In the modern digital world, servers are the cornerstone of all network services. However, when core components such as CPUs, GPUs, memory, and hard drives operate at astonishing speeds, a large amount of heat is generated. If this heat cannot be dissipated in time, servers may experience reduced clock speeds and performance degradation, or even crash and shorten their lifespan. In server cooling solutions, cooling fans are the most common and crucial component. What technological intricacies lie behind this small fan?

I. Why is server cooling fan so “difficult”?

The root cause of server cooling fan  challenges lies in three main characteristics: “small space, high density, and uninterrupted operation.”

Take a typical 1U server as an example; its chassis height is only about 44.45mm. After deducting the sheet metal, rails, internal modules, and vibration damping structures, the installation space left for fans is extremely limited, often requiring the use of 40×40×28mm or even smaller fans. Space is compressed, but heat generation is not reduced—the densely packed CPU heatsink fins, the compact arrangement of memory modules, and the stacked power supply modules and motherboard components all contribute to a high-resistance airflow environment. More problematic is that servers typically need to run 24/7. Fan failure can lead to system throttling, service interruptions, and even increased thermal stress on the motherboard and power supply.

II. Two Key Fan Parameters: Airflow and Static Pressure

When selecting server cooling fans, two core parameters must be understood: airflow and static pressure.

Airflow measures the volume of air a fan can move per unit of time (unit: CFM or m³/min), determining “how much heat can be removed,” essentially the “ability to move air.” Static pressure represents the airflow’s ability to overcome resistance (unit: Pa), determining “whether it can penetrate complex airflow channels,” similar to “the ability to propel air.”

In high-density server scenarios such as 1U, static pressure often takes precedence over airflow. This is because the nominal airflow is measured under “free air” conditions, while in high-resistance airflow channels, the actual airflow drops significantly. If the static pressure is insufficient, air cannot enter critical areas such as the heatsink fins and memory modules, leading to severe localized hotspots.

III. Fan Types: The Different Roles of Axial and Centrifugal Fans

Cooling fans are mainly divided into two categories based on their working principle and airflow direction: axial and centrifugal (blowers).

Axial fans have the same airflow direction (intake and exhaust), with the blades propelling air along the axis. They offer advantages such as large airflow, long lifespan, and wide application; most fans in server racks belong to this category. However, their airflow capacity is significantly reduced in high-resistance airflow ducts.

Centrifugal fans (blowers) have their air inlet located at the center of the flat surface. Air is expelled from the single outlet of the casing after the centrifugal force generated by the rotating blades. The inlet and outlet form a 90° angle. They are characterized by their small size and relatively low noise, making them suitable for the “side exhaust” requirements of flat-space equipment. Generally, centrifugal fans are more effective than axial fans in delivering airflow to high static pressure environments, and are therefore often used in air handling systems requiring penetration capabilities.

In practical selection, the fan type needs to be matched according to the airflow characteristics of the equipment—low-resistance systems are suitable for axial flow fans with a gentler P-Q curve, while high-resistance systems should choose blowers with a steeper P-Q curve.

IV. Bearing Technology: The “Heart” Determining Lifespan and Noise

Fan bearings are the core support structure for motor rotation, directly affecting the fan’s lifespan, noise, and reliability. Common bearing types each have their advantages and disadvantages:

Double ball bearings utilize rolling friction instead of traditional sliding friction, resulting in low friction, no need for lubrication, and good anti-aging performance. They are suitable for high-speed, long-term operation, with a lifespan of 50,000-100,000 hours or even higher. The disadvantage is that they have the highest noise level at the same speed, and are more difficult to manufacture and more expensive.

Hydraulic bearings (oil-impregnated bearing family) use lubricating oil as a drag reducer, resulting in low initial noise and low cost. However, the lubricating oil evaporates over time and at high temperatures, leading to a shorter lifespan, generally between 30,000-50,000 hours. In terms of noise level: double ball bearings > single ball bearings > oil-sealed bearings, but their lifespans are exactly the opposite.

For devices like servers that operate 24/7, dual ball bearings are the most popular choice due to their superior lifespan and reliability.

V. Redundancy and Intelligence: The Design Wisdom of Server Thermal Architecture

At the server level, heat dissipation is not just a matter of a single fan working alone, but a meticulously designed system engineering project.

Redundancy is a core feature of server thermal architecture. A single server typically has multiple independent fans arranged in a push-pull airflow pattern—if any fan fails, adjacent fans automatically increase their speed to compensate for the airflow. Simultaneously, fans support hot-swapping, allowing replacement without interrupting operations if a single fan fails.

Intelligent control further refines heat dissipation. Modern server fans generally support PWM (Pulse Width Modulation) control, automatically adjusting fan speed based on device temperature and connecting to the server’s baseboard management controller for speed monitoring, fault warnings, and temperature control linkage. The control system monitors the speed and current of each fan in real time; if abnormal speed or power consumption deviates from the baseline, it automatically degrades usage and triggers an alarm.

VI. From Air Cooling to Liquid Cooling: The Evolution of Thermal Technology

With the explosive growth of AI computing power, the power consumption of high-performance chips, represented by NVIDIA, is increasing at a rate of doubling generationally. From 1000W for the B200 to 4000W+ for the future Rubin R300 series, the power density per rack has jumped to over 140kW, far exceeding the economical cooling limit of 15kW for air-cooled systems. Against this backdrop, liquid cooling technology is moving from “optional” to “essential.”

Currently, cold plate liquid cooling dominates the liquid cooling solution market—it requires minimal modification to the existing server ecosystem, has controllable costs, and is easy to maintain, currently holding over 70% of the market share. Immersion liquid cooling, on the other hand, is considered the ultimate solution for ultra-high heat flux density scenarios. Single-phase immersion liquid cooling can achieve a PUE as low as below 1.1. Although its current cost is higher, its long-term trend is clear. Meanwhile, emerging phase change materials (PCMs) have also injected new ideas into the field of heat dissipation. They can absorb excess heat during load surges, narrowing temperature fluctuations to within ±2°C, helping to reduce the energy consumption of cooling systems by 15%-20%.

It is worth noting that air cooling will not be completely replaced. In practical applications, the “air-liquid combination” solution has become an important trend—the liquid is responsible for removing most of the heat from the core heat sources (CPU, GPU), while the air cooling system is responsible for handling the remaining heat load, cooling the power supply module, memory modules, and key components in the liquid cooling loop. With its advantages of simple structure, convenient maintenance, and controllable cost, air cooling will coexist with liquid cooling for a long time, jointly building a complete thermal management ecosystem.

VII. Summary and Purchase Recommendations

The small cooling fan bears the heavy responsibility of stable server operation. From a selection perspective, the following points are particularly crucial:

Prioritize matching air pressure: In high-density server applications, air pressure is often more important than airflow. Don’t just look at the nominal airflow parameters; pay more attention to whether the fan’s P-Q curve can penetrate the system’s high-resistance airflow channels.

Bearings determine lifespan: Servers operating 24/7 should prioritize dual ball bearing fans, which can last for tens of thousands or even hundreds of thousands of hours.

Intelligent control support: 4-pin PWM fans can dynamically adjust their speed based on temperature, balancing heat dissipation and noise.

Redundant configuration is key: Prioritize fan modules that support hot-swapping and have redundant backup mechanisms.

Excellent thermal design often lies in achieving the most precise heat transfer with minimal fan overhead—making the most of every airflow.