PC Benchmark X measures system performance and internal latency using its own synthetic tests. It runs a CPU simulation, three GPU scenarios, a combined load and two latency tests. Results are saved locally; a valid Candidate run can be published to the leaderboard.
How the benchmark works
Section titled “How the benchmark works”- Methodology: preparation, profiles, warm-up and durations.
- CPU and GPU loads: what work is performed and why the desktop resolution does not set the difficulty of the GPU test.
- Latency measurement and PresentMon: the measured path and its limitations.
- How scores are calculated: CPU, GPU, Latency, weights and the role of 1% Low.
- Repeatability check: an archive of eight runs of version 0.5.2 with data and calculations.
- Comparing runs: before/after, spread and version differences.
- Leaderboard: overall, CPU and GPU leaderboard, filters, badges and publishing.
Before you start
Section titled “Before you start”Open BoosterX → Benchmark → Test. Check the selected profile and the displayed run parameters. The program automatically fetches the methodology from the server: available profiles, durations and scoring parameters can be updated without reinstalling the client, as long as the installed version supports these changes.
- Close games and other heavy programs. Wait for updates and app installations to finish.
- On a laptop, connect power and keep the same performance mode throughout the series.
- Between repeats, keep power, cooling, overclocking and driver versions unchanged. When testing a setting, change only that setting.
- Wait for the test to finish and compare saved results, not intermediate numbers.
There are no separate SSD and RAM scores in this suite. Memory and its settings can affect the CPU test and latency, but there is no standalone RAM throughput test here.
What the result means
Section titled “What the result means”In the CPU/GPU/Latency model, the final PC Score combines CPU Score, GPU Score and Core Latency Score. The base weights are 60/20/20, but the effective methodology is determined by the server. Consistency remains an indicator of evenness within a run and is not added as a fourth multiplier to this formula.
For scores, higher is better; for latencies in milliseconds, lower is better. A score is not equal to the FPS of a specific game, the quality of the whole system or a percentage of Windows acceleration. Even a high CPU Score does not guarantee the same position in CS2: the composition of the work and the dependence on memory, driver and GPU differ.
In the archived series of version 0.5.2 on a Ryzen 7 7800X3D and RTX 5070, the spread of the final score was small. This observation applies to the old model and a single PC; it is not a repeatability check of the new formula. Details are preserved in the study.
Help verify the methodology
Section titled “Help verify the methodology”To check repeatability, send full series to admin@boosterx.org. Specify the program and driver versions, hardware and conditions. Before sending, check the files for personal data and state whether the anonymized measurements may be published.
Publishing to the online leaderboard is voluntary. The list of submitted information can be reviewed before publishing. Badges and contacts are available only when the corresponding feature is enabled by the server.
Updated: 2026-10-02. The description of the new model refers to engine 0.8.0; active server parameters may differ from the base examples.