Lacksan community

Lacksan HP Performance Forum

Workstation tuning, measured and written down.

Read-only preview

Benchmarks

Reading the UX-ROM results honestly: one sample is not a trend

Pinned Started · 2 replies · 1640 views

  1. Measured result

    The automated UX-ROM runs publish themselves, which means they publish their weak results too. That is intentional, but it does require readers to understand what the numbers can support.

    Take the most recent experimental run. It applied one reversible change and recorded one comparison:

    FieldValue
    MachineHP ZBook Firefly 14 G8
    Windows build26200
    Change appliedHP System Information manual-start functional validation
    Samples per window1
    Window lengthapproximately 120 seconds each
    PowerAC, plan not recorded
    Median processor use before18.5%
    Median processor use after18%

    A 0.5 percentage point shift on a single 120 second sample is inside the noise of an idle Windows desktop. It is not evidence that anything improved. The run is published because the procedure ran correctly, not because the result is meaningful. Full published run.

    Do not quote this figure as a Lacksan performance claim. It is a single sample from an experimental run and the published post says so explicitly.

  2. Hypothesis

    The obvious fix is more runs, but I suspect the bigger problem is the window length rather than the count. Two minutes of idle on a modern Windows desktop is dominated by whatever maintenance task happens to wake up.

    Hypothesis, untested: repeating the same 120 second window five times would produce a spread wide enough to swallow the entire reported difference. If someone has the harness set up, that is a cheap experiment and it would settle the question.

  3. Documented fact

    Agreed that it is the right next experiment, and it is queued. The diagnostic-only runs are useful here as a control, because they changed nothing at all and still show movement between windows. Diagnostic-only run.

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