Measure Mem Reduct Before and After: Test Protocol
Use a repeatable task, paired runs, and a record of all changed settings. Compare task duration and pauses as well as memory readings. This site provides a local worksheet for your own measurements; it does not publish invented benchmark scores or claim tests on your PC.
What matters for this task
A before/after screenshot is easy to produce but can be misleading if the workload changed. Keep the machine, application version, open content, and background activity comparable. A benchmark should also retain unsuccessful and inconclusive runs rather than selecting only the largest memory reduction.
Steps to follow
- Define the task and success metric before starting. Examples include a document-open duration or observed pauses during the same offline game scene, not an unspecified feeling of speed.
- Record a baseline with automatic cleanup disabled. Note physical-memory and commit observations separately and keep the active application list.
- Run one cleanup with documented settings, repeat the task, and record the same metrics. State when readings were taken relative to the cleanup.
- Repeat the paired comparison and use the worksheet to calculate differences. Treat small inconsistent differences as inconclusive and preserve all runs for review.
Check the result
A persuasive result is reproducible, useful to the real task, and larger than normal variability. The worksheet can calculate arithmetic but cannot establish causation.
Limitations and precautions
Do not label a result “tested” without the device, build, method, and actual observations.
Questions and answers
Does this site contain laboratory FPS results?
No. It provides a method and blank tools for recording your own results.
Can an immediate memory reduction still be a poor outcome?
Yes. If the next task takes longer or stutters, the operational result can be worse despite the lower counter.
Sources and evidence
Product statements use the tagged release documentation. Diagnostic steps are proposed checks, not claimed laboratory results.