Menu
Hardware
Homelab & NAS
Tools & Builders
Prices & Data
Troubleshoot

Inputs

Exact lookup

Choose one measured configuration

Start with a dataset so results from different providers and test scenes stay separate.

Live result

Loading measured results…

What this FPS calculator does

I built this as a benchmark lookup. Its 197 measured rows across seven games stay inside two separately identified datasets:

Pick a setup that was tested, and the page returns the average FPS and 1% low from that exact row. It does not combine the two providers, scale a 1440p result into 4K, substitute a nearby GPU, or fill a missing setting with an estimate — and each result separates the benchmark test date from the source’s publication date, so a TechSpot row whose test date was not stated says it is unspecified.

How to use it

  1. Pick the dataset first. This fixes the provider, test system, and source boundary.
  2. Choose the game, CPU, GPU, resolution, and preset from the remaining supported values.
  3. Read the average and 1% low together with the test date, scene, settings, and source link.
  4. Use Copy result link to keep that exact six-key selection. Reset returns to a supported default measurement.

The menus narrow after each choice. An unsupported shared link clears the form and shows no FPS. Choose Reset to load the supported default, then select a measured setup.

Which results actually state a test date?

The lookup preserves each source separately. A publication date can tell you when a chart appeared; it does not establish the test date when the provider did not report one.

ProviderMeasured rowsGamesTest date statedTest date unspecified
TechFuelHQ5350
TechSpot19240192

These rows are exact lookup tuples, not a prediction model. Do not blend providers, game scenes or rendering settings to fill an unsupported combination.

Reproduce this comparison

To recalculate the published table offline, save the Python script as reproduce.py in an empty folder. Inside that folder, create any subfolders shown below and save each input at its displayed path.

From that folder, run this command with Python 3.10 or later. It prints the rows and input hashes; it does not run the hardware or start a server.

python reproduce.py --data-root . --product fps-calculator-catalogue

Compare the output with this product’s entry in the published JSON and input hashes.

For the November 5, 2026 correction, download the corrected comparison package or the November 4 package with the original files. The tuning labels changed; these comparison tables did not. To check every input and table in either ZIP, extract it, open the extracted folder, and run python benchmark-continuations/reproduce.py --data-root . --check.

To collect a new measurement: Keep provider, chart hash, hardware, scene, preset, resolution, metric and date basis. Validate the source transcription separately from any formula experiment.

Why the dataset selector matters

Benchmark numbers inherit the reviewer’s whole method. A different scene, patch, driver, memory kit, CPU power policy, warm-up path, or run length can change the result even when the game, CPU, and GPU names match, so provider separation keeps each number attached to the system and method that produced it.

TechSpot’s results link to the review and original chart. The TechFuelHQ rows use the first-party dataset’s CC BY 4.0 license, which does not cover the TechSpot rows.

Settings are stated, off, or unspecified

The result card does not turn missing documentation into an on/off claim. It uses three different ideas:

  • Specified on or specified off means the source explicitly establishes that state.
  • Specified means the source names a mode such as DLSS Quality, native resolution, Native TAA, or 100% render scale.
  • Unspecified by source means the available source record does not establish the setting. It does not mean disabled.

This distinction matters most for frame generation, ray tracing, and upscaling because a blank chart label is not enough evidence that a feature was off.

The measured scope

The current catalog has seven games: Rainbow Six Siege X, Kingdom Come: Deliverance II, Marvel Rivals, Warhammer 40,000: Space Marine 2, Cyberpunk 2077, Alan Wake 2, and Black Myth: Wukong. Each menu narrows to the hardware and settings present in its dataset.

Unsupported hardware gets no universal FPS range. Find a benchmark that tests the same game version and settings, or measure the scene on your own system; the CPU and GPU benchmark lookup can compare separate component indexes, but it cannot supply game FPS. If your measured result is much lower than a matching benchmark, work through the low GPU usage checks before treating a component swap as the answer.

Bounded formula experiment

Before choosing exact lookup as the public behavior, I tested one pre-specified formula inside the TechSpot grid. For each chart, the formula predicted a lower-CPU/lower-GPU cell as the smaller of its CPU anchor and GPU anchor; across 24 charts it held out three target CPU×GPU cells per chart for 72 average-FPS and 72 1%-low comparisons, while each target GPU remained present in the separate higher-CPU anchor.

MetricMedian absolute relative error90th percentileMaximumWithin 10%Within 15%Within 20%
Average FPS0.00%2.44%10.44%98.61%100%100%
1% low FPS0.60%3.02%17.07%98.61%98.61%100%

The full 72-row validation table retains each prediction, actual value, signed error, absolute error, and source key. This was a bounded formula experiment on one provider’s two-CPU, four-GPU grid; it neither tests lookup accuracy nor supplies a confidence interval, and it provides no evidence that the formula works for another game, scene, provider, CPU, GPU, or setting. The public calculator therefore uses a measured row when one exists and returns no number otherwise.

Compare with your PC

Average FPS summarizes throughput. The 1% low helps expose slower frames that the average can hide. Compare both within the same source method; neither number guarantees what your own system will produce or describes every moment of gameplay.

Match the game version and scene first. Then match the preset, resolution, render scale, upscaler, ray tracing, frame generation, CPU, GPU, and test conditions. If the source marks a field unspecified, leave it unknown instead of assuming a setting.

The catalog records source dates and chart hashes so the transcription can be checked against the captured originals; the public page includes the structured numeric facts and citations, but not the third-party chart images.

Frequently asked questions

How many FPS will my CPU and GPU get?
This calculator can answer only when the selected dataset contains that exact CPU, GPU, game, resolution, and preset. It returns the source’s measured average and 1% low. If the tuple was not tested, it prints no FPS rather than estimating one.
How accurate is this FPS calculator?
It is an exact lookup of published benchmark rows, not a prediction model. The displayed number accurately reproduces the cited source row, but your PC can differ because of the game version, scene, drivers, memory, cooling, background load, and settings.
Does the result include DLSS, FSR, ray tracing, or frame generation?
Only when the source explicitly documents that setting. The result card distinguishes a stated setting from one the source did not specify. Unspecified never means disabled.
What does 1% low FPS mean?
The 1% low summarizes slower frames in the tested run and helps show smoothness that an average can hide. Use it to compare rows from the same provider and method; it is not a guarantee for every moment of play.

Sources and corrections

Last updated
Methodology
See our methodology for research and review standards. It draws on 2 cited sources, listed below, each checked against the original page on 2026-10-02. The public catalog is generated deterministically from a 192-row independently checked TechSpot chart transcription and five canonical TechFuelHQ first-party rows. The interface returns exact tuple matches only.
Update log
  • 2026-11-05 — Attached the manual GPU profile, rendering conditions and a versioned source download to the five June Astral results.
  • 2026-11-04 — Counted stated and unspecified test dates separately for each provider in the public lookup.
  • 2026-10-02 — Published the exact-match lookup with 197 sourced benchmark rows, explicit provider separation, and numeric-free handling for unsupported combinations.
Corrections
Spotted an error? Email a correction with the affected line and what you found. The draft includes this page’s address. Verified changes are recorded here. Contact details.