DC cooling challenges due to increasing rack power density

For the past two decades, mainstream servers had consistent power and cooling needs, making it easier to plan and design data centers. This stability has helped attract investment and allowed organizations to keep using data centers beyond their design lifespan without major upgrades. Data center designers were able to plan for an average of 4-6 kW of power per rack, following industry guidelines.

However, the rapid increase in IT power density means that future power and cooling assumptions may deviate from these standards. This creates technical and business risks, as incorrect design assumptions can lead to a data center becoming obsolete or overspending on underutilized capacity. The uncertainty creates a challenge for engineers and decision-makers who must grapple with the technical requirements of data centers for the next 10-15 years.

Data center cooling challenges that data center designers will face

Uptime Institute research shows a consistent increase in rack power density over the past decade, which is expected to accelerate further. This trend is driven by higher power consumption per server, especially with the emergence of higher-powered server processors. As semiconductor technology approaches its physical limits, power delivery, and thermal management will be major challenges.

IT silicon demand is driving up server power and rack power density, with extreme-density racks becoming more common in technical computing and AI. Industry standards have been established around air cooling, but the trend toward higher rack power density is challenging existing facility design assumptions.
Soon, mainstream servers are expected to consume more power due to the increasing demand for high-powered processors, with some product roadmaps projecting 500W to 600W TDP processors. This escalation in power consumption will put a strain not only on cooling but also on power delivery within the server chassis.
The cooling challenges that mainstream servers will encounter as their power consumption increases can be predicted by studying servers used in high-performance computing (HPC) applications. HPC accelerators, including graphics processing units, can consume hundreds of watts each at peak power, adding to the power consumption of server processors, memory modules, and other electronics.
Managing the temperature of high-density IT and upcoming next-generation servers is becoming increasingly challenging. Many older data centers may not be able to provide the necessary airflow to cool them. In addition, the restricted temperatures required by high-density racks demand higher cooling power, which can negatively impact IT performance as modern processors will throttle upon exceeding temperature limits.

How to solve DC cooling challenges to increase power density?

The data center cooling challenge is actually a heat removal challenge. Here are some methods data center designers can adopt to restrict temperatures as required by high-density racks.

Redesign traditionally static systems to handle varying densities and variable power draws. Use a cooling system that captures and removes heat at its source for a hyper-scalable and ultra-efficient environment that dynamically adapts to IT loads.
Decouple space from power using a mechanical, electrical, and plumbing (MEP) design to support density increases vertically within the rack or horizontally with additional rack positions. Supplement the data hall with additional fan arrays without impacting live load to support higher watts per square foot of capacity without requiring additional cooling or wasting data center space.
Adopting modular cooling systems can simplify the addition of more cooling arrays as data center density grows, effectively boosting cooling capacity as needed. With adaptive and modular cooling technology in place, a company can expand its power density without wasting unused space or capacity.

As high-performance workloads demand higher power density per rack, it is crucial to invest in the right cooling technology. By doing so, companies can save data center space and reduce energy costs through increased efficiencies. As we move forward, it is imperative to adopt innovative approaches to cooling that are scalable, flexible, and adaptable to meet the ever-evolving needs of data centers.