Here is the cruellest disguise in PC repair: a hard drive in its last months does not usually die loudly. It slows down - reads that took milliseconds take seconds of silent retries, and every program, update and boot inherits the wait. Owners blame Windows, RAM, viruses - everything except the one part quietly failing. That is why the engineer's method includes a drive health check before deep software testing: no other measurement on a dying drive can be trusted.
The check itself is a reading, not a repair: the drive keeps its own health diary, called SMART, and your job is simply to read it. Two routes give the same diary - the free tool CrystalDiskInfo draws it as a clear health verdict, or PowerShell asks Windows directly. Run Get-PhysicalDisk | Select-Object FriendlyName, MediaType, HealthStatus and Windows will answer Healthy, Warning or Unhealthy in one line. But the headline verdict is only the summary - the numbers beneath it tell the real story.
In the SMART detail, three readings carry the conviction on a mechanical drive:
On an SSD the diary speaks differently: a health percentage counting down its write life, and wear figures instead of sector drama. A low percentage with no errors is an old age warning, not a corruption verdict - read which language your drive speaks.
Now connect the paperwork to the pain. A Caution drive plus the classic pattern from the disk case page - 100% active time, tiny MB/s, enormous response time - is a conviction, not a coincidence. Add the supporting witnesses: freezes where the disk light stays solid, Event Viewer System log errors from source disk (Event ID 7 for bad blocks, 51 for paging errors, 153 for retried IO), and files that open on the second attempt.
One witness can lie; three agreeing witnesses inside one week is the drive telling you its retirement date.
This is the rule this whole method is strictest about: on a suspect drive, data comes off before any further testing. No full-surface scans, no chkdsk marathons, no defragmenting a drive that is already retrying reads - every hour of heavy work can be the hour it stops answering entirely.
Copy the irreplaceable data first - documents, photos, accounts data - to another device, and confirm the copies open. Only then continue diagnosing, in the calm knowledge that even a total failure tomorrow is now an inconvenience, not a loss.
Watch the video read a real drive's diary and match it against the slowness the owner described.
Watch Full VideoA Good reading clears the drive of the known, recorded failures - it does not guarantee the future or catch every fault. If the symptom pattern still matches a struggling drive (100% active time, huge response times), keep the evidence from all readings together and watch whether the numbers move between checks.
Not on a suspect drive, and not as a first test. A full chkdsk works the drive hard for a long time - exactly the stress a failing drive may not survive. Read SMART first, back up the data, and only then consider heavier scans with the data already safe.
That percentage is a fuel gauge for the SSD's rated write life, not a fault counter. 90% after years of use is normal ageing. Worry when it falls fast, reaches very low figures, or arrives with errors and disappearing data - and keep backups regardless, because gauges do not schedule failures politely.