# Methodology Overview | Compute methodology
Source: https://greenpixie.com/docs/methodology/cloud-compute/methodology-overview

How the methodology works at a high level.

Greenpixie's cloud compute methodology calculates the electricity drawn by the processors, memory, attached drives, and accelerators of public cloud workloads for each unit of cloud usage.

Server vendors publish whole-server power measurements under a public benchmark's run rules, each disclosing the processor, the memory modules installed, and the storage devices of the machine measured. Where a published measurement exists, Greenpixie matches each cloud instance to measurements of servers built on the same or a comparable processor.

A processor a cloud provider designs for its own fleet has no published measurement. For such a processor, Greenpixie takes the measured curve of the closest related processor that does have one. That curve is then scaled to the processor being modelled.

Memory, drives, and accelerators are then added back at the configuration the cloud host is modelled as carrying. One operating point is read from the resulting curve for each unit of cloud usage.

The methodology operates through six parts, being processor power derivation, memory power derivation, accelerator power derivation, local storage power derivation, operating point selection, and billing unit conversion. Section 7 describes each in detail. Together they produce the compute electricity figure for each unit of cloud usage.