Most drive failures aren't instant — they're preceded by weeks or months of warning signs that are visible if you know where to look. Both hard drives (HDDs) and solid-state drives (SSDs) constantly track their own health data internally. Here's how to read it, and what actually warrants backing up immediately versus what's normal wear.
SMART data: the built-in health report every drive keeps
Nearly every modern HDD and SSD tracks a set of internal metrics called SMART (Self-Monitoring, Analysis, and Reporting Technology) — things like reallocated sectors, read/write error rates, power-on hours, and temperature history. This data exists specifically to give an early warning before a drive fails outright, and it's accessible through free tools without needing to open the machine.
How to actually check it
- Windows: free tools like CrystalDiskInfo read SMART data and translate it into a plain health status (Good, Caution, Bad) rather than raw numbers, which makes it usable without technical background.
- Mac: Disk Utility shows a basic SMART status ("Verified" vs a warning) for internal drives; for more detail, third-party tools like DriveDx read the full SMART data set.
- Linux: the
smartctlcommand (part of the smartmontools package) reads full SMART data directly from the terminal.
Run a check every few months on any drive you depend on, and definitely before you'd otherwise assume everything's fine on an older drive.
The specific numbers worth paying attention to
- Reallocated sector count (HDDs). This counts sectors the drive has quietly retired because they were failing. A small, stable number isn't alarming; a number that's climbing over time is a real warning sign.
- Pending sector count. Sectors currently flagged as unstable, waiting to be either recovered or reallocated. Any non-zero and growing value here deserves attention.
- Media wearout indicator / percentage used (SSDs). SSDs have a finite number of write cycles per cell; this figure tracks how much of that lifespan has been used. Under normal desktop or laptop use, this climbs very slowly and rarely becomes a real concern within a drive's practical lifetime — but it's worth checking on a drive handling constant heavy writes.
- Temperature history. Drives run reliably within a specified range; a drive that regularly runs hot (from poor case airflow, for example) tends to wear out faster than one that stays cool.
- Uncorrectable error count. Errors the drive couldn't fix internally. Any non-zero value here, especially if it's increasing, is one of the more serious warning signs.
Non-SMART warning signs worth acting on too
- Unusual clicking or grinding noises from a mechanical hard drive — a well-known and fairly reliable sign of imminent failure.
- Files that take noticeably longer to open, or that occasionally fail to open at all.
- The operating system reporting disk errors during startup or file operations, especially if this is new behavior.
- A drive that intermittently disconnects or isn't recognized, particularly for external and portable drives.
What to do when you see a warning sign
Back up anything important on that drive immediately, before doing anything else — troubleshooting, running repair tools, or even just continuing normal use risks losing data on a drive that's actively failing. Once backed up, then decide whether the drive is worth continuing to use for less critical data or should be retired. A tool showing a "Caution" or "Bad" SMART status, or a climbing reallocated sector count, is a reason to treat the drive as unreliable going forward even if it still technically works today.
Quick drive-health checklist
- Check SMART status every few months using CrystalDiskInfo, Disk Utility, or smartctl.
- Watch reallocated/pending sector counts and uncorrectable error counts specifically.
- Treat unusual noise, slow file access, or intermittent disconnects as warning signs, SMART data or not.
- Back up immediately at the first sign of trouble, before troubleshooting further.
- Keep routine backups running regardless — SMART monitoring reduces risk but doesn't catch every failure.