Lacksan community

Lacksan HP Performance Forum

Workstation tuning, measured and written down.

Read-only preview

Benchmarks

A benchmark procedure you can actually repeat next month

Started · 1 replies · 774 views

  1. Documented fact

    Most enthusiast benchmark posts are unrepeatable six weeks later because the conditions were never written down. A minimum viable procedure:

    • Fixed thermal start: idle at the desktop until package temperature is stable for two minutes.
    • Fixed power state: AC, one named power plan, screen brightness fixed.
    • Fixed background: same set of running applications, updates paused, no browser.
    • Five runs minimum. Discard nothing. Report the median and the range.
    • Record the Windows build every single time, because it will change under you.

    The range matters more than most people think. If your five runs span more than the effect you are claiming, you have not measured the effect.

  2. User observation

    Adding one from bitter experience: record whether the machine has been rebooted since the last run. On our Z8s the first post-reboot run is consistently unrepresentative because indexing and telemetry catch up in the background.

    This is an observation across our fleet rather than a measured study, but it has been consistent enough that we now always discard run one and state that we did.

Replying is not available in this build. Accounts, posting and moderation arrive in the next phase; until then, corrections go to admin@lacksan.com.