Server Energy Efficiency: Five Key Insights

Overview

This document is UII Briefing Report 98, dated April 2023, prepared by T. Rahkonen and J. Dietrich and published by Uptime Institute Intelligence. The report quantifies server energy-saving potential through processor utilisation, hardware refresh and power management. The report sits behind a registration form or subscription; the information in this record rests only on the key-insights summary on the public landing page, and the full text was not read.

The analysis is based on server data from The Green Grid's SERT database. According to the findings, server efficiency generally increases by 50% when processor utilisation is doubled from low levels (20% to 30%). This shows how costly low-utilisation servers are in energy terms and explains why workload consolidation matters. This finding also shows why utilisation should be an input to IT load estimates.

Power management can improve server efficiency by 10% to 15%. A two-generation refresh yields efficiency gains of 150% to 300% for AMD and 50% to 100% for Intel. However, a one-to-one migration to newer servers may yield little efficiency gain unless the new servers carry larger workloads. The gain therefore depends less on the hardware itself than on how workload placement changes.

The record contains only these summary-level figures; method details, sample and breakdowns are in the full report. Because the figures differ by processor vendor, they should be re-evaluated for the server family in use before being applied to a facility. Even so, the report is a useful reminder that IT load depends not only on installed capacity but also on utilisation and power-management settings.

Key points

1. Server efficiency generally increases by 50% when processor utilisation is doubled from low levels (20% to 30%).
2. Power management can improve server efficiency by 10% to 15%.
3. Two-generation refresh: 150% to 300% (AMD) and 50% to 100% (Intel) efficiency gains.
4. One-to-one migration to newer servers may yield little efficiency gain unless they carry larger workloads.
5. The report is UII Briefing Report 98, April 2023.
6. The analysis is based on The Green Grid SERT database server data.
7. The report sits behind registration or subscription; the record rests on the public page summary.

Using it in design

• In the IT load estimator, add utilisation and power-management inputs that affect IT kW.
• In refresh scenarios, compute the gain separately by server generation and vendor.
• Warn the user that a one-to-one replacement may not give enough gain on its own.
• Reflect workload consolidation and utilisation increases in the IT load and cooling demand calculation.

Who should read it

Data centre designers planning IT load, energy managers and operators deciding on server refresh.

Publication details

Authors

T. Rahkonen; J. Dietrich

Publisher

Uptime Institute Intelligence

Year

2023

Document number

UII Briefing Report 98

Document language

English

Source type

Industry body

Access

Citation only

Licence

Copyright retained; citation only — the publisher’s terms apply to the original document

Citation