Before repairing anything, spend a minute finding out whether the hardware is sound. It costs nothing, it cannot make matters worse, and it is the step most people skip before escalating to repairs that assume a software fault.
SMART only reports overall failure once a manufacturer’s threshold is crossed, which is usually far too late to be useful. The individual counters underneath start moving considerably earlier.
Before repairing anything, find out whether the hardware is sound. Most people skip this and escalate on a guess.
CrystalDiskInfo on Windows, DriveDx on macOS, or smartctl from a terminal. Install it somewhere other than the drive you are worried about, because installers write.
Ignore the overall “Good / Caution” verdict; it is a threshold the manufacturer chose. Look at reallocated sector count and current pending sector count.
Sectors the drive found bad and retired, remapping them to spares. A handful over years is normal. Dozens, or a number that is climbing, means active deterioration.
Sectors it cannot read now but has not yet retired, because it cannot recover the contents to move them. Any pending count above zero on a drive that matters deserves attention today.
The rate of change tells you more than the value. A count that moves between two readings means the drive is failing now, not at some point in the past.
SMART is genuinely useful and it is not a health certificate. These are the gaps.
SMART reports flash wear well — percentage used, bytes written, spare blocks. It barely reports controller health, and the controller is what fails catastrophically. A drive can show 96% life remaining and be undetectable the next day.
Many external enclosures do not pass SMART data through, so the tool reports nothing at all. That is the bridge, not the drive. Removing the disk to test is unwise on WD units, which encrypt at the bridge.
A head crash or a seized bearing happens in an instant. No counter climbs beforehand. SMART predicts gradual wear, not accidents.
Hardware RAID controllers often hide the individual disks, so per-drive SMART is invisible to the operating system. The controller has its own tools, and its own alerting that nobody has configured.
Frequently more informative than the counters, and free to observe.
Freezes that clear on their own. The machine locks for twenty or thirty seconds with the disk light solid, then continues normally. That is the drive retrying a sector it cannot read while the operating system waits. It reads as an old computer and it is a dying disk.
Copies that stall at the same point. A transfer that runs fast then halts at a consistent percentage is hitting the same unreadable region every time.
Files that open on the second attempt. First try fails or takes twenty seconds; second try is instant. The drive succeeded eventually and cached the result.
Slowness that arrived suddenly. Gradual decline over years is software accumulating. A machine that became slow over a fortnight is usually hardware.
And one of them does not involve paying anybody.
Counts healthy, machine misbehaving. Software troubleshooting is reasonable. Run the repairs, reinstall if you need to, and treat it as an operating system problem — which it probably is.
Counts climbing. Stop repairing and start copying. Not a disk image, not a clone — open a file manager and copy the things that matter to another physical drive, most valuable first. Then stop using the drive entirely.
What not to do in the second case: chkdsk /r, defragmentation, or a deep recovery scan. All three are hours of sustained reading across the whole surface, which is precisely the load a deteriorating drive cannot take. Drives regularly die partway through a scan intended to save them. If the drive still reads, you are the best recovery tool available — and you cost nothing.
Install CrystalDiskInfo on Windows or DriveDx on macOS — to a different drive than the one you are checking — and look at two figures: reallocated sector count and current pending sector count. Ignore the overall Good/Caution verdict, which is just a manufacturer threshold.
Reallocated sectors are ones the drive found bad and retired to spares — damage already handled. Pending sectors are ones it cannot read now and has not yet retired, because it cannot recover the contents to move them. Pending is the more urgent number.
Not necessarily. That verdict is a threshold, and a drive can pass it with a rising reallocated count. It also tells you little about SSD controllers, nothing about drives behind USB bridges that do not pass SMART through, and nothing at all about sudden mechanical failure.
Because the drive is retrying a sector it cannot read, which can take many seconds, and the operating system simply waits. It reads as an old computer and it is usually a failing disk.
No. chkdsk with /r reads every sector on a disk already struggling, for hours, and repairs by discarding what it cannot reconcile. If the drive still reads, copy your files off first — that is the only step that helps.
Nobody can tell you honestly. Some run for months, others fail within days. What is certain is that the counts never go back down, so the window only narrows from here.