BottleneckCalculator.biz
Check my PC
Measurement playbook

Build a Performance Baseline You Can Trust After Every Change

Short answer: A baseline is a saved description of how the PC performs under known conditions. It turns a vague impression such as “the game feels slower” into a comparison that can survive a driver update, settings change, repair or hardware upgrade.

CPU, GPU and PC performance stages shown as one connected system

Define one workload, target and repeatable route

Name the application and version, scene or project, resolution, internal render scale, quality preset and target. For games, record display refresh, frame cap, V-Sync, upscaling and frame generation. For batch work, define the input file, output settings and stage whose waiting time matters.

Choose a built-in benchmark when it represents the real problem, or record a manual route that can be repeated closely. Use the diagnostic checklist to preserve the target and system state. A busy recurring scene is more useful than an easy route selected because it produces a flattering number.

  • Record application, operating-system and driver versions.
  • Save exact graphics, codec, model or project settings.
  • Describe the route, duration and warm-up condition.
  • State the experience target before looking at the result.

Capture delivery, operating state and capacity together

For games, save average FPS, 1% low and the raw frame-time trace. Add CPU and GPU duration where available, per-core activity, effective clocks, temperatures, power and RAM/VRAM allocation. The frame-time guide explains why a stable average can hide disruptive slow frames.

For rendering, exporting, compiling or AI, split the job into meaningful stages. Record load time, interactive responsiveness, steady throughput and total completion time separately. An upgrade that accelerates export but leaves the slow timeline unchanged should not be credited with solving the original problem.

  • Keep one lightweight logger and a consistent sampling interval.
  • Capture ambient or starting temperature when sustained boost matters.
  • Save individual runs rather than only their arithmetic average.
  • Exclude interrupted runs and write down why they were rejected.

Measure normal variance before claiming an improvement

Run the unchanged baseline at least three times after a defined warm-up. Compare the spread and use the median as a practical center. The 20-minute testing guide shows why repeated, comparable captures are necessary before a small change becomes convincing evidence.

Store the baseline before troubleshooting PC stutter or deciding between a CPU and GPU upgrade. After a change, repeat the original run before altering anything else. This keeps a genuine gain from being mixed with a new preset, route or cache state.

  • Use the same warm-up pass and test order.
  • Compare medians, distributions and visible trace changes.
  • Require the improvement to exceed ordinary result spread.
  • Keep a dated known-good result for future regression checks.

Choose the next test from the evidence

ScenarioSignal or scaleBest next check
Gaming throughputAverage FPS plus raw frame timeSame route, cap and warm-up
Gaming consistency1% low and spike distributionExclude loading and menus consistently
Interactive creationDropped previews or response latencySame project and cache state
Batch workPer-stage and total completion timeSame input, output and software backend
Sustained operationClock, temperature and power timelineSame duration and starting condition
Use tools to test a question, not manufacture certainty.

Start with the CPU and GPU bottleneck calculator, convert refresh targets with the FPS to frame-time calculator, or work through the diagnostic checklist.

Focused evidence

Continue with a specific guide

View the complete PC bottleneck guide →