CPU or GPU bottleneck? Use evidence, not guesswork.
Enter observations from two runs of the same game scene. A change in render resolution, combined with GPU activity, helps identify likely constraints without inventing universal bottleneck percentages.
Enter your measurements
Test the exact same scene twice. Keep game settings unchanged except render resolution or render scale. Record average FPS, not a single instant.
Try examples: · ·
No install required. Values are processed in your browser; this page cannot scan your PC.
Enter two FPS measurements.
First test at your normal render resolution. Then reduce render resolution while keeping everything else as similar as possible.
- Shows the observed FPS change
- Considers GPU utilization and frame caps
- Explains when the evidence is not decisive
Three steps for a cleaner bottleneck test
- Establish a baseline. Run the same scene at your normal display resolution. Disable V-Sync, frame-rate caps and frame generation during comparison where possible. Note average FPS and GPU usage.
- Reduce GPU rendering demand. Lower the game’s internal render scale or resolution. Do not change CPU-intensive settings, graphics preset, player position or scene conditions between runs.
- Repeat and interpret cautiously. A clear FPS gain while the GPU is heavily loaded points toward GPU rendering pressure. Little FPS change with low GPU utilization can point toward CPU, engine, a cap or another constraint. Compare per-frame CPU/GPU timing for stronger evidence.
Why model-name calculators cannot prove a bottleneck
Performance depends on the game engine, scene complexity, target FPS, ray tracing, memory, drivers and cooling. A CPU that limits an esports game at 1080p may be entirely sufficient for a graphics-heavy game at 4K. A single percentage from two product names does not express that variation.
Limitations and safety checks
Unstable test scenes, background activity, thermal throttling, dynamic resolution, upscaling, V-Sync and game FPS caps can invalidate the comparison. Utilization is only one signal; GPU busy time, per-thread CPU activity and frame-time traces can provide stronger evidence. The practical thresholds on this page are not derived from a controlled validation dataset.
Benchmark and manufacturer references
Read Intel Graphics Performance Analyzers guidance, Tom’s Hardware GPU hierarchy, and Tom’s Hardware CPU hierarchy. These are external sources; individual results differ from your system.