How to Check SSD Health: SMART, TBW and Free Tools

Unlike a hard drive, an SSD rarely gives you an audible warning before it fails. There is no clicking head, no grinding platter — just a drive that works perfectly one day and is gone the next. The good news is that every modern SSD monitors itself constantly, and learning how to check SSD health takes about five minutes. This guide shows you which numbers actually matter, which free tools to trust, and what to do when the numbers start looking bad.

Why SSD health checks are worth your time

SSDs wear out in a predictable way: every NAND flash cell survives a finite number of write cycles. Drive firmware tracks this wear and reports it through the SMART system (Self-Monitoring, Analysis and Reporting Technology). Checking it twice a year is enough for most people — and it is the difference between migrating your data calmly on a weekend and losing it on a Tuesday morning.

There is also a financial angle. In a market where NAND prices are climbing, replacing a drive prematurely wastes money, while running a dying drive risks data that is worth far more than the hardware. A quick health check tells you which situation you are actually in. If your drive is healthy but you are simply running out of room, a capacity upgrade is the smarter move than a panic replacement — our breakdown of 2TB vs 4TB SSD capacity can help you decide how much space to buy.

The SMART attributes that actually matter

SMART exposes dozens of attributes, but only a handful are worth your attention. Different vendors name them slightly differently, and NVMe drives report a shorter, standardized set than older SATA drives. Here are the ones that tell you the real story:

The single most informative number is Percentage Used (sometimes shown as "SSD Life Left" or "Wear Leveling Count"). It tells you what fraction of the drive's rated write endurance you have consumed. A drive at 12% used after three years is barely broken in; a drive at 85% used deserves a replacement plan.

Free tools to check SSD health

You do not need to pay for monitoring software. The best options are free, and each platform has a strong default choice:

On Windows, start with CrystalDiskInfo: it shows a simple "Good / Caution / Bad" status plus the raw values behind it. For NVMe drives on Linux, sudo nvme smart-log /dev/nvme0 gives you the authoritative health log straight from the controller, including percentage used and temperature extremes.

Understanding TBW and percentage used

TBW (terabytes written) is the manufacturer's endurance rating — the total amount of data you can write to the drive over its warranty life. A typical 1TB consumer TLC drive is rated around 600 TBW. Your actual lifetime writes are reported by the drive itself, so the math is simple: if you have written 60 TB to a 600 TBW drive, you have used roughly 10% of its rated life.

In practice, most desktop users write 5–15 TB per year, which means a 600 TBW drive lasts decades on paper. Heavy workloads — video editing scratch disks, Chia-style plotting, constant VM churn — can push 50–100 TB per year and genuinely wear a drive out. Endurance matters far more in write-heavy scenarios, which is why our guide to SSD endurance in the AI era digs into DWPD ratings and workload planning for demanding users.

Step by step: checking your drive right now

Windows

Download CrystalDiskInfo, run it, and select your SSD from the drive list. Note the health status, temperature, total host writes, and power-on hours. Then open your SSD maker's own toolbox (most major brands offer one) to see the vendor-specific "life remaining" figure and check for firmware updates.

macOS

Open Disk Utility, select the physical drive, and look at SMART status — though macOS only shows a pass/fail verdict. For real numbers, install smartmontools via Homebrew and run smartctl -a /dev/disk0, or use DriveDx for a readable dashboard.

Linux

Install smartmontools and nvme-cli from your package manager. For NVMe: sudo nvme smart-log /dev/nvme0. For SATA: sudo smartctl -a /dev/sda. Look for "Percentage Used", "Data Units Written", and the temperature fields.

Warning signs that mean act now

Any one of these deserves immediate action — back up first, diagnose second:

Media and data integrity errors above zero. On NVMe drives this counter should never move. Non-zero means the drive has failed to read data even with error correction.

A growing reallocated block count. A few retired blocks are normal over years; a count that climbs week to week means NAND is dying faster than the spare pool can absorb.

Percentage used approaching 100%. The drive will not necessarily die at 100%, but you are past the rated life and running on borrowed time.

Frequent freezes, disappearing drive, or files that corrupt. Symptoms trump numbers — if the OS intermittently loses the drive, trust the symptom over a "Good" SMART status.

