Benchmark

PCBenchmarkX measures system performance and internal latency using synthetic tests. The BoosterX benchmark runs its own CPU simulation, three GPU scenarios, a combined load and a separate latency test. Results are saved locally; compatible runs can be compared.

How the benchmark works

• Methodology: preparation, warm-up, durations and data collection.
• CPU and GPU loads: computational work, three rounds and internal resolution.
• Latency measurement and PresentMon: software signal, QPC, D3D12 and output events.
• How scores are calculated: source metrics, reference values, formulas and weights.
• Repeatability check: eight runs, reboots, three resolutions, CSV and reproducible calculation.
• Comparing runs: how to separate a change from normal variance.
• Rating: publication and comparability of results.

Before you start

• Close games and other programs that put a load on the computer. Wait for software installations and updates to finish.
• If you have a laptop, connect it to the charger and do not disconnect it until the tests are finished.
• Between repeated tests, do not change the Windows power mode, cooling settings, overclocking or drivers. If you are checking the effect of a specific setting, change only that one.
• Wait for the test to finish. Compare saved results, not the intermediate numbers on the screen.

There are no separate RAM and disk tests in the described set of engine 0.5.2. Memory and driver performance can affect the CPU/GPU scenarios, but it is not measured here with a separate SSD or RAM score.

What the result means

PC Score combines Performance, Core Latency Score and Consistency. The higher the scores, the better the result. For latencies in milliseconds it is the opposite: a lower value is better. The final score is not the FPS of a specific game or a percentage of Windows acceleration.

In the published series on Ryzen 7 7800X3D and RTX 5070, the coefficient of variation (CV) of the final score of six runs at 1920×1080 was 0.256%. This result applies to one configuration; on other PCs the variance may differ. All measurements and limitations are given in the repeatability check.

Help verify the methodology

Send complete series of results to admin@boosterx.org. With your permission, we will check the results and add them to the article without personal data. What to attach to the series.

Submitting a result to the online rating is voluntary. Before submitting, check what information about the computer will be published.

Verified for engine 0.5.2: 2026-09-20.