How PC Balance estimates a bottleneck
A bottleneck percentage is not an industry-standard measurement. PC Balance treats it as a comparative estimate: which component is more likely to limit the selected workload, how large the mismatch appears, and how much confidence the available data supports.
The short version
The gaming model compares normalized CPU and GPU pairing strength on a common internal scale, applies resolution behavior, then adjusts context for target FPS. A dead zone prevents tiny score differences from becoming fake bottlenecks. Separate workload profiles are used for streaming, video editing, and productivity.
The displayed result is deliberately described as an estimate, and the confidence layer widens the expected range when the underlying hardware data is weaker.
Inputs the model uses
Component profiles
- CPU gaming and pairing strength
- CPU single-core and multi-core context
- GPU raster gaming strength
- GPU render and ray-tracing profiles where available
User context
- 1080p, 1440p, or 4K resolution
- Gaming, streaming, editing, or productivity workload
- Optional target FPS
- Optional RAM capacity
How resolution is handled
Resolution does not simply add or subtract a fixed percentage. For gaming, PC Balance uses the selected GPU's stored native-resolution scaling when that benchmark coverage exists. Higher resolutions generally increase GPU rendering pressure; lower resolutions and high frame-rate targets can expose CPU limits more often.
A contextual resolution adjustment can reduce an existing mismatch toward balanced, but it is not allowed to invent a new opposite-side bottleneck by itself.
Confidence and expected range
Not every processor and graphics card has equally strong modern benchmark coverage. The confidence engine scores the evidence behind the selected CPU and GPU, whether native multi-resolution GPU data is available, how old the hardware is, and whether optional FPS/RAM context was supplied.
- Validated: directly calibrated against a common benchmark reference.
- Derived: based on a closely related validated sibling or family reference.
- Legacy bridged: measured in an older benchmark hierarchy and mapped into the current comparison domain.
- Provisional calibrated: useful coverage with explicitly lower confidence while direct same-suite data is unavailable.
The current CPU database includes 138 provisional-calibrated entries; the GPU database includes 55 cards with directly validated raster status. Those labels are used by the confidence system instead of being hidden from the user.
Validation process
Every engine revision is checked against known anchor pairings and against the full hardware matrix. With the current database, that means 92,070 CPU/GPU/resolution checks for non-finite values, impossible percentage ranges, and resolution-only side flips, plus targeted regression cases that previously exposed bugs.
Passing the matrix tests means the model is internally consistent. It does not mean every game will produce the same percentage; game engines, settings, drivers, cooling, background tasks, and memory configuration still matter.
Public reference sources
These sources inform hardware calibration, scaling behavior, specifications, or model sanity checks. PC Balance does not copy another calculator's proprietary formula.
Important limitations
- There is no universal industry-standard bottleneck percentage.
- Game-specific CPU behavior can differ substantially even at the same resolution.
- Ray tracing, upscaling, frame generation, mods, and graphics presets can change the limiting side.
- Thermal throttling, power limits, drivers, BIOS settings, and background software are outside this static model.
- Older provisional hardware should be interpreted with a wider confidence range.
Shopping and retailer disclosure
Retailer buttons on PC Balance currently open normal search-result pages for the selected or recommended hardware. They do not represent live prices, stock, guaranteed compatibility, or a claim that a retailer is the best place to buy. If affiliate relationships are introduced later, qualifying links will be identified clearly and this disclosure will be updated.
Review policy
Methodology changes are regression-tested before deployment. The public methodology date is updated when the model or data policy receives a meaningful review rather than automatically changing on every site build.
Last reviewed: August 13, 2026