An SSD reports dozens of SMART attributes. Four of them carry almost all of the signal, and knowing which four is the difference between a useful five-minute check and a screenful of numbers nobody acts on.
The four that matter
Percentage used, or wear levelling count. How much of the drive's rated write endurance has been consumed, counting up toward 100. It plans your replacement. It does not predict tonight.
Available spare. Every SSD ships with reserve blocks it quietly uses to replace failing ones. This is the pool. Its whole purpose is to hide degradation from you until it can no longer hide it, which is exactly why a falling value is worth more attention than a low one.
Media and data integrity errors. Unrecoverable read errors. Any non-zero value deserves a look. A value that grows deserves a backup, today.
Unsafe shutdowns. Not a fault, context. A high count on a drive with rising errors points at a power supply or a cable rather than at the flash.
Power-on hours and total bytes written are history, not prognosis. They are interesting and they predict very little on their own.
How to read them
Linux and macOS. smartctl -a /dev/sdX from the smartmontools package. On NVMe drives, use
smartctl -a /dev/nvmeX to get the NVMe health log, which is where percentage used and available
spare live.
Windows. wmic diskdrive get status returns a single OK or not, which is not enough to act on.
A small utility such as CrystalDiskInfo shows the same attributes the other platforms expose.
Whichever tool you use, go to the raw attribute values. The overall verdict is a summary of thresholds set by the manufacturer, and thresholds are the part you least want to delegate.
This is the right mental model, including its limits. The light comes on when a specific condition is detected. Nothing about a dark dashboard promises the car is fine, and nobody has ever been reassured by a warning light that stayed off after the failure.
The asymmetry nobody tells you about
A SMART fail is strong evidence of trouble. A SMART pass is weak evidence of health.
The reason is structural. SMART measures the flash. It does not meaningfully measure the controller or the firmware, and controller failure is one of the most common ways an SSD dies. That mode of failure produces no gradual signal, no rising counter, no warning at all. The drive works, and then the computer does not see it.
So when a guide tells you to run a health check monthly and relax if it passes, the second half of that advice is doing damage. The check tells you when to act early. It never tells you that you can stop having backups.
What to actually do with the numbers
- Available spare falling week over week -> copy your data off now, replace the drive.
- Media errors above zero and rising -> back up today, plan replacement.
- Percentage used past 80 -> budget for a replacement, no emergency.
- Everything flat for months -> nothing to do, and your backup strategy is still the thing that saves you.
Trend beats absolute value on all four. A number you read once is a number you cannot interpret, which is the strongest argument for checking twice a year rather than in a panic.
Where this stops being a health check
If the drive is already not being detected, making noises, or mounting read-only, this article is the wrong one. Reading SMART from a failing drive means powering it on and issuing commands to it, and on a drive that is actively degrading, every power cycle can cost you data. At that point the priority is a bit-for-bit image of the drive onto healthy storage, and the health check comes after, on the copy.
Everything above you can verify yourself in five minutes, on your own hardware, with tools that cost nothing. That is the only kind of check worth building a habit around.
Recover your deleted files → EaseUS
Free scan · deleted, formatted & lost files · Windows & Mac
