Measured, not promised. How the bench works.
Every figure on a solutions page links back here, to the procedure that produced it and the date it ran.
eclean does not promise numbers. When a page shows one, it is a record of a run on our own documented machine, published with its method. No recorded run, no number.
The rules every run follows.
- The machine is documented with each run: processor, graphics, memory, storage, and the exact Windows build.
- Both states are measured the same day, on the same machine, from a fresh boot each time.
- Measurements repeat several times and the published figure is the median run.
- Figures are published verbatim with their run date, and they stay tied to it. A figure is never restated without a new run.
- Your hardware is not our hardware. A published figure records what happened on the bench, never what will happen on your PC.
Startup and boot
- What is measured
- The number of enabled startup apps before and after a review in Booster, and the time from power button to a settled desktop.
- How
- Both states are counted from the same list Booster and Task Manager show. Boot timing comes from a Windows Performance Recorder boot trace, and the run notes state the trace settings.
- What gets published
- Both counts, both times, and the machine they came from.
Frame consistency
- What is measured
- Frame time behavior in a fixed, repeatable scene of a named game, before and after applying a named set of tweaks.
- How
- PresentMon captures the run. We look at 1 percent low frame times and the count of frames beyond double the median, which is where a stutter lives.
- What gets published
- The 1 percent lows and spike counts for both states. Never a promised FPS gain, because average FPS belongs to the hardware, not to Windows settings.
Background memory
- What is measured
- Resident memory in use at an idle desktop after a fresh boot, before and after trimming the background load.
- How
- The machine sits idle for a stated number of minutes after login before the reading, so startup work has finished. The figure is the one Task Manager reports.
- What gets published
- Both readings and the idle wait used.
Disk space
- What is measured
- Free space on the system drive before and after one cleanup pass.
- How
- The pass uses only what the run names (the clutter categories selected, hibernation disabled or not), and the freed amount is checked against Windows' own free-space figure, not only the preview's estimate.
- What gets published
- Free space before and after, and what the pass included.
DNS lookups
- What is measured
- Resolver response times before and after switching resolvers with Booster.
- How
- Booster's own benchmark runs several passes and the run records the median, the location, and the connection type, because lookup times are a property of a place as much as of a resolver.
- What gets published
- Median response times for both resolvers, with location and connection stated. Faster lookups are the claim, never lower in-game ping.
Battery
- What is measured
- Estimated full-charge runtime across an identical workload, before and after a change that claims battery impact.
- How
- Windows' own battery report (powercfg /batteryreport) supplies the estimates after repeated identical sessions. Battery numbers are noisy, so a figure is only published when the change is larger than the spread between repeat runs.
- What gets published
- Both estimates, the workload, and the number of sessions behind them.
Hold us to it.
Figures produced this way appear on the solutions pages, each with its run date beside it. If a published figure looks wrong, or a rerun on comparable hardware tells a different story, tell us and we will rerun it or take it down.