Disk Detector
Windows 10 & 11 Buy license · $29.95
Reference

SMART Attributes Explained: The Values That Matter and What They Mean

SMART attributes explained in plain English: reallocated, pending and uncorrectable sectors, CRC errors, wear and temperature, and which mean trouble.

Updated 5 minute read The Disk Detector team

SMART attributes are the individual health readings a hard drive or SSD records about itself, such as how many sectors it has had to replace, how hot it runs and how many hours it has been powered on. Only a handful really predict failure. On hard drives, watch Reallocated Sectors Count (5), Current Pending Sector Count (197), Offline Uncorrectable (198), Reported Uncorrectable Errors (187) and Spin Retry Count (10). On SSDs, watch wear level, available spare and media errors. Anything above zero in the hard drive sector counts deserves attention, and a number that keeps rising means the drive is failing.

This reference explains how to read a SMART table, lists the attributes you are most likely to see, and shows which values are normal and which mean trouble.

How to read a SMART table

Each attribute in a SATA drive's SMART data has several columns:

  • ID: a number identifying the attribute, often shown in hexadecimal as well (for example 5 is 05, 197 is C5).
  • Raw value: the actual reading, such as a count of sectors or hours. This is usually the most useful number.
  • Current (normalized) value: a score set by the manufacturer, typically starting at 100 or 200 and falling as things get worse.
  • Worst: the lowest normalized value ever recorded.
  • Threshold: if the normalized value falls to this level or below, the drive declares that attribute failed.

The drive's overall SMART status only says "failed" when a value crosses its threshold, and manufacturers set thresholds conservatively. That is why a drive can report "OK" while its raw counts already show clear warning signs. Reading the raw values, and watching how they change, gives you much earlier notice.

Some raw values are vendor-specific and look alarming without meaning anything. Raw Read Error Rate and Seek Error Rate on some brands show huge numbers on perfectly healthy drives, because the raw value packs several counters together. Judge those by their normalized value instead.

Critical hard drive attributes

These are the attributes most closely linked to drive failure. On a healthy drive, their raw values are zero.

IDNameWhat it countsConcern level
5 (05)Reallocated Sectors CountBad sectors swapped for sparesAny non-zero value; urgent if rising
10 (0A)Spin Retry CountTimes the drive needed more than one try to spin upAny non-zero value
187 (BB)Reported Uncorrectable ErrorsErrors the drive could not fix with error correctionAny non-zero value
188 (BC)Command TimeoutOperations that timed outRising values; can also be power or cable issues
196 (C4)Reallocation Event CountNumber of remapping attemptsRising values
197 (C5)Current Pending Sector CountUnstable sectors waiting to be remappedAny non-zero value
198 (C6)Offline UncorrectableSectors that failed the drive's own offline scanAny non-zero value

Pending sectors deserve a special note. When a sector cannot be read, the drive marks it pending. The next time that sector is written, the drive either finds it is fine (and the count goes down) or replaces it with a spare (and Reallocated Sectors goes up). A pending count that will not clear means data in those sectors is currently unreadable. Our guide to the reallocated sectors count covers what to do about it.

Informational attributes

These describe the drive's age, usage and conditions. They do not predict failure on their own, but they add context.

IDNameWhat it tells you
9 (09)Power-On HoursTotal hours powered on. 8,760 hours is one year of continuous use.
12 (0C)Power Cycle CountHow many times the drive has been switched on
4 (04)Start/Stop CountSpindle start and stop cycles
3 (03)Spin-Up TimeHow long the platters take to reach speed
193 (C1)Load/Unload Cycle CountTimes the heads parked; high on drives that park aggressively
194 (C2)TemperatureCurrent temperature in degrees C
199 (C7)UDMA CRC Error CountData errors on the cable between drive and PC

CRC errors are a useful special case. A rising count almost always points to a bad SATA cable, a loose connection or a faulty USB enclosure, not the disk itself. Replace the cable and watch whether the count stops climbing. The count never resets, so look for changes, not the total. For safe temperature ranges, see our guide to normal hard drive temperatures.

SATA SSD attributes

SATA SSDs use the same table format, but many attributes are specific to each manufacturer. The names differ, but they usually cover the same ideas:

  • Wear or life remaining, under names such as Wear Leveling Count, SSD Life Left, Percent Lifetime Remaining or Media Wearout Indicator. Usually a normalized value counting down from 100.
  • Total data written, often as Total LBAs Written. Multiply by the sector size (usually 512 bytes) to get bytes, and compare with the drive's TBW rating.
  • Program and erase failures, which count flash blocks that failed to write or erase.
  • Available reserved space, the spare blocks left.

NVMe health information

NVMe drives do not use numbered attributes. They report a standard health log, with the same fields on every brand:

FieldMeaningWatch for
Critical WarningFlags for low spare, overheating, reliability or read-only modeAnything other than 0
Percentage UsedEstimate of rated endurance usedApproaching 100 (it can exceed 100)
Available SpareRemaining spare capacity as a percentageFalling toward the Available Spare Threshold
Media and Data Integrity ErrorsUnrecovered data errorsAny non-zero value
Composite TemperatureDrive temperatureSustained high readings
Unsafe ShutdownsPower lost without a clean shutdownRapid increases
Data Units WrittenTotal writes, in units of 512,000 bytesCompare with the TBW rating

Our guide on how to check SSD health shows how to read these on Windows.

How to use SMART values in practice

  1. Record a baseline when a drive is new or when you first check it: the critical counts, power-on hours and, for SSDs, wear level.
  2. Check again regularly, monthly for most PCs and more often for older drives.
  3. Compare, do not just read. A stable reallocated count of 2 is far less worrying than a count that went from 0 to 40.
  4. Rule out cables when only CRC or command timeout counts are rising.
  5. Back up and replace a drive whose critical counts keep climbing, or an SSD whose spare or wear values approach their limits.

Reading SMART data without decoding it yourself

Windows does not show a full SMART table. PowerShell shows a few reliability counters, and Settings shows NVMe life estimates, but none of it explains what the values mean. Disk Detector reads the full SMART data for hard drives, SSDs and NVMe drives and turns dozens of numbers into a simple Good, Caution or Bad rating, with the specific reasons, such as reallocated or pending sectors, and the advice that goes with them. When a USB enclosure does not pass SMART data through, Disk Detector tells you instead of showing an empty table.

Frequently asked questions

Which SMART attributes predict failure best?

Reallocated sectors, current pending sectors and offline uncorrectable sectors are the strongest warning signs on hard drives.

What is the difference between the raw value and the normalized value?

The raw value is the actual count or reading. The normalized value is a manufacturer score, usually counting down from 100 or 200, compared with a threshold.

Do NVMe drives use the same SMART attributes?

No. NVMe drives report a fixed health log with fields such as Percentage Used, Available Spare and Media Errors instead of numbered attributes.

Get three guides like this every month, plus software update notices.Subscribe →