Every debloating video promises frames. I wanted to know what I could actually measure on hardware I own, so I ran a before-and-after on one machine and wrote down everything, including the results that were boring.
The bench
Nothing exotic, deliberately — a mid-range desktop of the kind most people actually game on. Windows 11, current feature update, freshly installed, drivers current, then left to sit for two hours so that provisioned apps and initial indexing finished before I measured anything.
- Ryzen-class 6-core CPU, 32 GB DDR4, mid-range GPU, NVMe system drive
- Capture: in-game frame time logs, 3 runs per config, median reported
- Same resolution, same settings preset, same driver version throughout
- Measurements taken 10 minutes after boot, on mains power, no browser open
The "after" state was: recommended config imported, content delivery denied, telemetry at edition minimum, Store bloat removed, Widgets and Copilot removed, notifications trimmed, power plan set to the high-performance profile.
The numbers
| Metric | Before | After | Change |
|---|---|---|---|
| Idle RAM in use | 4.6 GB | 2.9 GB | −37% |
| Background processes at idle | 132 | 88 | −44 |
| Power-on to usable desktop | 47 s | 31 s | −16 s |
| Game A, average FPS | 142 | 145 | +2% |
| Game A, 1% low FPS | 88 | 101 | +15% |
| Game B, average FPS | 96 | 97 | +1% |
| Game B, 1% low FPS | 61 | 70 | +15% |
| Game C (CPU-bound sim), average | 54 | 58 | +7% |
Read that table carefully, because the headline is not in the average column.
Averages barely moved. Lows did.
Average frame rate changed by one to two percent in two of three games, which is inside run-to-run noise. If someone tells you debloating gained them thirty percent average FPS on a modern machine, ask what else they changed.
The 1% lows are a different story. A consistent 15% improvement in the bottom percentile is something you feel, because that percentile is exactly where stutter lives. A background indexer waking up, a telemetry batch uploading, a suggestion service reaching out — none of those cost you average frames. They cost you one bad frame at a bad moment.
Debloating did not make my machine faster. It made it more consistent. That is a better outcome and a worse headline.
Where the wins came from
I re-ran with changes applied one group at a time to find out what mattered. Roughly, in order:
The power profile
The largest single contributor to CPU-bound results, and it has nothing to do with bloat. A balanced plan with aggressive core parking costs frames in simulation-heavy games.
Search indexing and scheduled maintenance
Not disabled, but no longer firing during a session. Most of the stutter reduction traces here.
Content delivery and suggestion services
Small but real. These wake, reach out, and go back to sleep — on a schedule that does not care what you are doing.
Removing Store apps
Measurably nothing. It freed 6 GB of disk and made the Start menu tolerable, which is worth doing for its own sake.
About placebo
I ran the "after" configuration blind for a week without looking at the numbers, and I would have told you the machine felt dramatically faster. The numbers say it is marginally faster and noticeably steadier. Both things are true, and the gap between them is the reason this whole genre of content is so unreliable.
A cleaner Start menu, fewer notifications and a desktop that appears sixteen seconds sooner change how a machine feels far more than three frames per second do. I think that is a legitimate reason to debloat. It is just not a performance claim.
What I would do differently
Next time: more games, at least five runs each, and a proper thermal soak before measurement. I would also log for an hour of ordinary use rather than benchmark loops, because the stutters that annoy people happen when a browser, a chat app and a game are all resident — which is the state no benchmark reproduces.
If you run your own numbers, I would genuinely like to see them. Send them over and I will publish the ones that contradict me.
Written by Joseph Savastano. Measurements come from my own machines and are not a promise about yours. Corrections are welcome and get published — send one.