IT explained · 8 min read
How to check SSD health: SMART data, wear and warning signs
Written for: IT staff and technically confident users checking whether a drive is the cause of a slow or unstable machine, or planning replacements before failure.
Reading drive health
In short
An SSD keeps records about its own condition and will tell you how worn it is if you ask. The figures that matter are percentage used (an estimate of consumed write endurance), total bytes written, available spare blocks, and any reallocated or pending sectors. Read those four and you know whether a drive is healthy, ageing, or about to become a data-loss incident.
Key takeaways
- Percentage used and available spare are the two figures that predict remaining SSD life most reliably.
- SSDs fail differently from hard disks: usually by wearing out writes or dying suddenly, rarely with warning noises.
- A drive reporting reallocated or pending sectors should be scheduled for replacement, not watched.
- Windows built-in tools give a pass/fail answer; reading actual SMART attributes needs a vendor or third-party tool.
- Checking health across an estate remotely is the only version of this that scales beyond a handful of machines.
What SSD health actually measures
Flash memory cells tolerate a finite number of write cycles. The drive's controller spreads writes across the available cells, keeps spare blocks in reserve, and retires cells that stop holding data reliably. It tracks all of this in a set of self-monitoring attributes known as SMART data.
So SSD health is mostly a question of consumed endurance rather than mechanical wear. A drive that has been written to heavily for five years is closer to the end of its life than an identical drive that has mostly been read from, even though both look the same and neither makes a sound.
- Percentage used: the controller's own estimate of consumed endurance. Above roughly 80 per cent, plan replacement.
- Total bytes or host writes: how much has been written over the drive's life, comparable against the manufacturer's rating.
- Available spare: reserve blocks left. A falling figure is a genuine warning.
- Reallocated and pending sectors: cells already retired or suspected. Anything above zero deserves attention.
- Power-on hours and unsafe shutdowns: context, particularly for machines that lose power abruptly.
Checking on Windows
Windows has a quick built-in answer and a slower detailed one. The quick answer comes from wmic diskdrive get status or, on newer builds, Settings, System, Storage, Disks and volumes, which shows an estimated remaining life and available spare for NVMe drives. That is enough to triage.
For the underlying attributes you need a vendor utility (Samsung Magician, Crucial Storage Executive, Intel or Kingston equivalents) or a general SMART reader such as CrystalDiskInfo. PowerShell's Get-PhysicalDisk and Get-StorageReliabilityCounter give scriptable access to wear and temperature, which is what you want if you intend to check more than one machine.
Checking on macOS and Linux
On macOS, System Information under Storage reports SMART status as Verified or Failing, which is a blunt instrument. smartmontools installed through Homebrew gives the real attribute list, and for NVMe drives the nvme-cli tools report percentage used and available spare directly.
On Linux, smartctl -a /dev/sda for SATA drives and nvme smart-log /dev/nvme0 for NVMe give everything, including the wear figures. Both are trivial to run from a remote session and easy to wrap in a script that reports back only the drives worth worrying about.
Reading the numbers without panicking
Context matters. A drive at 15 per cent used after three years in a standard office machine is entirely healthy. The same figure after four months means something is writing constantly, and it is worth finding out what — a misconfigured log, a swap file thrashing because of insufficient memory, or a backup agent rewriting the same data.
Temperature is worth a glance in laptops and thin clients, because sustained heat shortens endurance and NVMe drives in tight enclosures run hot. Sustained figures above the manufacturer's stated range point to a cooling or placement problem rather than a drive problem.
Symptoms that suggest the drive before anything else
SSDs rarely announce themselves. When they do misbehave, the pattern is distinctive: the machine freezes briefly and recovers, files fail to save with input/output errors, the drive suddenly presents as read-only, or the system stops seeing it entirely after a reboot.
A drive that has flipped to read-only is often a controller protecting the remaining data. Treat it as a recovery job, not a repair job: copy what matters off it first, and replace it.
A five-minute health check
1. Get the pass/fail answer
Windows: wmic diskdrive get status. macOS: System Information, Storage. Linux: smartctl -H. Anything other than OK or Verified moves straight to the backup step.
2. Read percentage used and available spare
Use the vendor utility, CrystalDiskInfo, Get-StorageReliabilityCounter, smartctl or nvme smart-log. These two figures tell you where the drive is in its life.
3. Check for retired sectors
Any reallocated, pending or uncorrectable count above zero means cells have already failed. Schedule a replacement rather than monitoring it hopefully.
4. Sanity-check total writes against age
Unusually high writes for the machine's age means something is writing constantly. Fix the cause or the replacement drive wears out just as fast.
5. Confirm the backup before doing anything else
Verify a recent restorable backup exists for that machine. On an ageing drive this is the only step that cannot be repeated later.
6. Record and re-check
Note the wear figure with the date. The rate of change over a few months is far more useful than any single reading.
Reading the figures
| Percentage used under 50 | Healthy. Re-check annually. |
|---|---|
| Percentage used 50 to 80 | Ageing. Include in the next refresh cycle. |
| Percentage used above 80 | Plan replacement now. |
| Available spare falling | Replace. Cells are being retired actively. |
| Reallocated or pending above zero | Back up and replace. |
| Drive is read-only | Recover data immediately; the drive is protecting itself. |
Select any column heading to sort, or filter with the box above.
Common mistakes
- Trusting a single pass/fail status, which stays green until very late on many drives.
- Running a full surface test on a drive that is already failing, adding load when you should be copying data off.
- Replacing a worn drive without finding out what wrote it out, so the new one follows.
- Checking health only when someone complains, rather than as a scheduled estate-wide report.
Frequently asked questions
- How do I check SSD health on Windows?
- For a quick answer, run wmic diskdrive get status or open Settings, System, Storage, Disks and volumes. For real detail use a vendor utility, CrystalDiskInfo, or PowerShell's Get-StorageReliabilityCounter.
- What is a good SSD health percentage?
- Anything reporting under 50 per cent used is comfortable. Between 50 and 80 per cent, plan for replacement in the normal refresh cycle. Above 80 per cent, replace it deliberately rather than waiting.
- How long does an SSD last?
- In typical office use, well beyond the machine's service life: commonly five to ten years. Write-heavy workloads such as databases, video editing or busy logging consume endurance far faster.
- What are the signs of a failing SSD?
- Brief freezes that recover, input/output errors when saving, files that will not open, the drive turning read-only, or the system failing to detect it after a reboot.
- Can I check SSD health remotely?
- Yes. The same commands run fine over a remote session, and the scriptable versions on Windows, macOS and Linux can be run across an estate so you see wear before a user notices anything.
How this works in 247connect
Checking drive wear one machine at a time does not scale. With 247connect, an operator can open a session on any managed device in around eight seconds and run the same check without arranging access first.
More IT explainers
What is a data breach?
A data breach is unauthorised access to information you hold. How breaches actually begin, the four stages they follow, why detection takes so long, the first hours of response, and the controls that reduce the damage most.
What is a botnet?
A botnet is a network of compromised devices under one operator's control, used for DDoS attacks, spam, credential stuffing and crypto mining. How devices get recruited, why infections go unnoticed, and the signs worth watching for.
What is patch management?
Patch management is the process of finding, testing, deploying and verifying updates across an estate. The five-stage cycle, deployment rings, realistic timescales, and why verification is where most patching programmes actually fail.