No single SMART field proves that an SSD is healthy. A green badge, a high “health” percentage or a low write count can all be useful, but each is only one part of the evidence. The sensible check combines the drive’s own warnings, its error history, written endurance, behaviour under ordinary use and a backup of anything that matters.
If the drive is disappearing, becoming read-only, reporting a critical warning or producing new media errors, protect the data first. Do not make a failing drive work harder just to complete every test in this guide.
Start with a three-minute first check
Identify the exact model, capacity and firmware version before interpreting anything. The capacity matters because two drives with the same family name can have different endurance ratings. Then open a reputable read-only health tool: the manufacturer’s own utility is a good starting point when one exists, while Windows and Linux also expose drive reliability information in different ways.
Record the result rather than taking a photograph of only the large health badge. You want the model, overall warning state, percentage used or remaining life, total host writes, temperature, power-on hours, unsafe shutdowns and media or data-integrity errors. Some drives expose more; some USB enclosures and storage controllers expose less.
Use an evidence ladder, not one percentage
- Critical device warnings: a non-zero NVMe critical warning or an explicit manufacturer alert deserves immediate attention.
- Errors and real behaviour: new media errors, repeated disappearance, read-only operation or corrupted reads outweigh a reassuring headline percentage.
- Endurance context: percentage used and total writes become meaningful only against the exact model and capacity specification.
- History: temperature exposure, unsafe shutdowns and power-on hours add context, but do not independently condemn a drive.
- Tool presentation: a green “Good” label is a summary chosen by the software, not a separate measurement supplied by the NAND.
This order matters. A drive does not become trustworthy because a neat dashboard says 98% if the underlying log contains an active reliability warning.
SMART interpretation table
| Field or symptom | What it can tell you | Concern trigger | Next verification |
|---|---|---|---|
| Critical warning | NVMe summary flags for spare capacity, temperature, reliability, read-only mode and related conditions. | Anything other than zero, or a named warning in the vendor utility. | Back up important data and decode the individual warning before further load. |
| Percentage used or life remaining | An endurance estimate produced from the drive’s own model. | A rapid change, a value at or beyond its rated life, or conflict with other evidence. | Compare with total writes, the exact specification and error history. |
| Available spare | The remaining reserve capacity available to replace unusable NAND blocks. | At or below the drive’s reported spare threshold. | Check the critical-warning state and manufacturer guidance. |
| Data or host writes | How much data the host has asked the SSD to write, although units vary by interface and tool. | Only meaningful when unusually high relative to the exact model’s rating. | Convert using the tool’s documented units, then perform a model-and-capacity check. |
| Media and data-integrity errors | Errors the controller associates with media or data integrity. | Non-zero, increasing, or accompanied by failed reads and warnings. | Preserve data and investigate with the manufacturer’s diagnostics. |
| Unsafe shutdowns | Power losses or shutdowns the drive judged unsafe. | A growing count that matches freezes, cable faults or sudden power loss. | Check power, cabling, enclosure and system stability; the count alone is not proof of NAND failure. |
| Temperature | Current temperature and, on some drives, time above warning thresholds. | Sustained threshold events or throttling under an ordinary workload. | Check airflow, heatsink contact and the drive maker’s permitted range. |
Compare writes with the correct TBW figure
TBW means terabytes written. Manufacturers commonly use it as one limit in a warranty or endurance specification, but it is not a countdown timer that makes the drive fail at the next terabyte. Equally, being far below the figure does not rule out a controller fault, firmware problem or accidental damage.
Find the specification for the exact model and capacity. Do not borrow the rating from a larger version of the same range. Check whether your tool reports bytes, host writes, NVMe data units or an already-converted terabyte figure before comparing the numbers. The Samsung SSD warranty tables, for example, show why both model and capacity need to match.
A useful assessment states the evidence plainly: “exact 1TB model, approximately X TB written, manufacturer rating Y TBW, no active warning, no recorded media errors”. That is more informative than “97% healthy”.
Allow for SATA and NVMe differences
NVMe defines a standard health log with fields such as critical warning, percentage used, available spare and media errors. The NVM Express NVMe-CLI guide explains those fields and the device self-test facility.
SATA SMART attributes are less uniform. Two manufacturers can assign different meanings or scales to a raw value with the same-looking label. The smartctl documentation distinguishes raw and normalised values and describes the limitations of access through some controllers. Prefer the maker’s interpretation for the exact drive rather than copying a generic threshold from a forum.
A missing field is not a clean bill of health. USB bridges, RAID controllers, operating-system permissions and tool support can all hide information. If the health log cannot be read, record that limitation instead of turning “unknown” into “good”.
Warning signs that matter outside SMART
- The SSD intermittently disappears from the operating system or firmware.
- Ordinary reads fail, files repeatedly become corrupt or the drive unexpectedly becomes read-only.
- A manufacturer diagnostic or supported self-test fails.
- New media or data-integrity errors appear between two recorded checks.
- The system reports a storage reliability warning or the drive remains excessively hot in normal use.
Some of these symptoms can come from power, a SATA cable, an enclosure, a controller or the wider system. That is why the next step is controlled diagnosis, not a confident guess that the NAND is dead. Windows’ Get-StorageReliabilityCounter can expose temperature, errors, wear and usage information where the device and driver provide it.
CHKDSK checks a file system; it does not restore worn flash cells or prove SSD endurance. A successful file-system check and a healthy device log answer different questions.
Reach a three-signal decision
For a drive you plan to keep, look for agreement between three signals: no active critical warning, no concerning error trend or real-world symptom, and endurance that makes sense for the exact model. Keep a backup regardless. For a drive you are considering buying, ask for the full health record and exact label rather than a cropped green badge.
For a drive you want to sell, save the model and capacity, the unedited health output and a short description of any symptoms. Our used SSD selling guide explains what to include, while the wider computer-component quote route covers mixed hardware. A member of our team can provide a personalised quote in as little as 10 minutes once the item is identified.
The honest outcome may be “no warning found, but the available evidence is incomplete”. That is still a better decision than pretending one attractive percentage can predict the future.
Sources and scope
Primary technical guidance checked on 12 September 2026. The exact SSD manufacturer’s specification and diagnostic guidance take precedence. This page interprets condition evidence; it does not cover secure erasure, physical compatibility or data recovery.