How to Test a New PC for Stability: A Staged Checklist

1. Save a reproducible baseline

Before testing, record the CPU, GPU, memory kit and occupied slots, BIOS version, operating-system build and main drivers. Start at manufacturer-standard settings with optional overclocks, undervolts and memory profiles off. Check that all the hardware is detected and that cooling works. Save open work and back up important files. Use one monitoring setup consistently. If you are investigating a crash rather than checking a new build, also write down the task and settings that produce it; that is the test you will need to repeat.

2. Check ordinary operation first

Try browsing, video playback, file access and the application you bought the PC to run. Check cold starts and restarts, plus sleep and resume if you use them. Test the network, audio and front ports. If ordinary use already produces a restart, missing device or freeze, work on that fault first. Running a heavier benchmark will not repair it. Keep the time and visible symptom of the failure so you can tell whether the next attempt behaves the same way.

3. Run a memory test from a prepared USB drive

For a modern UEFI PC, MemTest86 is one concrete starting point. Use the "MemTest86 download and requirements" link in Sources below. On Windows, extract the package and use its included imageUSB tool to prepare a spare USB drive. Writing the image erases that USB drive: confirm the selected device before continuing. Use the motherboard boot menu to start its UEFI USB entry, following the linked PassMark instructions.

Let the configured test sequence finish and save the result report to the USB when prompted. Record the tool version, test selection and completed passes; an interrupted run is not the same as a completed pass. Any reported error needs investigation, but it does not identify a bad DIMM on its own. Restore standard memory settings before testing modules individually in the board's prescribed slot. Keep the original report so you can compare the repeat.

4. Exercise the processor and graphics separately

For a supported Intel processor, the Intel Processor Diagnostic Tool is one option; check its supported-products list before running it. For a separate stress workload, OCCT offers CPU and CPU + RAM tests. Use its official download and test-description links below, choose an edition licensed for your use, and check its current system requirements. The CPU + RAM test exercises both areas; it cannot by itself identify which part is faulty.

Save the selected test settings before starting, monitor temperatures and stop for the faults in the next section. Avoid changing clock or voltage settings to make a test pass. Then check graphics separately with a familiar game benchmark or rendering task at fixed settings. Record artifacts, driver resets or crashes as well as completion. Repeat a failing task before changing another component.

5. Monitor and know when to stop

Watch temperatures, fan or pump operation, reported errors and performance while a test runs. Interpret the sensors using the component maker's limits. Stop for test errors, a cooling failure, repeated crashes, severe abnormal noise or unexpected shutdowns. Shut down and disconnect power safely if there is smoke, a burning smell or a leak. Keep the error message and configuration before trying a fix. Raising voltages or disabling protection just to reach the end of a test hides the problem you are trying to find.

6. Test combined real workloads after individual checks

After the separate checks, run the combination you actually use: for example, a game with voice chat and recording, or an editing session followed by export. Include sessions long enough to resemble your work and let the machine warm up. Record the task and duration instead of treating an arbitrary number of hours as proof of stability. For storage, use a drive maker's non-destructive diagnostics where available. Destructive write tests have no place on a drive containing data you want to keep.

7. Change one setting and retest

Treat an XMP or EXPO profile as a new configuration. Repeat the relevant checks after enabling one; reaching the desktop is only the start. If a task fails only with the profile enabled, restore the previous settings and rerun that task. This separates a settings-dependent failure from one that also occurs at standard settings.

Copy the blank record below into your build notes and fill it in for each run. It contains no RigAtlas test results. Keep completed, interrupted and failed runs distinct, and preserve the report or screenshot alongside the entry. A pass describes those conditions; it cannot cover every future driver, application or room temperature.

Your checklist

• Record the hardware, software versions and standard settings.
• Resolve failures in everyday use before adding heavy test loads.
• Save memory and CPU/GPU test results with their settings.
• Stop for errors or cooling faults and keep the failure evidence.
• Retest after tuning, and state which tasks and durations actually passed.