How to Check SSD and HDD Health with S.M.A.R.T.

What S.M.A.R.T. is, and what it cannot tell you

S.M.A.R.T. is a self-monitoring system built into the drive's own firmware. The drive counts events it considers meaningful — sectors it had to remap, read commands that had to be retried, hours powered on, data written — and exposes those counters to the operating system. Nothing interprets them for you by default; Windows only surfaces a single overall pass/fail flag, which the drive raises when one of its counters crosses a manufacturer threshold.

That single flag is why “Windows says my drive is fine” is weak evidence. The threshold is set to a point at which failure is already very likely, so a drive can accumulate a worrying number of reallocated sectors for months and still report OK. Reading the individual attributes is what turns S.M.A.R.T. from an obituary into a warning.

It also has a hard limit that no amount of reading fixes: S.M.A.R.T. describes wear and error trends, not sudden events. A controller board failure, a firmware bug, a power surge or a laptop drop can take a drive from perfect attributes to unreadable without any intermediate stage. Good attributes mean “no evidence of degradation”, not “safe”, and that is why backups are not optional regardless of what the counters say.

What Windows shows for free

Start with the built-in commands. Each of these runs on Windows 10 and 11 without installing anything; the PowerShell ones need an elevated window to read reliability counters. Run them in the order listed — the first two take seconds and often settle the question.

• wmic diskdrive get model,status,size — the drive's own pass/fail verdict per disk
• Get-PhysicalDisk | Select FriendlyName, MediaType, HealthStatus, OperationalStatus
• Get-PhysicalDisk | Get-StorageReliabilityCounter | Format-List — wear, temperature, error counts
• Get-Disk | Select Number, FriendlyName, HealthStatus, OperationalStatus, Size
• chkdsk C: /scan — an online scan of the file system without taking the volume offline
• Optimize-Volume -DriveLetter C -ReTrim -Verbose — re-issues TRIM on an SSD
• Event Viewer → Windows Logs → System, filtered on the disk and Ntfs sources

Reading the numbers: Windows tools versus the drive's own

Get-StorageReliabilityCounter is the most useful built-in view. On an SSD it typically returns Wear (roughly the percentage of rated endurance consumed), Temperature, PowerOnHours, ReadErrorsTotal and WriteErrorsTotal. On a mechanical drive it returns the error counters and temperature but often leaves Wear empty, since the concept does not apply. Values of zero for the error counters on a drive with thousands of power-on hours is the picture you want.

The raw attribute table with ID numbers is not exposed by Windows itself; a dedicated S.M.A.R.T. reader shows it, and every major drive vendor also ships a free utility for their own hardware — Samsung, Crucial, Western Digital, Seagate and Kingston all publish one, and the vendor tool is the most reliable source for its own drives because it knows how that firmware defines each attribute. Kleaner PRO's drive health page reads the same attributes and presents them per drive in plain language, which is enough to decide whether you have a hardware problem or a software one.

One warning about interpreting raw values: vendors do not normalize them consistently. The same attribute ID can carry a count on one drive, a count plus a timestamp packed together on another, and a manufacturer-specific meaning on a third. Trust the direction of travel — a counter that was 0 last month and is 14 now — more than the absolute number.

Hard drive attributes that matter

For a mechanical drive, five attributes carry almost all of the predictive value. Reallocated Sector Count (ID 05) counts sectors the drive found bad and remapped to spare area; any non-zero value on a consumer drive is a reason to start watching, and a value that grows week over week is a reason to replace. Current Pending Sector Count (ID C5) counts sectors that failed a read and are awaiting a verdict — this is the attribute most associated with files that suddenly will not open, and it should be zero.