What to do when health is declining

First, verify your backups are current and complete — not "mostly", completely. Then plan a migration rather than an emergency reinstall: cloning your existing drive to a new SSD preserves your OS, applications, and settings exactly as they are. Our walkthrough of cloning a drive to a new SSD without reinstalling covers the process step by step, including how to handle larger or smaller target drives.

When buying the replacement, consider sizing up if you were regularly above 80% full — a fuller drive wears faster because the controller has less free NAND to work with for wear leveling and garbage collection.

Who should check regularly — and who can relax

Check quarterly if you: run a write-heavy workload (video editing, databases, local AI training), use QLC drives as primary storage, operate without complete backups, or run a home server or NAS where drives work 24/7.

Twice a year is plenty if you: are a typical desktop user with a TLC drive, keep good backups, and mostly read data rather than write it. Your drive will almost certainly become obsolete before it wears out.

You can mostly relax if: the drive is under two years old, percentage used is in single digits, and temperatures are normal. Just make sure backups exist — health monitoring is not a substitute for them.

Reading a real SMART report: a worked example

Abstract attribute names become much clearer with a concrete example. Below is what a typical three-year-old 1TB NVMe drive looks like in nvme smart-log output, annotated the way a technician would read it:

The verdict on this drive: healthy, with years of life remaining. The only action item is the temperature peak — 68°C is not dangerous, but it suggests the drive works hard during long writes, and a five-minute check of case airflow would be cheap insurance. This is the entire value proposition of SMART in one paragraph: numbers that turn vague anxiety into a specific, small task.

Note how the fields corroborate each other. Low percentage used plus 100% available spare plus zero media errors is a consistent healthy picture. When fields disagree — say, low percentage used but a climbing reallocated count — trust the worse number. Percentage used measures expected wear; reallocated blocks measure actual NAND defects, and defects do not care about the rating. Save or screenshot this first report, too: a baseline captured when the drive is healthy makes every future comparison dramatically easier, because you will be looking at changes rather than trying to remember what "normal" looked like.

SMART on external and USB-connected SSDs

Here is the catch that surprises many users: put that same healthy drive in a USB enclosure and your monitoring tools may suddenly report nothing. Whether SMART data passes through depends on the enclosure's bridge chip and whether it supports the right passthrough commands.

Modern bridge chips based on the Realtek RTL9210 or JMicron JMS583 generally pass SMART through correctly over UASP, and most tools will read the drive as if it were internal. Older or bargain-bin enclosures using outdated firmware often block it entirely — CrystalDiskInfo will show the drive but with blank health fields, or the drive may not appear at all.

If you rely on a portable SSD for important data, test SMART passthrough before you need it: connect the enclosure, open your monitoring tool, and confirm you can see percentage used and temperature. If the enclosure blocks SMART, you have three options: swap to an enclosure with a known-good bridge chip (the chip model is usually listed in the product specs or reviews), accept that you will monitor the drive only when it is installed internally, or treat the external drive as expendable storage with rigorous backups. For a drive holding a photo archive or a backup set, the first option is worth the small premium.

Thunderbolt enclosures behave better on average, since Thunderbolt exposes the NVMe device more directly, but the rule is the same: verify, do not assume.

Automating health monitoring so you do not have to remember

Twice-a-year manual checks work only if you actually do them. For drives that matter — a NAS, a home server, a workstation with client work on it — automate the watching:

On Linux and NAS systems, use smartd. It is part of smartmontools and can run scheduled self-tests plus email you when attributes cross thresholds. A minimal configuration that watches all drives and mails warnings takes about ten minutes to set up and then runs silently for years. Most dedicated NAS operating systems expose the same capability in their GUI under disk health or SMART tasks.

On Windows, schedule a script. A small PowerShell script that queries SMART via WMI or parses smartctl output, run monthly through Task Scheduler, can log percentage used to a file. You do not need a fancy dashboard — a text log you glance at twice a year beats a dashboard you never open. Some manufacturer toolboxes also include background monitoring with tray warnings; if yours does, leave it enabled.

