With an SSD there is no degrading as with a hard drive: no noise and no gradual slowdown. The useful signals are all in the numbers, and only a few of them matter.
These are the three attributes to watch: percentage used, reallocated (bad) or reserved blocks, and uncorrectable errors. The other few dozen will be white noise for most users.
Windows built-in options are limited: wmic diskdrive get status provides a simple yes/no result and Optimise-Drives will indicate whether TRIM is enabled. For actual numbers to be reported you can use CrystalDiskInfo, which pulls data from the SMART attribute directly, or your manufacturer’s tool (Samsung Magician, Crucial Storage Executive, WD Dashboard) which accurately interprets vendor specific values as opposed to generic tools guessing.
On macOS, the System Information app will report a general SMART status. However, if you want to know what is really going on with the drive you’ll have to run smartmontools through Terminal. Apple has intentionally limited the information they provide regarding health and reliability of SSDs, and due to the soldering of the Apple Silicon SSDs, there is also very little in terms of statistics available.
Percentage used (Media Wearout Indicator) indicates an approximate amount of writing endurance of the SSD that has been utilised. Less than 80%, do not worry about replacing. More than 80%, begin planning for replacing. Although, please keep in mind that this is a ‘wear’ estimation rather than a prediction of when the drive may fail, and many times these drives last longer than their predicted “end-of-life”.
Reallocated or reserved block count measures the total amount of excess storage space that has been used to “retire” faulty blocks. Normal is a slow increase in this number over the years. An unexplained spike in this number requires immediate attention.
Uncorrectable error count is the biggest indicator of potential failure. This indicates there are files that have been corrupted by bad sectors and were unable to be repaired via the drive’s built-in error checking/correction software, which means real data loss, however small. Any non-zero reading of this metric increasing over time will indicate your disk needs to be replaced rather than continued monitoring.
SMART reporting on an SSD reports how much flash is wearing out but reports very little about how much damage the controller (the brains of the SSD) has taken. Most catastrophic failures of an SSD occur in either the controller’s hardware, or its firmware. If that happens there will be almost no warning: it was fine on Monday and gone by Tuesday. A perfect SMART report simply shows that the flash is still good; it doesn’t show that your SSD is good.
That’s why the ‘practical’ approach to using SSDs will be different to using traditional hard drives: when your data resides on a mechanical disk (a ‘spinning disk’) you can usually hear if something is wrong and take action. When your data is stored on an SSD you won’t know anything is wrong until it is too late; therefore, backups have to be made before things go wrong as opposed to having backups organised at the point of discovering problems.
If the controller fails but all of the NAND storage remains viable, then many times the drive can be accessed via the vendor interface and then copied. This would be a recoverable scenario; it is also very common. If the flash memory has degraded or if files have been removed from a functional drive and those files have been subsequently deleted due to TRIM then there may be nothing left to read. The cost of recovering your SSD/NVMe drive is from £300 +VAT after we’ve performed our complimentary 48-hour diagnostic process which will tell you what is happening with your specific SSD/NVMe device.
Use CrystalDiskInfo (Windows) or smartmontools (macOS) to read SMART information from your drive. Or check using software provided by the manufacturer. This will help you understand the specific SMART attributes as they are defined by each vendor. Look at these 3 metrics: percentage used, reallocated/reserved blocks and uncorrectable errors.
This indicates how much of your drive’s total write cycle (endurance) has been expended. Less than 80% is normal. More than 80%, get ready to replace. While this metric measures an estimate of wear on the drive, many drives have written beyond their estimated endurance before failing.
A sudden increase in reallocated/reserved block count. An increasing number of uncorrectable errors over time. Uncorrectable errors are the most alarming type of error. If the error correction process cannot correct errors then there is a loss of data, even if it is just a few bytes.
Not entirely. Flash wear is reported fairly accurately but reporting controller health is marginal to none. Controller failure is typically the largest single cause of catastrophic failures without notice. So while your SMART results indicate that the flash has not reached end-of-life, it does not mean the entire drive is safe.
Often, yes. When the controller has failed but the NAND on the drive remains good, the drive can be accessed in vendor mode (for an image) to be read as read only. However, if the flash has degraded or TRIM has already removed deleted files, nothing may remain. We provide a complimentary 48-hour evaluation of whether anything remains.