Power usage effectiveness, or PUE, is the most widely used efficiency metric for data centers. It compares the total energy a facility uses with the energy delivered to the IT equipment. A PUE of 1.0 would mean every kilowatt-hour goes to computing, with nothing spent on cooling, power conversion or lighting.
PUE was introduced by The Green Grid and is now standardized internationally in ISO/IEC 30134-2. It is a useful tool for tracking a single facility's performance over time, but it is easy to calculate inconsistently and easy to misuse in comparisons. This guide explains how to measure it properly and how to improve it.
The definition
PUE = total facility energy / IT equipment energy
Both quantities are energy (kWh), normally measured over a full year. Using energy rather than instantaneous power matters, because cooling energy varies strongly with weather and IT load.
- Total facility energy includes everything serving the data center: IT equipment, cooling plant, air handlers, pumps, humidifiers, UPS and transformer losses, lighting and support spaces that are part of the data center.
- IT equipment energy is the energy consumed by servers, storage, network equipment and other computing devices, ideally measured at the input to the IT equipment.
The reciprocal of PUE, IT energy divided by total energy, is sometimes called data center infrastructure efficiency (DCiE).
Measurement levels
The Green Grid and ISO/IEC 30134-2 describe measurement categories based on where IT energy is metered. Measuring closer to the IT equipment is more accurate because it excludes losses in upstream distribution.
| Measurement point for IT energy | Accuracy | Notes |
|---|---|---|
| UPS output | Basic | Includes PDU and distribution losses as IT |
| PDU output | Intermediate | Excludes PDU transformer losses |
| IT equipment input (rack or server) | Advanced | Most accurate; needs branch circuit or intelligent rack metering |
When reporting PUE, state the measurement point and the period. A PUE calculated from a single snapshot on a cold day at UPS output is not comparable with an annualized figure at the rack.
Common mistakes
- Using instantaneous readings. A winter reading can look far better than the annual result.
- Leaving out loads. Omitting the chiller plant serving the data center, the generator block heaters or the office HVAC that shares a meter understates PUE.
- Counting non-IT loads as IT. Rack fans in rear doors, CDU pumps or in-rack cooling sometimes get counted as IT when they are cooling.
- Ignoring mixed-use buildings. In a multi-tenant building, shared plant energy must be allocated to the data center using a defensible method.
- Comparing across climates and designs without context. A facility in Vancouver and one in Toronto face different weather, and a site with high redundancy at low load will have a higher PUE than an identical site at full load.
What drives PUE up
- Cooling energy: compressors, fans, pumps and cooling towers usually dominate overhead.
- Electrical losses: UPS, transformers and distribution losses, especially at low load.
- Low IT load: fixed overheads are spread over less IT energy, so PUE rises when a facility is lightly loaded.
- Overcooling and bypass air: supplying colder or more air than needed.
- Humidification: steam or electric humidifiers running to meet unnecessarily tight humidity limits.
How to improve PUE
Airflow management
Fixing bypass and recirculation is usually the cheapest improvement. Install blanking panels, seal cable cutouts and floor openings, and contain hot or cold aisles. Once mixing is controlled, supply air temperature can rise and fan speeds can drop. See airflow management and hot and cold aisle containment.
Raise supply temperatures
The ASHRAE TC 9.9 recommended envelope allows server inlet temperatures up to 27 °C (81 °F). Many facilities still run at 18 to 20 °C (64 to 68 °F). Raising supply air and chilled water temperatures improves chiller efficiency and increases economizer hours. Check that every rack inlet stays within the envelope.
Use free cooling
Canada's climate is well suited to economizers. Airside economizers, waterside economizers with plate heat exchangers, and dry coolers can carry much of the load for much of the year, especially when supply temperatures are raised.
Variable speed everywhere
Fans and pumps on variable frequency drives or EC motors, controlled on actual demand, save large amounts of energy because power varies roughly with the cube of speed. Constant-speed CRAC fans running at full flow are a common opportunity.
Optimize the chiller plant
Sequence chillers to run near their best efficiency, reset chilled water and condenser water temperatures, and avoid running more pumps and towers than needed. See chilled water plant design for data centers.
Relax humidity control
Current ASHRAE guidance allows a wider humidity range than older practice. Removing unnecessary humidification and dehumidification, and stopping CRAC units from fighting each other, cuts energy. See humidity control and static electricity.
Electrical efficiency
High-efficiency UPS modes, right-sizing UPS modules to load, and efficient transformers reduce losses.
PUE in context
PUE does not measure computing output, carbon intensity or water use. A facility can reduce PUE by switching to evaporative cooling while using a lot of water, so pair PUE with water usage effectiveness (WUE). Replacing server fans with liquid cooling reduces IT energy and can make PUE look slightly worse even though total energy falls. Track total energy alongside PUE.
ASHRAE Standard 90.4 sets energy requirements for data centers using mechanical and electrical loss components rather than a single PUE target. Some Canadian jurisdictions reference ASHRAE 90.1 or the NECB for buildings, and requirements for data center spaces vary by province.
Frequently asked questions
What is a good PUE?
It depends on climate, density, redundancy and load. Older enterprise rooms often operate well above modern purpose-built facilities. The more useful question is whether your PUE is improving over time at similar load.
How often should PUE be measured?
Continuously, if metering allows, with results reported as rolling 12-month values. Monthly values help show seasonal patterns and the effect of changes.
Key takeaways
- PUE is total facility energy divided by IT energy, normally over a year.
- State the measurement point and period; snapshots are misleading.
- Cooling is usually the largest overhead, so airflow management, warmer setpoints and free cooling deliver the biggest gains.
- Low IT load raises PUE because fixed losses are spread over less IT energy.
- Use PUE alongside WUE, carbon and total energy, not as the only measure.
Talk to an engineer
Have a project or a problem this guide touches on? Tell us about the building and we’ll be in touch within one hour.