Set thresholds that mean something. Alert at 70% used (start planning), 85% used (buy the replacement), and immediately on any media error or a reallocated count that increases between checks. Temperature alerts at 70°C sustained catch cooling problems before they become throttling problems.

The goal is not to obsess over numbers — it is to convert monitoring from a chore you forget into a system that taps you on the shoulder only when something actually needs attention.

Health checklist before buying a used SSD

A used SSD can be a genuine bargain, but only if you verify what you are buying. NAND wear is invisible from the outside, so the SMART report is the inspection:

Ask the seller for a SMART screenshot before paying. A legitimate seller with nothing to hide will provide one. Refusal is itself an answer. Look for percentage used under 30%, zero media errors, a stable reallocated count, and power-on hours consistent with the claimed history.

Do the wear math against the price. A drive at 60% used has roughly 40% of its rated life left — price it accordingly, not against a new drive. As a rule of thumb, do not pay more than half the new price for a drive past 50% used, no matter how healthy it looks otherwise.

Watch for ex-mining and ex-plotting drives. These workloads write staggering amounts of data. A drive that spent a year plotting can show 90%+ used while looking cosmetically perfect. The SMART report exposes this instantly — which is exactly why a seller who will not share one should be avoided.

Secure-erase on arrival. Once the drive is yours, run the manufacturer's secure erase before trusting it with data. This resets the NAND to a known-good state, restores performance, and — on most drives — gives you a clean baseline for future health comparisons. Then run your own SMART check to confirm the numbers match what the seller showed you.

Remember the warranty usually does not transfer. Most SSD warranties apply to the original purchaser. Factor the absence of warranty coverage into the price, and never put irreplaceable data on a used drive without a backup somewhere else.

FAQ

How often should I check my SSD health?

Twice a year for typical use; quarterly for write-heavy workloads like video editing, servers, or AI training. Set a calendar reminder — the check itself takes five minutes.

What is a good SSD health percentage?

Above 90% remaining life is excellent. Between 70–90% is fine with years of service left for normal use. Below 50% remaining, start planning a replacement within the next year or two depending on your workload.

Can a "Caution" status in CrystalDiskInfo be ignored?

Sometimes — it often triggers on a single threshold like reallocated sectors or high temperature. Investigate which attribute triggered it: a temperature warning may just mean poor case airflow, while reallocated sectors warrant a real replacement plan.

Does a full SSD wear out faster?

Yes, indirectly. A nearly full drive forces the controller to work harder during garbage collection and wear leveling, increasing write amplification. Keeping 15–20% free space is the simplest endurance habit you can adopt.

Will the manufacturer warranty cover a worn-out drive?

Only within the warranty terms — most warranties cover a set number of years or the TBW rating, whichever comes first. A drive that hits 100% of its TBW rating in year two is typically out of warranty even if the five-year period has not elapsed.

Why does my SSD report 100% health but feel slow?

SMART health measures endurance and error counts, not performance. A drive can be perfectly healthy yet slow because it is nearly full (high write amplification), overheating (thermal throttling), attached via a slower interface than you assumed, or simply a DRAM-less or QLC design hitting its sustained-write limits. Check temperatures and free space before suspecting the NAND.

Can SMART data be faked on a used drive?

Not through any normal means — the counters live in the drive's firmware and increment monotonically. Extremely sophisticated attackers could theoretically reflash firmware, but this is vanishingly rare in the consumer used market. The realistic scam is simpler: a seller who refuses to share a SMART report at all. Treat missing data as a failed inspection.

What temperature is actually dangerous for an SSD?

Occasional peaks to 70°C under heavy load are normal and harmless. Sustained operation above 75°C is where controllers begin throttling to protect themselves, and regular excursions past 80°C suggest a cooling problem worth fixing — usually case airflow or a missing heatsink rather than a faulty drive. NAND itself tolerates heat better than the controller does, so the controller temperature is the number to watch.

Do I need to check health on a brand-new drive?

One baseline check at installation is worthwhile: confirm zero media errors, 0–1% used, and sensible temperatures. This takes two minutes and gives you a reference point for every future check. After that, the drive needs no attention until your regular twice-yearly routine — or until it gives you a reason.