The cable or connection
The cheapest explanation and worth eliminating first. A marginal USB or SATA cable corrupts data in transit, so the checksum fails on perfectly healthy storage. Swap it, and go direct rather than through a hub.
Check first
The message names the test that failed rather than the reason — which is why the same error can mean a marginal cable or a dying platter. It is a smoke alarm, not a diagnosis, and the useful next step is working out which.
Every sector carries a checksum written alongside it. When what comes back off the platter fails that check, the drive raises this error rather than passing you data it already knows is wrong.
chkdsk /f /r as a first response. On a drive with failing sectors it forces repeated re-reads of exactly the areas that are struggling, and it can relocate or discard data while “repairing” the file system. If the file matters, copy what you can first.The message names the test that failed, not the reason it failed — which is why it appears in such different situations.
Every time data is written to a drive, a short mathematical value is calculated from it and stored alongside. When that data is read back, the value is recalculated and compared. If the two disagree, the drive reports a cyclic redundancy check error: what came back is not what went in.
That is a genuinely useful mechanism — it means the drive tells you about corruption rather than silently handing you damaged data. But it is a smoke alarm, not a diagnosis. It reports that a read went wrong and says nothing about whether the cause was a failing platter, a bad cable, a dying controller or a single sector that has degraded over years.
Which is why the same message can mean a five-minute fix or a serious hardware failure, and why the next step is establishing which.
In rough order of how easy they are to rule out.
The cheapest explanation and worth eliminating first. A marginal USB or SATA cable corrupts data in transit, so the checksum fails on perfectly healthy storage. Swap it, and go direct rather than through a hub.
One area of the platter that no longer reads reliably. The rest of the drive may be entirely sound. Common on older drives and on regions untouched for years.
The serious version. Reallocated and pending counts climbing, errors appearing in more places, transfers stalling. The drive is deteriorating and the errors are a symptom rather than the problem.
On SSDs particularly, a failing controller produces read errors that look identical to media damage. The flash may be fine behind it.
Work through these before running any repair.
Two minutes, no risk, and it eliminates the one cause that has nothing to do with your data. Go straight into the machine rather than through a hub or dock.
CrystalDiskInfo or DriveDx, installed to a different drive. Reallocated and pending sector counts tell you whether this is one bad area or a drive in decline. Any pending count above zero is worth acting on today.
A CRC error on a single file that always fails, with everything else reading normally, points at a localised defect. Errors across unrelated folders point at the drive.
Separates a drive fault from a controller, driver or motherboard fault. If the errors follow the drive, the drive is the problem.
It is the standard advice for CRC errors and it is the standard advice that finishes drives off.
chkdsk with the /r switch performs a full surface scan — reading every sector on the disk, for hours, and attempting to recover the contents of any it finds bad. On a healthy drive with one degraded sector that is reasonable. On a drive with spreading bad sectors it is the heaviest sustained load you can apply to hardware that is already struggling, and drives regularly fail partway through.
Worse, chkdsk repairs by making the file system consistent. Where it cannot reconcile a record it discards it — which on a damaged volume means deleting exactly the structures a recovery would rebuild from. You can end up with a clean, consistent file system containing considerably less than it did before.
The order that works: establish whether the drive is healthy, copy the important data off if it is not, and only then repair anything. If the drive still reads, you are the most effective recovery tool available. Where it does not, a single drive in Manchester is from £300 +VAT after a free 48-hour diagnostic, with imaging done in graduated passes so the readable majority is secured before the difficult regions are attempted.
That a checksum failed — the value calculated when data was written does not match the value calculated when it was read back. It tells you a read went wrong, not why. The cause can be anything from a marginal cable to a failing platter.
No, and that is what makes it awkward. It can be a faulty cable corrupting data in transit on a perfectly healthy drive, a single sector degraded over years, or a disk in active decline. The checks are quick and worth doing before assuming the worst.
The cable and the port — two minutes, no risk, and it eliminates the one cause unrelated to your data. Then SMART, to see whether reallocated and pending sector counts are climbing. Then whether it is one file failing or many.
Not before you know the drive is healthy. chkdsk /r reads every sector for hours on a disk that may be struggling, and it repairs by discarding records it cannot reconcile — producing a consistent file system rather than your files.
Frequently. Imaging in graduated passes captures everything that reads immediately, then returns to the difficult regions with longer timeouts. Where a sector genuinely cannot be read, that portion of that file is lost while the rest recovers.
Because the damaged sectors happen to fall within that file. A localised defect on an otherwise sound drive produces exactly that pattern — one file that always fails while everything around it reads perfectly.