Thermal Strategy and Cooling Systems Guide

Introduction

Thermal management is the second-largest infrastructure cost after electrical systems and has the greatest impact on operational efficiency. This guide helps engineering teams evaluate cooling strategy options, understand their implications for PUE and WUE, and select the right approach based on site conditions, rack density targets, and operational priorities.

Cooling Strategy Overview

Data center cooling strategies fall into three broad categories, each with distinct cost, efficiency, and complexity profiles. The right choice depends on rack density, climate zone, water availability, and the operator's tolerance for mechanical complexity.

Air-Cooled Systems

Air-cooled systems reject heat to the ambient air using direct expansion (DX) units, computer room air handlers (CRAHs) with dry coolers, or roof-mounted packaged units. They require no water infrastructure and have the simplest maintenance profile, making them well-suited for edge, container, and small modular deployments.

• Typical rack density support: up to 8-12 kW per rack
• PUE range: 1.4-1.8 depending on climate and economizer hours
• Pros: No water dependency, simpler permitting, lower mechanical complexity
• Cons: Higher PUE in warm climates, limited density ceiling, larger physical footprint

Chilled Water Systems

Chilled water systems use centralized chillers to produce cold water that is distributed to air handlers or in-row cooling units throughout the facility. They offer better efficiency at scale and support higher rack densities than air-cooled systems alone.

• Typical rack density support: up to 20-30 kW per rack with in-row units
• PUE range: 1.2-1.5 with free cooling integration
• Pros: Higher density support, better PUE at scale, centralized maintenance
• Cons: Water infrastructure cost, piping complexity, leak risk management

Direct Liquid Cooling

Direct liquid cooling (DLC) brings coolant directly to the heat source at the chip or server level. This includes cold plate systems (where coolant flows through plates attached to CPUs/GPUs) and immersion systems (where servers are submerged in dielectric fluid).

• Typical rack density support: 50-100+ kW per rack
• PUE range: 1.02-1.15
• Pros: Highest density support, best PUE, enables GPU/AI compute workloads
• Cons: Highest capital cost, specialized maintenance, limited vendor ecosystem maturity

PUE Estimation Methodology

Power Usage Effectiveness (PUE) quantifies the ratio of total facility power to IT equipment power. A lower PUE indicates more efficient cooling and power distribution infrastructure. GridCore uses an 8,760-hour annual simulation approach that accounts for hourly ambient conditions, load profiles, and equipment part-load efficiency curves.

Key factors in PUE estimation include: chiller or DX compressor efficiency curves, fan power for air distribution, pump power for fluid distribution, free cooling switchover temperature setpoints, UPS efficiency at operating load point, and lighting and miscellaneous building loads.

Water Usage Effectiveness

For systems using evaporative cooling (cooling towers, evaporative condensers, or direct evaporative economizers), Water Usage Effectiveness (WUE) measures the annual water consumption per unit of IT energy. Sites with water scarcity, high water costs, or sustainability commitments should evaluate WUE alongside PUE.

Climate Zone Adjustments

Cooling system performance varies significantly by geography. This guide uses ASHRAE climate zone classifications to adjust cooling system recommendations. Cool climates (zones 5-8) enable extensive free cooling hours and may justify air-side economization. Warm climates (zones 1-3) favor high-efficiency mechanical cooling with minimal evaporative supplement.

Plant-to-Load Thermal Pathway

Regardless of the cooling strategy selected, the thermal pathway from the central plant to the IT load must be defined with clear interface boundaries, flow requirements, temperature differentials, and redundancy provisions at each stage.

• Heat rejection plant: chillers, cooling towers, dry coolers, or DLC distribution units
• Primary distribution: piping mains, pumps, valves, and isolation points
• Secondary distribution: zone-level air handlers, in-row units, or CDUs
• Tertiary delivery: supply and return airflow paths, containment strategy, or direct-to-chip connections

Containment Strategy

Airflow containment (hot aisle, cold aisle, or chimney) dramatically improves cooling efficiency by preventing supply and return air mixing. The guide covers containment options, their compatibility with different rack configurations, and the impact on fire suppression agent distribution and detection response times.