There are two potential risks when using an SSD. A risk you can see coming slowly is flash wear. Every single health test available will tell you this. But there is also a switch which has given you absolutely no warning: the controller. In order to test an SSD fully you have to read the slow gauge and always remember the silent switch that follows it.
Percentages related to health tracking can accurately reflect flash-wear but have absolutely no bearing on either controller or firmware faults which are common reasons why SSDs fail. Don’t interpret this number as an indicator of how long your SSD will function; simply use it as an indicator of current status.
When using Windows simply downloading CrystalDiskInfo will give you everything you need. This program lists every SATA and NVMe drive with the drive’s health percentage, the temperature of the drive, as well as the entire attribute table below. If your SSD is manufactured by Samsung, Crucial, or WD, their respective software tools (Magician, Storage Executive, as well as the WD Dashboard) will provide you with your drive’s total number of writes throughout its lifetime and allow you to update the firmware, but in terms that require no user manual. For use on a Mac, Disk Utility provides you with the SMART status, and smartctl (which is part of smartmontools and can be easily installed via Homebrew) provides you with the remainder including the built-in storage of Apple silicon.
Regardless of which tool you are reading from, it is four values that perform all of the actual “work” in terms of measuring your SSD’s health. Percentage Used represents the official NVMe wear meter, increasing upward from zero, or you will see its opposite counterpart, the “health” score. Total bytes written only has relevance when compared to the rated TBW value listed in the drive’s specifications. In other words, the vast majority of SSDs are replaced before the available TBW budget is utilised. Available Spare represents the amount of replacement blocks that the controller uses as flash blocks fail due to usage, and the controller would like to have this value at or close to 100% available. When you notice a decline in available spare, it is the best indication of SSD wear you’ll receive. Media and Data Integrity Errors has just one acceptable reading: zero errors, regardless of what percentage you’ve seen listed in terms of headline numbers.
As well as being an inconvenient truth confirmed by every recovery bench, there is little doubt that most of the SSDs that arrive dead are not worn out. Instead, they are typically found to have perfect wear statistics and a controller that simply quit responding, resulting from either a firmware crash or an electronics fault that transitions the drive from a perfectly good condition to complete failure over two boot cycles. It should go without saying that since the component that crashed was responsible for writing the health report, none of these gauges had any indication of the impending failure. This is merely a design principle and not bad luck: an SSD’s health check reports wear, while a backup is the only method to protect against a hardware failure.
Your consolation lies in the fact that a failed controller typically leaves the actual NAND flash and your files intact. They were left physically within a drive which will refuse to communicate. This is why we have methods such as firmware level repairs and reading the chips directly, to rebuild data that the controller can no longer provide: this forms the basis for much of our SSD not detected work, and it is why a missing SSD is considered a recoverable issue versus a lost item.
Numbers clean: write down today’s bytes-written number, schedule another check six months later, and that is all of the maintenance needed. Spare dropping, or integrity errors off zero: the flash is giving you unusual honesty at this point; copy all of your data off immediately while it will read for you, replace it, and refrain from using stress test methods to double-check because each complete scan uses up even more of the limited resources. Stutters, freezes, a drive that disappears and comes back: these are controller symptoms, and power-cycling will only make them worse; back up everything in one single move. Already gone: stop reflashing, stop changing cables around hoping to find something, and get a bench-read done; our SSD & NVMe recovery service starts with a free diagnostic followed by a written quote. Internal SSD note: files that have been deleted will have their blocks cleared by TRIM within moments; therefore, if you delete an SSD file accidentally, a back-up was likely the answer anyway.
No. Health measures both the physical wear on your flash and the overall health of your controller. Formatting an SSD places a new file system atop the existing structure; if anything, this process consumes slightly more of your remaining write budget. If formatting seems to fix a malfunctioning SSD, then the underlying problem was likely a logical one, and one you would want to recognise as a potential warning sign rather than a solution.
No. Defragmentation is something that you should never do on a solid-state drive (SSD) because defragmentation uses your SSD’s write cycles, which are limited and can cause damage to your SSD over time. Defragmentation also causes unnecessary overhead to move data around inside your SSD. As long as you have Windows installed, it knows this too and instead of running a defragmenter on your SSD, it runs TRIM. Let it run.
Don’t wait for some timer to go off; instead treat this as a point at which your drive has reached its end of service. When a drive reaches exhausted endurance, it will tend to degrade into a read-only state with increasing error counts; therefore get anything you can from it immediately while reads are possible. And do not continue running an exhausted drive until “it finally fails” because that last failure may be far from being gentle.
In our laboratory we offer recovery options at the controller, firmware and chip levels on SSDs. As always, this includes a complimentary 48-hour diagnostic as well as a written price quote prior to working on your unit. In most cases you only pay if we recover your data.