Energy Consumption and Carbon Footprint Spreadsheet

Created by CMIP IPO with Scribe

This spreadsheet is distributed to modelling centres participating in CMIP7 AFT and ongoing CMIP7 experiments.

Its purpose is to collect HPC performance information that enables the Task Team to:

  • Estimate energy consumption
  • Calculate the carbon footprint
  • Compare efficiency across platforms and experiments

Please refer to the Guidance notes available on Zenodo.

Quick Start (Experienced Users)

If you already understand how to collect HPC performance metrics:

  1. General Information tab

    • Enter modelling group and contact details
    • Add HPC platform name(s)

  2. HPC centre information

    • For each machine: location, PUE, emission factor, energy mix %, source

  3. Experiment Metrics tab

    • Complete Tier 1 (Strongly recommended)
    • Add Tier 2 (Recommended)
    • Add Tier 3 (Optional)

  4. Add any comments if needed, for example where metrics cannot be provided due to institution regulations.

  5. Use Metrics short descriptions tab as glossary

For definitions or formulas → open the Guidelines at top of the spreadsheet.

Step-by-Step Instructions

General Information Tab

Open the spreadsheet to locate the key guidance links

Step 1 — Open the tab

Select the General Information tab.

Step 2 — Guidelines document (Zenodo)

The Zenodo link provides:

  • Tier 1–3 metric definitions
  • Calculation methods
  • PUE and emission factor explanations and links
  • Best practice
Link to the full Guidelines and Best Practice document.

Step 3 — FAQ and contact

  • FAQ link (updated over time)
  • Task Team support email
FAQ and contact information.

Modelling Group Information

Enter details about your modelling centre

Fill in:

Field What to enter Example
Modelling group name Organisation or group ECMWF
Location City + country Hamburg, Germany
Contact email Person completing form name@email
HPC platform name Primary system name Levante

Screenshots

Modelling group name.
Location and country.
Contact email.
HPC platform name.

HPC Centre Information

Describe each HPC machine

Add one row per machine.

Required information:

Field Description
Platform name Machine identifier
Location City + country
PUE Power Usage Effectiveness
Emission factor Carbon intensity of electricity
Energy mix (%) Source breakdown
Source Link or text reference
Multiple rows allow several machines.

Experiment Metrics Tab

Select Experiment Metrics.

Experiment Metrics tab.
Core inputs required to compute carbon footprint

A Setup = experiments of same type, same mode, same machine.

Enter for each setup:

Field Example
Platform name dropdown
Experiment type Production / Tuning / Discarded
Model name ICON
Simulated years 100
Core hours 13000
Data output (GB) 4000
Energy consumption (MJ) calculated
Performance and efficiency metrics
Metric Example
SYPD 12
QSYPD 10.5
RSYPD 9.5
Parallelisation (cores) 500
Data output intensity (GB/core-hour) 0.05
Number of grid points 1.5E+06

Provides deeper insight into performance and resource use.

Tier 3 — Optional

Advanced/diagnostic metrics
Metric Example
Complexity 32
Data output cost 0.12
Coupling cost 0.12
Memory bloat 50

Complete only if this data is available.

Comments

Add any notes or explanations related to:

  • Assumptions
  • Estimates
  • Missing values
  • Special configurations
Comments section.

Metrics Glossary

Open Metrics short descriptions for definitions of all fields.

Glossary of metric descriptions.

For detailed explanations or formulas, please refer to the Guidelines document in the General Information tab.