Immersion Cooling vs. Direct-to-Chip Liquid Cooling: The Role of Secondary Airflow

Article 46: Immersion Cooling vs. Direct-to-Chip Liquid Cooling: The Role of Secondary Airflow

Introduction: The Liquid Cooling Revolution
As TDPs (Thermal Design Power) for CPUs and GPUs climb toward 500W and beyond, traditional air cooling is reaching its physical limits. In response, data center operators are increasingly turning to liquid cooling—specifically Immersion Cooling and Direct-to-Chip (DTC) Liquid Cooling.

While liquid is vastly more efficient at carrying heat than air, a common misconception is that "liquid cooling eliminates the need for fans." In reality, most high-performance liquid-cooled systems still require a sophisticated secondary airflow strategy to protect the components that the liquid loops miss.

Immersion Cooling: The Fanless Myth?
In single-phase or two-phase immersion cooling, the entire server is submerged in a dielectric fluid. While this eliminates internal server fans, it moves the thermal burden to the "Heat Rejection Unit" (CDU/Dry Cooler).

- **The Fan's New Home**: Large-diameter (800mm+) EC axial fans are required at the dry cooler outside the building to reject the heat from the dielectric fluid to the atmosphere.
- **SXDOOL Role**: We provide high-efficiency EC fans for these dry cooler units, ensuring the fluid temperature remains within the optimal range.

Direct-to-Chip (DTC) Cooling: The Hybrid Reality
DTC cooling uses "Cold Plates" mounted directly on the primary heat sources (CPU/GPU). However, a modern server motherboard contains hundreds of other heat-generating components:
- **VRMs (Voltage Regulator Modules)**: These convert high-voltage power to the low-voltage required by the chips. They can generate significant heat.
- **Memory (DDR5/HBM)**: High-speed RAM is extremely temperature-sensitive and is rarely covered by a liquid loop.
- **Storage and IO**: NVMe drives and networking cards still require active cooling.

In a DTC system, **Secondary Airflow** is mandatory. This is a hybrid approach where liquid handles 80-90% of the load, and air handles the remaining 10-20%.

Designing for Secondary Airflow
Because the CPU/GPU heat sinks are replaced by flat cold plates, the internal airflow paths are often wide open, which can actually *reduce* the velocity of air over the VRMs.

1. Strategic Blower Implementation
Small, high-pressure blower fans (centrifugal fans) are often used to target VRM banks specifically. Their focused airflow "jets" can scrub heat from small components that axial fans might miss.

2. High-Temperature Tolerance
Because the ambient air inside a liquid-cooled server is often pre-heated by other components, secondary fans must be rated for continuous operation at +70°C or +80°C.

3. Redundancy in Airflow
Just as the liquid pump is redundant, the secondary air cooling must be as well. N+1 fan configurations ensure that a single fan failure doesn't lead to a VRM meltdown.

Conclusion: The Synergy of Air and Liquid
The future of high-density computing is not "Liquid vs. Air," but rather "Liquid + Air." Understanding the critical role of secondary airflow in protecting VRMs and RAM is the difference between a reliable system and a catastrophic failure. At SXDOOL, we specialize in the high-reliability fans and blowers that complement the world's most advanced liquid cooling architectures.