What Is SMART Data and How to Read Drive Health Honestly

Where the numbers come from and how to see them

SMART lives in the drive's own firmware: every completed pass, every retry, every sector that had to be remapped bumps a counter inside the disk itself. Windows surfaces almost none of it — the stock HealthStatus in Settings or Get-Disk tells you only OK/Warning/Unhealthy, which is a fire alarm, not a stethoscope. For the full table you need a reader such as CrystalDiskInfo, or PowerShell: Get-PhysicalDisk | Get-StorageReliabilityCounter, which pulls wear, temperature and error counts without installing anything.

The catch is that attribute IDs are not standardized across vendors. ID 5, 0xC5 and 0xC6 mean one thing on a WD drive, something subtly different on Seagate, and on many SSDs the old spinning-disk attributes are simply padded zeros. NVMe drives sidestep the mess with a clean standard: Percentage Used, Available Spare and Media Errors — three numbers that actually mean what they say.

• 05 — Reallocated Sectors Count: sectors the drive remapped to spare blocks
• C5 (197) — Current Pending Sectors: unstable sectors awaiting a verdict
• C6 (198) — Uncorrectable Sectors: reads that failed outright
• C7 (199) — UltraDMA CRC Errors: a cable or controller problem, not a dying drive
• NVMe — Percentage Used, Available Spare, Media and Error Information Log

Raw values lie more often than they tell the truth

Vendors pack raw values in vendor-specific ways: some store a rate, a total and a lifetime count in one 64-bit number, so a perfectly healthy Seagate shows a terrifying-looking Read Error Rate of eighty million. The normalized value and the threshold columns exist for exactly this reason — but even those are conservative tripwires that usually fire long after a careful owner would have acted. A single reading, on any drive, tells you almost nothing.

Trend is the only honest signal. A reallocated count that has sat at four sectors for three years is a curiosity; the same counter climbing every week is a countdown, and you should treat it as one. This is why tools that log history beat one-off snapshots: the number that matters is the slope, not the value. One pending sector after an unexpected power cut is often self-healing debris; a hundred that grow toward uncorrectable are not.

What to do when the numbers turn bad

The order of operations matters. Back up first, before running more tests — every full scan of a genuinely failing drive is a small gamble with its remaining life. Then rule out the cheap causes: C7 errors usually mean a loose or cheap SATA cable, not a condemned disk. Check the event log alongside SMART, because Windows logs disk resets and bad blocks (events 7, 51 and 153) that corroborate what the counters claim.

Growing reallocations, pending sectors that never clear, or an NVMe drive whose Available Spare has dropped below its threshold all mean the same thing: shop for a replacement and stop stress-testing. A stable, small reallocated count on an old HDD is not a death sentence — run it until the trend moves, with backups current. Monitoring once a week is plenty; SMART feeds on nothing and never needs refreshing on your side.

Questions and Answers

Read next