PC Bottleneck Diagnostic Checklist

Why the checklist starts with a target

Write down one clear outcome

Every system has a slowest stage. It becomes a useful diagnosis only when that stage prevents a defined outcome. Record the exact game, application, resolution, settings and frame-rate or completion-time goal so every later test answers the same question.

A repeatable scene reduces noise. Built-in benchmarks are convenient, but a saved camera path or representative project can be more relevant. Warm the workload, keep background activity stable and repeat captures before treating a small difference as real.

Capture the entire performance pipeline

Save signals from the same run

CPU and GPU percentages are not enough. Record per-core activity, GPU utilization, effective clocks, temperatures, power or limit flags, RAM and VRAM pressure, storage activity and frame time. For productivity work, measure the duration of the exact decode, preview, render, compile or inference stage.

• Use the same monitoring interval for every run.
• Separate memory allocation from active use.
• Document frame caps and synchronization settings.
• Save software, firmware and driver versions with the baseline.

Change one variable and look for the predicted response

Reject a cause that does not respond

Lower resolution to test graphics sensitivity, reduce a CPU-heavy simulation setting to test processor pressure, close background tasks to test contention, and compare early versus warmed-up runs for thermal behavior. Restore the baseline before moving to the next variable.

If the suspected component does not respond, reject or refine the hypothesis. The goal is not to prove the calculator correct. The goal is to find a repeatable cause before money is spent.