Where the data comes from
Component performance scores are sourced from PassMark CPU and GPU benchmarks, cross-referenced against real-world FPS data from trusted benchmark channels. Each CPU and GPU in the database has a normalised performance score that accounts for both synthetic benchmark performance and real-world game loads. We weight single-threaded performance more heavily for gaming workloads, and multi-threaded performance more heavily for rendering and streaming workloads.
How the bottleneck percentage is calculated
We calculate the performance ratio between your CPU and GPU at your selected resolution:
- At 1080p: CPU weight 60%, GPU weight 40%
- At 1440p: CPU weight 40%, GPU weight 60%
- At 4K: CPU weight 20%, GPU weight 80%
The bottleneck percentage represents how much the weaker component is limiting the stronger one. A 20% CPU bottleneck means your CPU is delivering 20% less throughput than your GPU can consume — upgrading the CPU would unlock roughly that much additional performance.
How RAM speed affects the result
RAM speed has a significant impact on bottleneck percentage at 1080p. A system running DDR4-2133 can show a 15 to 20% larger CPU bottleneck than the same system running DDR4-3600 with XMP enabled. This is because faster RAM reduces CPU memory latency and increases data throughput to the GPU. The calculator accounts for this when you select your RAM speed.
How often we update
The database is updated monthly. Major GPU and CPU launches are added within 2 weeks of sufficient public benchmark data becoming available. The current database was last updated June 2026 and covers 760 CPUs and 96 GPUs, including hardware through RTX 50-series and RX 9000-series.
Accuracy and limitations
Accuracy is highest for popular component combinations that appear in hundreds of real-world benchmarks. Accuracy decreases for rare or very new hardware with limited public data. No online calculator can replicate the exact performance of your specific system — your game engine, driver version, RAM timings, and cooling all affect real results. Use this calculator for upgrade planning before a purchase, not as a live system diagnostic.
On our published Afterburner validation set (three correctly configured builds × three games at 1440p Ultra), mean absolute error (MAE) is 6.2 percentage points between the calculator bottleneck estimate and implied bottleneck from measured GPU utilization (n=9). Typical AAA titles without CPU-extreme engines land near 5.0 pp MAE. Best-case matches are often within 2–4 points; intentional BIOS misconfig (XMP/ReBAR off) jumps to ~14 pp MAE and is excluded from the primary figure. Treat results as planning guidance and verify on your PC.
Read the full validation write-up: Are PC bottleneck calculators accurate?
Utilization self-check matrix
After you run the calculator, confirm the story on your PC with MSI Afterburner + RivaTuner (or similar). These OSD patterns match the teaching thresholds in our free-software bottleneck check guide:
| OSD pattern | What it usually means |
|---|---|
| GPU ≥95%, CPU <80% | GPU-limited / healthy for that resolution — the card is the ceiling. |
| GPU <85%, CPU ≥90% | CPU bottleneck — the processor cannot feed the GPU fast enough. |
| Both CPU and GPU relatively low | Often settings, drivers, or BIOS (XMP/ReBAR) — not a simple CPU vs GPU mismatch. |
Step-by-step monitoring: How to check for a PC bottleneck using free software.
Methodology FAQ
Where do CPU and GPU scores come from?
Primarily PassMark CPU and GPU benchmarks, cross-checked against real-world FPS data from trusted review channels, then normalised for gaming vs multi-thread workloads.
Why does resolution change the bottleneck percentage?
Higher resolutions shift work to the GPU. We weight CPU/GPU at 60/40 (1080p), 40/60 (1440p), and 20/80 (4K) so the percentage reflects how games actually load each chip.
Is the result as accurate as Afterburner?
No live OSD can be replaced by an online estimate. Our published validation set shows MAE of 6.2 percentage points (5.0 pp on typical AAA titles). Treat the calculator as planning guidance and verify on your PC.
Does RAM speed matter?
Yes — especially at 1080p. Slow RAM can inflate a CPU bottleneck by roughly 15–20% versus a fast XMP kit on the same CPU/GPU pair.
How often is the database updated?
Monthly, with major launches added within about two weeks of enough public benchmarks. The live stats line on this page shows the current CPU/GPU counts and last update date.
Ready to check your build?
Run the bottleneck calculator →