How to check the health of an SSD or hard drive (and which values really matter)

A drive rarely dies without warning. Hard drives build up bad sectors for weeks before giving up; SSDs wear with every write and keep count of how much life they have left. All of this is recorded on the drive itself, in its SMART data. The problem is that nobody looks at it until the PC no longer boots.

Here’s how to read it on Windows, Mac and Linux, which values deserve attention and which only look scary.

On Windows 11 without installing anything

For NVMe SSDs, the “chewing-gum” drives in recent PCs, Windows has a little-known screen:

1. open Settings > System > Storage;
2. click Advanced storage settings > Disks & volumes;
3. choose the drive and click Properties.

Under Drive health you’ll find the estimated remaining life, the available spare and the temperature. A yellow warning means Windows found a value over the threshold.

With PowerShell, for every drive

Open PowerShell as administrator and type:

Get-PhysicalDisk | Get-StorageReliabilityCounter | Select-Object DeviceId, Wear, Temperature, ReadErrorsTotal, WriteErrorsTotal, PowerOnHours

The Wear column is the wear percentage reported by the drive (0 = new), PowerOnHours the hours it has been on. Some drives don’t fill in every field: a blank value isn’t an error.

The tool I recommend: CrystalDiskInfo

For a full picture on Windows, the reference tool is the free CrystalDiskInfo. Download the standard edition from the official site (avoid download sites with “enhanced” installers). It shows a Health Status for each drive: Good, Caution or Bad, with a percentage for SSDs. Below is the attribute table, which is where you really see what’s going on.

The values that matter

For traditional hard drives and SATA SSDs

• 05 – Reallocated Sectors Count: damaged sectors the drive has replaced with spare ones. A low, stable value can be tolerated; a value that grows week after week is the most reliable sign of a coming failure.
• C5 – Current Pending Sector Count and C6 – Uncorrectable Sector Count: sectors the drive couldn’t read. Even a few are a serious warning, especially alongside reallocated sectors.
• C7 – UltraDMA CRC Error Count: communication errors between drive and motherboard. They look serious but almost always come from a faulty or loose SATA cable. Replace the cable: if the number stops growing, the drive is fine.

For NVMe SSDs

• Percentage Used: how much of the life rated by the manufacturer has been used. Reaching 100% doesn’t mean the drive switches off, only that it has passed its rated endurance.
• Available Spare: the spare cells left. It must stay above the threshold shown next to it; if it gets close, it’s time to replace the drive.
• Media and Data Integrity Errors: should be zero. Any other value means back up immediately.
• Total Host Writes: the data written over the drive’s life. Compare it with the TBW rating in the manufacturer’s spec sheet. Example: a 1 TB SSD rated for 600 TBW that has written 60 TB has used about 10% of its rated endurance.

For perspective: normal home use writes a few dozen terabytes a year. With a modern SSD, the end of rated endurance almost always comes after the computer has been replaced for other reasons.

On a Mac

The basic check is in System Information (hold Option, click the Apple menu, then System Information): under NVMExpress or SATA, S.M.A.R.T. Status should say Verified. It’s only a yes or no, though. For details, install smartmontools with Homebrew (brew install smartmontools) and run smartctl -a disk0.

On Linux

Install the smartmontools package and use:

• sudo smartctl -a /dev/nvme0 for an NVMe SSD, sudo smartctl -a /dev/sda for a SATA drive;
• sudo smartctl -t short /dev/sda to start a short test (a couple of minutes); the result then appears in the Self-test log section of the previous command.

When to replace the drive

• reallocated or pending sectors increasing on a hard drive;
• integrity errors on an NVMe drive, or available spare close to the threshold;
• a “Caution” or “Bad” status that isn’t caused by a cable;
• rhythmic noises, clicking or freezes while copying files, on a mechanical drive.

The most important advice, though, is something else: SMART data warns you in many cases, not all. A controller that fails suddenly or a drive killed by a power surge gives no warning. That’s why checking the drive doesn’t replace a backup: it complements it. If you don’t have one, start with the 3-2-1 rule: three copies, on two different media, one of them off site.

Questo articolo è disponibile anche in italiano: leggi la versione italiana.