Calculator Methodology
How PCBuildCheck sources hardware specifications and produces bottleneck, FPS, and PSU estimates.
Database updated: 2026-07-18 · Score methodology: 2026.07
Hardware data
CPU and GPU names, clocks, core counts, memory, and power figures are checked against manufacturer product pages where available. PCBuildCheck maintains a documented 0–100 planning index within each component class; those normalized scores are editorial inputs, not official manufacturer ratings or raw lab results.
How estimates are calculated
Bottleneck: the tool compares normalized CPU and GPU scores. The displayed percentage is their relative score gap, not a measured percentage of lost performance.
FPS model 2026.07: each game has a 1080p High reference profile. CPU and GPU scores scale separately, with resolution-sensitive weighting plus quality, supported upscaling, anti-aliasing, RAM and VRAM pressure. Results are shown as an uncertainty range with a planning midpoint and estimated 1% low; they are not measured benchmarks.
PSU: published CPU and GPU power figures are combined with a broad allowance for selected components, then 15–50% headroom is applied and the planning result may be rounded to a common PSU size. Always check GPU-vendor, board-partner, and PSU-model guidance.
Important limitations
- Game patches, drivers, BIOS settings, thermals, RAM, background tasks, power limits, and silicon variation affect performance.
- Laptop parts and factory-overclocked cards can have very different power limits despite similar names.
- FPS figures are model outputs, not measurements collected from the selected hardware.
- Prices are launch references or editorial snapshots, not live offers.
- Use results as a shortlist aid and verify compatibility plus independent benchmarks matching your exact game and settings before purchasing.
Reproducible evidence
Follow one calculation from inputs to decision
This example is generated by the same versioned functions used by the calculators. It is a transparent model walkthrough—not a claim that this hardware was laboratory-tested.
Worked model trace
Fixed inputs
AMD Ryzen 5 5600X · NVIDIA GeForce RTX 4070 · Cyberpunk 2077 · 1440p High · native rendering · FXAA · 16GB DDR4-3200 · NVMe SSD
| Model stage | Multiplier | FPS after stage |
|---|---|---|
| Game reference | — | 105 FPS |
| CPU scale | ×0.948 | 99.53626803383156 FPS |
| GPU scale | ×0.859 | 85.48359310216617 FPS |
| Resolution | ×0.760 | 64.96753075764629 FPS |
| Quality preset | ×1.000 | 64.96753075764629 FPS |
| Upscaling | ×1.000 | 64.96753075764629 FPS |
| Anti-aliasing | ×1.000 | 64.96753075764629 FPS |
| VRAM pressure | ×1.000 | 64.96753075764629 FPS |
| RAM capacity | ×1.000 | 64.96753075764629 FPS |
| RAM speed | ×0.980 | 63.642847871951986 FPS |
| Storage | ×1.000 | 63.642847871951986 FPS |
Estimated range
56–72 FPS
Estimated 1% low
49 FPS
Equivalent frame time
15.63 ms / 20.41 ms
Likely constraint
GPU
Adjusted CPU / GPU index
73 / 69
Relative planning gap
5%
Broad PSU estimate
498W
How to interpret the output: The range is a model output from version 2026.07; the frame-time values are exact reciprocal conversions of that output. The balance gap and PSU estimate answer different planning questions and must not be combined into a benchmark claim.
Repeatable on-device verification protocol
Use this process to test whether the model’s suspected limit appears on your actual PC.
- 1Record game version, driver, BIOS settings, resolution, preset, upscaling and frame cap.
- 2Use one repeatable route or built-in benchmark; discard the shader-compilation warm-up pass.
- 3Capture average FPS, 1% low, frame-time graph, per-core CPU load, GPU load, temperatures and clocks.
- 4Change one variable only: first resolution, then a CPU-heavy setting, then a GPU-heavy setting.
- 5Repeat at least three runs and act only on a consistent limit, not one spike.
Platform-cost evidence to record
A CPU or GPU price alone is not the upgrade cost. Record these items before comparing two paths.
- CPU or GPU street price and return policy
- Motherboard plus required BIOS-update path
- DDR4/DDR5 memory that cannot be reused
- Cooler, mounting kit, case or airflow changes
- PSU capacity, quality and native connector requirement
- Performance evidence from the exact game, scene and target refresh rate
Methodology update log
Reviewed 11 October 2026
- 2026-10-11: Added a code-generated model trace, repeatable frame-time verification protocol and full platform-cost checklist.
- Model versions: FPS 2026.07; hardware scoring 2026.07.
- Evidence boundary: no in-house laboratory benchmark is claimed; model outputs and exact unit conversions are labelled separately.