environments benefit more from immersive liquid cooling.