From 89602924d02517a89aae9b94d806b80209d551cf Mon Sep 17 00:00:00 2001 From: zhouzhou626 <255877794+zhouzhou626@users.noreply.github.com> Date: Wed, 20 May 2026 03:42:26 +0800 Subject: [PATCH 1/2] docs: document hardware measurement scope --- README.md | 2 ++ docs/explanation/methodology.md | 7 +++++++ 2 files changed, 9 insertions(+) diff --git a/README.md b/README.md index 08e3135a0..ace045914 100644 --- a/README.md +++ b/README.md @@ -76,6 +76,8 @@ Configuration precedence and examples: [configuration guide](https://docs.codeca We created a Python package that estimates your hardware electricity power consumption (GPU + CPU + RAM) and we apply to it the carbon intensity of the region where the computing is done. +CodeCarbon focuses on the main compute components it can measure or estimate directly. It does not separately model disk I/O, network transfers, displays, cooling, or other peripherals. + ![calculation Summary](docs/images/calculation.png) We explain more about this calculation in the [**Methodology**](https://docs.codecarbon.io/latest/explanation/methodology/) section of the documentation. diff --git a/docs/explanation/methodology.md b/docs/explanation/methodology.md index 673a70c3e..b85de7041 100644 --- a/docs/explanation/methodology.md +++ b/docs/explanation/methodology.md @@ -85,6 +85,13 @@ intervals. This is a configurable parameter `measure_power_secs`, with default value 15 seconds, that can be passed when instantiating the emissions tracker. +CodeCarbon focuses on the main compute components it can measure or +estimate directly: CPU, GPU, and RAM. It does not separately model disk +I/O, network transfers, displays, cooling, or other peripherals. For +most local code-level experiments, those sources are usually smaller +than CPU, GPU, and RAM consumption, but they can matter for workloads +that are dominated by data movement, storage, or distributed systems. + Currently, the package supports the following hardware infrastructure. ### Tracking Modes From ad4c52d54b0c8f0cbd36452e1c4cb6602c0e076c Mon Sep 17 00:00:00 2001 From: zhouzhou626 <255877794+zhouzhou626@users.noreply.github.com> Date: Wed, 20 May 2026 14:19:33 +0800 Subject: [PATCH 2/2] docs: clarify hardware scope rationale --- README.md | 2 +- docs/explanation/methodology.md | 10 ++++++---- 2 files changed, 7 insertions(+), 5 deletions(-) diff --git a/README.md b/README.md index ace045914..4e61fdf40 100644 --- a/README.md +++ b/README.md @@ -76,7 +76,7 @@ Configuration precedence and examples: [configuration guide](https://docs.codeca We created a Python package that estimates your hardware electricity power consumption (GPU + CPU + RAM) and we apply to it the carbon intensity of the region where the computing is done. -CodeCarbon focuses on the main compute components it can measure or estimate directly. It does not separately model disk I/O, network transfers, displays, cooling, or other peripherals. +CodeCarbon focuses on the main compute components it can measure or estimate directly: CPU, GPU, and RAM. It does not separately model disk I/O, network transfers, displays, cooling, or other peripherals because those sources are usually much smaller for local code-level experiments and are not exposed through the same low-overhead measurement interfaces. ![calculation Summary](docs/images/calculation.png) diff --git a/docs/explanation/methodology.md b/docs/explanation/methodology.md index b85de7041..968d55dc1 100644 --- a/docs/explanation/methodology.md +++ b/docs/explanation/methodology.md @@ -87,10 +87,12 @@ emissions tracker. CodeCarbon focuses on the main compute components it can measure or estimate directly: CPU, GPU, and RAM. It does not separately model disk -I/O, network transfers, displays, cooling, or other peripherals. For -most local code-level experiments, those sources are usually smaller -than CPU, GPU, and RAM consumption, but they can matter for workloads -that are dominated by data movement, storage, or distributed systems. +I/O, network transfers, displays, cooling, or other peripherals because +those sources are usually much smaller, and often negligible, for local +code-level experiments. They are also not exposed through the same +low-overhead measurement interfaces as CPU, GPU, and RAM. However, they +can matter for workloads dominated by data movement, storage, or +distributed systems. Currently, the package supports the following hardware infrastructure.