Overclock and XMP Instability Test Your PC at Stock Settings First
Every overclock is a loan taken against stability, and the interest comes due as random blue screens. The diagnostic rule professionals use is absolute: an unstable PC gets tested at stock settings first - everything else is guessing until that is done.
Watch the full method in the video - identifying a blue screen cause, step-by-step
XMP/EXPO memory profiles, CPU overclocks, GPU overclocks, and undervolts all push hardware past the settings its maker guaranteed. They can pass benchmarks for months and still corrupt memory once a week. That is why any crash investigation on a tuned PC starts the same way: load factory defaults, test, and only then re-tune. This is not being anti-overclock - it is separating two questions that are otherwise impossible to untangle: is a part faulty, or is the tune too aggressive? One honest note: passing a 10-minute stress test does not prove stability - marginal systems fail after hours, or only in one particular game.
BIOS settings that count as not stock (examples)
[x] XMP / EXPO memory profile enabled
[x] CPU multiplier or PBO limits changed
[x] Curve Optimizer undervolt applied
[x] GPU core or memory offset in a tuning app
If any box is ticked, your PC is overclocked - even if you never called it that. XMP counts.
The usual hidden overclocks. Loading BIOS defaults clears the first three; GPU tuning apps must be reset separately.
1
What Stock Actually Means - the Full Reset List
Partial resets give partial answers. Stock means all of it.
In BIOS/UEFI choose Load Optimized Defaults (or Load Setup Defaults), then save and exit
Confirm XMP/EXPO is Off afterwards - defaults sometimes keep it on
In Windows, open your GPU tuning app (Afterburner, Adrenalin, G-Helper) and reset core/memory offsets to zero, or uninstall the app for the test period
Remove CPU undervolts done in BIOS or tools like ThrottleStop / Ryzen Master - defaults in the tool itself
2
Run the Stock Test Long Enough to Mean Something
The test only works if it covers the conditions that used to crash.
Use the PC normally for several days at stock - same games, same workloads, same long sessions
Short stress tests help but do not replace real use; add an overnight MemTest86 run (4+ passes) for the memory side
Keep your crash diary going - date every crash, or celebrate every clean day
If the PC crashes at full stock too, stop tuning entirely: you have a faulty part or driver, and the other guides in this series apply
3
If Stock Is Stable - Re-Tune One Step at a Time
Now the tune itself becomes the experiment, reintroduced in controlled steps.
Enable only XMP first and test for days - memory profiles are the most common silent failure
If XMP alone crashes, drop the memory speed one step (for example 3600 to 3200) or accept base speed - the real-world performance difference is small, the stability difference is everything
Then add CPU tuning, then GPU tuning, days apart - the step that brings crashes back is your limit
Write down your last fully stable settings somewhere outside the PC
4
Why Tuned PCs Fool Their Owners
Understanding the trap helps you not fall into it twice.
Degradation is real: a tune that was stable for a year can fail as components age and as seasons change case temperatures
New games are new tests: a title that uses AVX instructions or fills VRAM differently can find instability every older game missed
Errors hide before they crash: silent corruption can damage files long before the first blue screen - one more reason the stock test comes first, not last
Be honest in support forums and RMA claims: state your tune - helpers and makers ask for a reason
No driver reinstall, Windows reinstall, or replacement part should happen before the stock test on a tuned PC. Shops will ask whether it crashes at stock - have the answer ready instead of paying for their time to find it.
Watch the Complete Video Guide
The full blue screen identification method - dump files, stop codes and the fix - shown live, step-by-step.
My overclock passed Prime95 and benchmarks. How can it be unstable?
Benchmarks test specific loads for short periods. Real instability is often rare - one bad memory access in hours of mixed use, in one game, at one temperature. Passing stress tests lowers suspicion; it never clears it. Days of crash-free real use at stock versus crashes when tuned is the deciding evidence.
Is XMP really an overclock? The RAM box advertises that speed.
Yes. The speed on the box is the kit's tested profile, but it runs the memory controller beyond the CPU maker's base specification. Most systems hold it fine; some combinations do not. That is exactly why disabling XMP is step one of memory-crash diagnosis.
Crashes happen at stock settings too. What is next?
Then tuning is innocent and something is genuinely faulty or corrupted. Go back to the dump pattern: same driver named means software, scattered codes mean hardware - test RAM with MemTest86, check drive health and temperatures, and test with one stick at a time.
This guide is based on the Technical Anshu blue screen video - stuck on any step, ask in the video comments and you will get an answer.