Skip to content

Benchmark Guide

Jump to bottom
Emalfai edited this page Mar 22, 2026 · 1 revision

Testing & Benchmarks

Winrift includes a built-in performance benchmark that measures real system metrics before and after applying tweaks.

Quick Start

Step 1. Run the benchmark before applying any tweaks:

. .\modules\system\Benchmark.ps1
Invoke-Benchmark -Phase Before

Step 2. Apply your tweaks through the Winrift menu.

Step 3. Run the benchmark after tweaks (reboot recommended first):

. .\modules\system\Benchmark.ps1
Invoke-Benchmark -Phase After

The comparison report is displayed automatically and saved to %USERPROFILE%\Winrift\benchmarks\.

To re-view the last report without collecting new data:

Invoke-Benchmark -Phase Compare

Metrics

Metric Source What It Shows
CPU idle load \Processor(_Total)\% Processor Time Background CPU usage at rest — fewer services = lower idle load
RAM usage Win32_OperatingSystem Physical memory consumed — disabled services free RAM
Committed memory \Memory\Committed Bytes Virtual memory pressure including page file
Running processes Get-Process Total process count — direct measure of background bloat
Running services Get-Service Active Windows services — tweaks disable unnecessary ones
Startup apps Win32_StartupCommand Programs that auto-launch on boot
Scheduled tasks Get-ScheduledTask Active tasks (Ready state) — maintenance tweaks reduce these
Disk read latency \PhysicalDisk(_Total)\Avg. Disk sec/Read Storage response time — SSD/NVMe tweaks improve this
Disk write latency \PhysicalDisk(_Total)\Avg. Disk sec/Write Storage write response time
DPC rate \Processor(_Total)\DPC Rate Deferred Procedure Calls — system latency indicator
Context switches/sec \System\Context Switches/sec CPU scheduling overhead — lower is better
Interrupts/sec \Processor(_Total)\Interrupts/sec Hardware interrupt rate — affected by interrupt steering tweak
Page faults/sec \Memory\Page Faults/sec Memory paging activity — memory optimization impact

All sampled metrics are averaged over 10 readings at 3-second intervals (~30 seconds total).

Interpreting Results

The report shows a comparison table:

 +------------------------------------------------------------------+
 ' Winrift Performance Report                                    '
 +------------------------------------------------------------------+
 '   Metric                    Before       After       Change      '
 +------------------------------------------------------------------+
 '   CPU idle load             3.2%         1.1%        ▼ 66%      '
 '   RAM usage                 2,800 MB     2,100 MB    ▼ 25%      '
 '   Running processes         142          98          ▼ 31%      '
 '   ...                                                            '
 +------------------------------------------------------------------+
  • ▼ indicates a decrease (typically better for all metrics)
  • ▲ indicates an increase (typically worse for resource metrics)
  • Percentage shows relative change from the baseline

Expected Ranges

These are typical improvements on a clean Windows 11 24H2 install:

Metric Typical Improvement
CPU idle load 30–70% reduction
RAM usage 15–30% reduction
Running processes 20–40% reduction
Running services 10–25% reduction
DPC rate 20–40% reduction
Context switches 15–30% reduction

Actual results depend on hardware, installed software, and which tweak categories were applied.

Output Files

All benchmark data is stored in %USERPROFILE%\Winrift\benchmarks\:

File Description
before_{timestamp}.json Raw metrics snapshot before tweaks
after_{timestamp}.json Raw metrics snapshot after tweaks
report_{timestamp}.md Markdown comparison report

Limitations

  • Reboot recommended between before/after snapshots — some tweaks require a restart to take effect
  • Close background apps before measuring — browsers, editors, and other apps affect metrics
  • Hardware-dependent — results vary significantly between systems
  • GPU tweaks are not measured — GPU performance requires specialized tools (FrameView, CapFrameX)
  • Network tweaks — throughput/latency changes require external test servers to measure accurately
  • Single-point measurement — for statistically significant results, run multiple before/after cycles

Pester Tests

Unit tests for the benchmark module are at tests/Benchmark.Tests.ps1. Run them with:

Invoke-Pester ./tests/Benchmark.Tests.ps1

Clone this wiki locally