Bench Lab
PC Benchmarking Guide: How to Test a New Gaming PC Without Fooling Yourself
A clean benchmarking routine for gaming PCs: 3DMark, Cinebench, OCCT, HWiNFO, CapFrameX, FrameView and real-game testing.
Edited and published by Christian Kirk, Founder and editor
Research basis Source-backed; hands-on only where explicitly shown
Photo by Atahan Demir on Pexels
Use 3DMark for a GPU baseline, Cinebench for CPU and cooling, OCCT for stability, HWiNFO for sensors, CapFrameX or FrameView for real-game frame times, and CrystalDiskMark for SSD sanity checks. Compare only the same benchmark, same settings and similar CPU/GPU combinations.
CapFrameX and NVIDIA FrameView documentation was checked on 2 October 2026 for capture and analysis capabilities. The repeated-run procedure is editorial guidance, not a published hardware test. The screenshot is an official CapFrameX interface example.
I chose the topic, scope and source list, and I am responsible for publishing corrections. Automation and AI can help with research sorting, first-draft structure and repetitive formatting. Facts, prices and claims must still trace back to the sources shown on the page. This page includes a real tool view or captured evidence where it is labelled.
Real tool view
What a useful frame-time capture contains
A trustworthy capture keeps the test route, hardware description, frame-time graph and percentile results together. This official CapFrameX example demonstrates the interface only; its Doom Eternal numbers are not used as a product benchmark on this site.
Official CapFrameX project screenshot
Shows whether GPU score, clocks, temperature and power look normal.
Good for CPU cooler mounting, fan curves and thermal limits.
Useful after a fresh build, undervolt, RAM tune or random crash.
Frame times and 1% lows tell more than a simple average FPS.
The benchmark stack
Use a synthetic benchmark to establish a repeatable baseline, then record a game you actually play. A score helps compare similar systems; a frame-time graph helps locate interruptions during play. Neither test alone proves the PC is stable.
| Tool | Use it for | Record |
|---|---|---|
| 3DMark | GPU baseline and online score comparison | Score, graphics score, settings, driver version |
| Cinebench | CPU load and cooling check | Multi-core score, CPU temperature, clock behavior |
| OCCT | Stability, power and error checks | Test type, duration, errors, max temperatures |
| HWiNFO | Sensor logging during every test | GPU hotspot, CPU package, clocks, power limits |
| CapFrameX | Real-game frame-time analysis | Average FPS, 1% lows, frame-time graph |
| NVIDIA FrameView | Frame rate, frame time and supported power metrics | FPS, power, performance per watt |
| CrystalDiskMark | SSD sanity check | Sequential and random results, drive slot notes |
Record a baseline before changing settings
Start with stock CPU and GPU settings. Record whether a memory profile is enabled, the driver version, resolution, preset, upscaling mode and frame-generation setting. Close unnecessary background apps and keep these conditions the same between runs.
Choose a repeatable game route or built-in benchmark. Warm up the game, capture the same route three times and keep all three results. A one-off hitch or unusually fast run should not decide whether an upgrade worked.
- Compare the same test and version on systems with similar hardware.
- Keep graphics score separate from total score when comparing GPUs.
- Change one setting, repeat the captures and compare both average FPS and frame-time spikes.
- Record crashes and test errors separately from performance scores.
How to read the result
A low score is a reason to investigate, not a diagnosis. First confirm that the benchmark version, preset, resolution and hardware match the comparison. Then check background load, clocks and temperature during the test.
If the GPU benchmark looks normal but one game stutters, repeat the same scene and check whether spikes coincide with loading or CPU load. If several unrelated tests perform poorly, investigate cooling, power settings and drivers before replacing parts.
- Compare graphics score separately from total score when checking GPU performance.
Keep a test log you can compare later
Record the date, game or benchmark version, CPU, GPU, RAM configuration, driver, display resolution and test settings. For game captures, include the route, capture duration, average FPS, chosen low-percentile metric and frame-time graph.
Use the same capture tool and metric definition in each comparison. This guide provides a procedure; NordPeek has not measured performance for your hardware. The CapFrameX screenshot below is an official interface example.
FAQ
Which benchmark should I run first on a new gaming PC?
Run 3DMark with HWiNFO logging in the background. It gives you a repeatable GPU baseline and sensor data in one pass.
Is 3DMark enough to prove my PC is stable?
No. 3DMark is useful for a performance baseline. Stability needs longer gaming sessions, sensor logging and tools such as OCCT when crashes or undervolts are involved.
Should I trust average FPS?
Average FPS is only part of the story. 1% lows and frame-time graphs are better for judging whether a game actually feels smooth.
Sources and methodology
I use official documentation, public hardware data and repeatable benchmark methods. Treat the settings as a starting point and measure your own PC before changing several variables at once.