• 05 Reallocated Sector Count — remapped bad sectors; zero is healthy, growth is fatal
• C4 Reallocation Event Count — how many times remapping was attempted
• C5 Current Pending Sector — unreadable sectors awaiting reallocation; should be 0
• C6 Offline Uncorrectable — sectors that could not be recovered at all
• C7 UDMA CRC Error Count — transfer errors; almost always a bad SATA cable, not the drive
• 0A Spin Retry Count — the motor needed more than one attempt to spin up
• BB Reported Uncorrectable Errors — errors the firmware could not fix internally

SSD and NVMe: wear, spare area and media errors

An SSD fails differently, so the attributes are different. On NVMe drives the two that matter most are Percentage Used, which counts up from 0 toward 100 as rated endurance is consumed, and Available Spare, a percentage of reserve blocks remaining that comes with a threshold — commonly 10% — below which the drive is formally considered degraded. Percentage Used above 100 is not an immediate failure; the drive is simply past its warranty endurance and should be treated as living on borrowed time.

Media and Data Integrity Errors is the counter that should always be zero. It records uncorrectable errors where the drive returned bad data or could not return data at all, and even a handful is a reason to migrate off the drive. Unsafe Shutdowns counting up is usually benign — it increments whenever the machine loses power without a clean flush — but a high count on a desktop can point at a failing power supply, which damages drives over time.

SATA SSDs expose broadly the same information under vendor-specific attribute IDs, most often as a “Percentage Lifetime Remaining” counting down from 100, plus Total LBAs Written. If the health percentage is falling noticeably month over month on a machine that is not doing heavy writes, something on the system is writing far more than you think — a misconfigured page file on a small drive, a logging application, or a sync client rewriting a large folder repeatedly.

TBW: what the endurance rating actually means

TBW — terabytes written — is the total volume of writes a manufacturer warrants for that model. Consumer drives publish it per capacity: a 1 TB TLC drive is commonly rated somewhere in the low hundreds of terabytes, and larger capacities are rated proportionally higher because the writes spread across more flash. It is a warranty figure rather than a cliff; drives routinely exceed it, and they also occasionally fail well before it for unrelated reasons.

The number that makes it meaningful is your own write rate. Take Total Host Writes from the drive's attributes, divide by PowerOnHours to get an hourly rate, and scale to a year. A typical desktop used for browsing, office work and games writes on the order of a few terabytes per year, which puts a modern consumer SSD's rated endurance decades away and makes wear an irrelevant concern. A machine running virtual machines, video editing scratch files or continuous database work can write an order of magnitude more, and that is where TBW turns into a real planning number.

This is the arithmetic that settles most SSD anxiety. If Percentage Used has moved from 1% to 3% over two years, the drive will outlast the computer. If it moved from 1% to 30% in the same period, find out what is writing — Task Manager's Details tab with the I/O write bytes column enabled, or resmon's Disk tab, will name the process within a minute.

Symptoms S.M.A.R.T. misses, and what to do first

Some failures announce themselves in behaviour before any counter moves. A mechanical drive that clicks rhythmically, spins up and down repeatedly, or disappears from the BIOS until a cold boot is failing regardless of what the attributes say. An SSD that freezes the whole system for seconds at a time, drops to a fraction of its normal copy speed, or remounts itself read-only is also failing — read-only mode is a deliberate last-resort behaviour that lets you copy data off, and it is the clearest possible signal to stop using the drive for anything else.

Event Viewer corroborates all of this. Filter the System log on the disk source and look for event ID 7 (a bad block), 51 (an error during a paging operation) or 153 (an I/O operation was retried). Repeated entries from the same drive, especially clustered around the times the machine stuttered, are stronger evidence than a single attribute reading.

The order of operations at the first bad sign is not negotiable: copy your data off before you investigate further. Diagnostics read the drive, and reading a drive with pending sectors is precisely the workload that turns a recoverable situation into an unrecoverable one. Once you have a verified copy, replace the drive if any of the fatal attributes are non-zero, re-seat or replace the SATA cable if the only problem is CRC errors, and run chkdsk /scan on the file system if the drive itself is clean and only the volume misbehaves.

Questions and Answers

Read next

All articles · FAQ