How to check SSD health — and what it can’t tell you.

For all SATA and NVMe devices connected to a Windows computer, CrystalDiskInfo can show the health percentage and temperature of the device on the main page. Manufacturers’ own software goes a level deeper into testing for their products, including Samsung Magician, Crucial Storage Executive and WD Dashboard. These programs add two additional items to test: the amount of total bytes written to the device, and whether firmware updates have been applied. When using a Mac, Disk Utility provides only a single word SMART answer, but when running smartctl from smartmontools, the program produces the complete NVMe health log, even when used with newer Apple silicon devices. Regardless of which program you use, you’re viewing the exact same reported ledger.

There are four specific lines to pay close attention to. Percentage Used (which mirrors the ‘Health’ value and displays how much of NVMe’s wear capacity has been consumed; 0% is “as new”). Data Written compared to the drive’s rated endurance (TBW rating printed on each manufacturer’s product specification). Available Spare: the remaining number of replacement blocks of flash memory. Healthy SSDs are nearly always reporting 100% availability of spare, so a decreasing availability rate is similar to increasing numbers of “reallocation” entries being recorded on a mechanical hard disk. Media and Data Integrity Errors: should report no errors; any increase indicates either faulty hardware (such as failing flash), or controller issues, regardless of the overall percentage claimed by the health-gauge.

Hard drives fail in a manner similar to engines. There are warnings, slow downs, and often loud noises, followed by the complete failure. Solid state drives tend to fail in a fashion similar to light bulbs. The controlling component (the controller chip) and/or the firmware that operates this component becomes faulty. As a result, a disk that was scoring at 98%, will no longer be visible in the BIOS. None of the SMART attributes were indicating potential issues with the disk, as the failed components were responsible for collecting and reporting their own status. Following the controller failure, there are two possible outcomes. The first consequence is that even though an individual disk may have passed testing and received a high health rating, it should never be relied upon as a substitute for a backup. In fact, due to the nature of SSD failures being silent prior to failure; it would appear that backups become much more critical when using solid state drives rather than less. The second consequence is that most of the time, the “disappearance” of a disk does not mean data loss. Typically, behind each failed controller are the NAND flash chips which retain all the information stored within them. Thus, SSD not detected recovery is simply the process of fixing/repairing the firmware issue causing the non-detection, or extracting data from the NAND flash chips via direct access to read the contents and re-assemble what the controller has broken apart.

In addition to being aware of various behaviours exhibited by SSDs, another behaviour of modern SSDs worth noting is something called TRIM. In essence, TRIM represents a means whereby systems notify an SSD to delete/erase blocks containing deleted file information immediately and permanently. Due to this behaviour, most ‘undelete’ software that is designed to restore deleted files and works well on hard drives and memory cards will result in nothing but empty results when attempting to undelete files on an internal SSD. Therefore, if you wish to attempt restoring deleted files from an SSD, your best option is creating a backup of your entire drive prior to deletion and utilising that backup as your recovery option.

Everything zero, spare at 100%: note your “Data Written” figure somewhere and verify it twice per year. That is the entirety of your maintenance routine. Spare falling, media errors appearing, or the drive stuttering with system freezes: that is amber. Copy your files off now while it reads, then replace it; do not waste any further goodwill on that drive via benchmark passes or vendor “repair” scans. Vanished, wrong capacity, or locked read-only: the controller has had its event. Do not power cycle, skip firmware update roulette, and send it to a bench. Our SSD & NVMe recovery service exists for exactly this ending.

The health-check has to earn its value most at the time of purchasing. Are you looking at a used laptop or a secondhand hard-drive? Run the reading before payment is made or in your return window: The amount of data written against the model’s Total Bytes Written (TBW) rating will tell you if this is going to be a very lightly used drive or a drive that was used for 3 years in a mining application. Both may look the same on the outside but could be thousands of terabytes different on the inside. Sellers typically do not have objections to running a two minute CrystalDiskInfo scan and if they do, their objection is often an indication of what you are getting into. The same applies to ‘refurbished’ drives: a reset operating system does not reset the wear ledger.