How to Stress-Test a New PC Build
Research-led guide — checked 16 October 2026: A new PC that boots successfully is not automatically stable. A sensible stress-test routine checks cooling, CPU stability, memory errors, graphics stability and whole-system behaviour before you trust the machine with important work or long gaming sessions. This guide is based on current tool documentation and practical PC-building principles; it is not a claim that The Tools and Tech has physically tested your individual hardware.
Quick answer
Start with the PC at normal settings, watch temperatures and voltages, then test one subsystem at a time before running a combined workload. A useful sequence is: idle and sensor check, CPU test, memory test, GPU test, then a combined CPU/GPU load. Any crash, calculation error, display corruption, unexpected restart or thermal problem should be investigated rather than dismissed.
Before you start
- Check the build physically. Confirm the CPU cooler is mounted correctly, fans are connected, GPU and motherboard power cables are fully seated and memory is locked into the recommended slots.
- Use known-good settings first. Establish stability before experimenting with manual overclocks, undervolts or aggressive memory tuning.
- Install current chipset and graphics drivers. Driver faults can look like hardware instability.
- Watch temperatures. Use a reputable monitoring tool or motherboard utility while tests run.
- Know when to stop. End the workload if temperatures rise abnormally, cooling stops, the system becomes unresponsive or the test reports errors.
A practical stress-test sequence
1. Check idle behaviour
Leave the PC at the desktop for several minutes. Confirm fans behave normally, storage is detected, memory capacity is correct and there are no repeated crashes or obvious temperature problems. This can catch a loose cooler, missing fan connection or configuration problem before a heavy workload makes it worse.
2. Test CPU stability and cooling
A CPU-focused workload checks processor stability and whether the cooler can cope with sustained heat. OCCT Personal is one current option for home systems; its official download page lists v17.1.8 dated 13 October 2026. OCCT states that its Personal edition is free for personal use.
Begin with a moderate CPU workload rather than immediately choosing the harshest test. Watch temperatures, clock behaviour and error reporting. If the CPU reaches its thermal limit unusually quickly, clocks collapse, the system shuts down or the software reports errors, stop and check cooler contact, fan or pump operation, power settings and any CPU tuning.
3. Test system memory separately
Memory instability can cause random crashes, corrupted files and errors that look like other component faults. A bootable memory test is useful because it runs independently of the normal Windows installation. PassMark MemTest86 currently lists v11.7 Build 1000 and provides a free edition for UEFI systems.
Complete a full test run and treat reported errors as a fault to investigate. If errors only appear with XMP or EXPO enabled, retest at platform defaults. A memory kit can be healthy while an aggressive profile is unstable with a particular CPU, motherboard, BIOS or DIMM configuration.
4. Test the graphics card
Run a sustained 3D workload or reputable GPU stability test and watch for driver resets, black screens, visual artefacts, crashes or abnormal temperatures. A short benchmark score alone is not a stability test: a system can complete a quick benchmark and still fail after heat builds up.
If a GPU test fails, return any GPU overclock or undervolt to default, then check power connectors, PSU capacity, case airflow and driver installation before assuming the graphics card is defective.
5. Run a combined workload
Once CPU, memory and GPU tests pass independently, a combined CPU/GPU workload can expose problems that only appear when power draw and case temperature rise together. This is useful for checking PSU behaviour, case airflow and whether one hot component is feeding heat into another.
You do not need to torture a normal PC indefinitely. The aim is to reproduce demanding real-world conditions long enough to see whether temperatures settle and the system remains error-free. Longer testing makes more sense for workstations, overclocked systems or machines intended for sustained rendering or compute workloads.
What counts as a failed stress test?
- an error reported by stress-test or memory-test software;
- a blue screen, unexpected restart or complete freeze;
- graphics corruption or repeated driver resets;
- temperatures that immediately reach a component limit with heavy throttling;
- fans or pumps that stop or fail to respond under load;
- any new physical symptom such as a burning smell.
A single repeatable failure is more useful diagnostically than hours of random testing. Return the system to stock settings, change one variable at a time and rerun the test that exposed the problem.
How long should you test?
There is no universal number of minutes that proves stability. For an ordinary gaming or home PC, roughly 20–30 minutes per major CPU/GPU test, plus a complete memory-test run and real applications or games, is a practical starting point. For sustained rendering, simulation or development workloads, test for longer and include the actual software the PC will run.
Common mistakes
- Testing an overclock before stock settings.
- Running every test at once before isolating components.
- Ignoring a single memory error.
- Comparing temperatures without accounting for CPU model, power limits, cooler, room temperature and case airflow.
- Assuming a benchmark pass means the whole PC is stable.
- Changing several BIOS settings at once while troubleshooting.
If the PC fails
- Return CPU, GPU and memory tuning to default.
- Repeat the same test to see whether the failure is reproducible.
- Check cooling, power cables, memory seating and GPU seating.
- Test memory at default settings if XMP or EXPO was enabled.
- Update relevant stable drivers and review motherboard BIOS notes before deciding whether a firmware update is justified.
- Test the suspected component separately where possible.
If the machine remains unstable at default settings after basic checks, use the retailer or manufacturer support route while the hardware is still within its return or warranty period rather than masking the problem with reduced performance settings.
Official resources
Final checklist
A new PC is ready for normal use when it can idle normally, complete CPU and graphics workloads without errors or abnormal thermals, pass a proper memory test and then run the games or applications you actually built it for. If you later enable XMP/EXPO, undervolt, overclock or change power limits, retest the affected parts rather than assuming the original result still applies.
Planning a new system? Use our Build a PC tool and PC compatibility checklist before ordering, then return to this guide once the machine is assembled.
