Gaming and performance settings: what actually moved the numbers
I benchmarked before and after, on the same drivers, on the same day. Most tweaks did nothing measurable. Three of them did.
How I measured
Same machine, same drivers, same afternoon, same room temperature. Three titles, each run five times per configuration with the first run discarded. I recorded average frame rate, the 1% low, and cold boot time to an interactive desktop. Between configurations I restarted and waited three minutes for background work to settle.
This is a sample of one PC. It is not a review, and your hardware will disagree with mine in places. The reason I am publishing it anyway is that most optimization content publishes no numbers at all.
Test system
8-core CPU · 32 GB DDR4 · current-gen mid-range GPU · NVMe system drive · Windows 11, fully updated, clean install two weeks prior.
The results
| Change | Avg FPS | 1% low | Boot to desktop |
|---|---|---|---|
| Baseline, clean install | 100% | 100% | 24.1 s |
| + Customization & notification changes | 100% | 101% | 23.8 s |
| + Privacy & telemetry toggles | 100% | 102% | 22.6 s |
| + High-performance power profile | 103% | 107% | 22.4 s |
| + Background app and startup cleanup | 103% | 111% | 18.9 s |
| + Update behaviour changes | 103% | 111% | 18.7 s |
Read the 1% low column, not the average. The average frame rate barely moved and I would not claim the 3% as meaningful. The 1% low — the number that corresponds to the stutter you actually feel — moved by eleven percent, and that I will defend.
The three that worked
1. The power profile
On a desktop that is never on battery, letting the CPU sit at a lower minimum state to save a few watts buys you nothing and costs you latency on the first frames after an idle moment. Switching to a performance-oriented profile was the single clearest win, and it is entirely reversible.
2. Startup and background apps
Not glamorous, but it accounts for almost all of the boot-time improvement and a good share of the 1% low gain. Every item that launches at sign-in competes with your game for the same disk and the same first seconds of CPU.
3. Taking control of when updates install
Zero effect on any benchmark. Enormous effect on the experience, because the machine stopped deciding to download a 3 GB update in the middle of a session. Deferral, not disabling — security patches still arrive.
The ones that did nothing
- Disabling visual effects. Measurable on hardware from a decade ago. Not on anything with a modern GPU.
- Removing store apps. They were not running. Disk space, yes. Frames, no.
- Most service tweaks. Idle services consume approximately nothing, and a few of the ones popular scripts disable are needed by things you will miss later.
- Game Mode, on this system. Within run-to-run noise in all three titles. It appears to help more on machines with fewer cores.
If a tweak cannot explain the mechanism by which it would help, it probably does not help. "It frees up resources" is not a mechanism.
What I would tell a friend
Do the power profile and the startup cleanup, take control of update timing, and stop. That is maybe fifteen minutes of work and it captures nearly all of the measurable benefit. Everything beyond it is preference, tidiness or privacy — all legitimate reasons, but not performance ones.
And keep the before-and-after configuration files, so that when a feature update quietly resets your power profile in November you can put it back in a minute rather than rediscovering the problem in a firefight.
Written on my own hardware, with my own money, on my own time. Nobody paid for a mention here. If a page ever does contain a paid placement or an affiliate link, it will say so at the top — see the disclaimer.