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

1. Save a reproducible baseline

Record the CPU, GPU, memory kit and slot arrangement, motherboard BIOS, operating-system build and key driver versions. Begin at manufacturer-standard settings with optional overclocks, undervolts and memory profiles disabled. Verify the hardware is recognised and cooling is working before applying sustained load. Back up important files and save work. Use one monitoring setup so competing utilities do not confuse the comparison. State the purpose of the test: a new assembly check, a memory-profile change or investigation of a repeatable crash.

2. Check ordinary operation first

Start with the tasks the PC must actually perform: browsing, video playback, file access and a representative application. Test cold starts, restarts and sleep/resume where used, as well as network, audio and front-panel ports. Note any repeatable symptom and its timing. If the machine already fails these simple checks, resolve that problem before escalating to heavy synthetic load. A missing device or random desktop restart is useful evidence even when a benchmark has not yet been run.

3. Test memory with a suitable dedicated tool

Use a memory test from its official publisher and follow its setup and full-run instructions. Creating a bootable test USB may erase that USB drive, so identify it carefully. Let the chosen standard suite complete and save its report, including test coverage and configuration. Any reported memory error needs investigation; it does not by itself identify a particular defective DIMM because settings, the processor's memory controller and the motherboard are also involved. Return to standard settings before comparing modules individually in the prescribed slot.

4. Exercise the processor and graphics separately

Choose diagnostics that explicitly support the hardware and operating system. Intel's Processor Diagnostic Tool, for example, performs processor checks on supported Intel systems; verify its current supported-products list rather than assuming every new Intel model is included. Save the functional-test result. For graphics, repeat a known game benchmark or representative rendering task with fixed settings and look for artifacts, driver resets or crashes. Keep CPU-focused and GPU-focused checks separate initially so the first failure is easier to reproduce and investigate.

5. Monitor and know when to stop

Observe relevant temperatures, fan or pump behaviour, errors and performance changes while testing. Use the component maker's limits and interpretation, not a universal temperature target. Stop the workload for a reported test error, repeated crash, obvious cooling failure, severe abnormal noise or unexpected shutdown. Disconnect power safely if there is smoke, a burning smell or liquid leakage. Do not disable protections or raise voltages to force a test to finish. Keep the error message and the last settings instead of immediately overwriting them with another test run.

6. Test combined real workloads after individual checks

Once individual checks complete, try the combination you intend to use, such as gaming while voice chat and recording are active. Include longer work sessions that resemble normal use and watch for faults that only appear after the system warms up. There is no single duration that proves every PC stable. Record what ran, for how long and whether it completed without observed errors. Storage checks should use the drive maker's non-destructive diagnostics where available; do not use destructive write tests on a drive holding wanted data.

7. Change one setting and retest

If you enable a supported XMP or EXPO profile or another optional adjustment, record it as a new configuration and repeat the relevant checks. Memory profiles are overclocking settings, and successful startup alone does not validate them. If errors appear only after that change, revert to the last working baseline and retest the same workload before blaming a component. Save a short final record of the tested configuration, completed checks and unresolved observations. A pass covers those conditions; future drivers, applications or warmer surroundings can expose a different problem.

Your checklist

• The exact hardware and standard-settings baseline were recorded.
• Everyday functions were checked before heavy testing.
• Memory and processor or graphics checks produced saved results.
• Errors and cooling faults trigger investigation rather than more aggressive settings.
• The final notes say what was tested and preserve any unresolved issues.