Before you begin
A stop error, a frozen application and a complete loss of power are different events. Record what actually happens: does Windows show a code, does sound continue, can the pointer move, and does the PC restart by itself? A photograph of the whole error screen is more useful than a description such as “it crashed”.
Save your work and back up important files while the system is usable. Note the last change before the problem: drivers, Windows updates, a new game, additional RAM or a tuning setting. If you smell burning, hear electrical crackling or see liquid, switch off and unplug the PC when safe, then contact support.
Work through the checks
-
Make a short incident record
Write down the time, application, workload and any exact stop code. Note how long the PC had been running and whether the fault happens at idle, under load or on waking. Reliability Monitor, found through Windows Search, can help match an incident to an application failure or update. A “Windows was not properly shut down” entry records the result; it does not identify the cause.
-
Return optional tuning to a known state
If you recently changed CPU, GPU or memory tuning, restore the previous stable configuration. Do not raise voltage or power limits as a general crash fix. If you do not know which settings changed, photograph them and ask support. Do not clear all firmware settings blindly because boot configuration and encryption recovery requirements may also be affected.
-
Separate an app fault from a system fault
If only one game closes, verify its files, check its official known issues and temporarily remove unsupported mods. Test a different ordinary workload. If Windows itself restarts across unrelated tasks, record that. Use current component drivers from the manufacturer; if the crashes began with a particular driver, a supported rollback may be more informative than installing multiple driver packages.
-
Check Windows and temperature symptoms
Review updates and use our system-file repair guide if Windows features also fail. Observe CPU and GPU temperatures alongside the workload, using the correct sensor and component limits. A hot number without the model or sensor name is not enough to diagnose overheating. Repeated thermal shutdowns need attention rather than being treated as normal long-session behaviour.
-
Investigate memory and storage carefully
Use the memory guide for compatibility and diagnostic checks. An error can involve the module, slot, memory controller or settings; it does not automatically identify a single failed stick. If storage health warnings or missing drives accompany crashes, back up and stop unnecessary writes. Do not ignore a test’s memory errors as “just running out of RAM” without reading its exact message.
-
Keep testing controlled
Reproduce the original fault with a normal task after each change. Avoid launching several stress tools simultaneously. A synthetic test can add substantial heat and power load, so use one only for a specific support-led question, monitor it and stop on errors or abnormal behaviour. Passing a short stress test is useful evidence, not a certificate that every component is healthy.
What the results tell you
IRQL_NOT_LESS_OR_EQUAL or KMODE_EXCEPTION_NOT_HANDLED
These codes can have multiple driver and hardware causes. Record any named file and the circumstances; do not delete the file or replace RAM based on the code alone.
WHEA_UNCORRECTABLE_ERROR
Hardware, firmware and unstable settings can be involved. Preserve the exact report and configuration for support.
Instant power-off without an error
Investigate power and cooling as well as software. A Windows log may record the unexpected shutdown without explaining why it happened.
Common questions
Should I disable Fast Startup for every crash?
It can be a targeted comparison for shutdown or startup symptoms, but it is not a universal cure for a particular stop code. Restarting is a useful initial comparison.
Is a test at 99% safe but 100% dangerous?
There is no universal safety boundary at those numbers. Workload, cooling, component limits and test behaviour matter; use the test developer’s instructions.
Can I share a crash dump?
Dumps can contain information from memory. Share them only through an appropriate support channel when requested, not in a public post. Start with the code and incident details.
When to contact ElitePCS
Contact us for repeated system crashes, failed memory diagnostics, shutdowns under ordinary use or any suspected electrical fault. Provide a concise incident record and the exact changes already tested.
Contact ElitePCS support. Include your order number, the exact symptom and the checks you have completed. Keep passwords, licence keys and recovery keys out of screenshots.
Technical references: Microsoft WHEA guidance · Intel shutdown diagnostics · Microsoft system-file repair. Reviewed 27 September 2026.