Why SSDs Fail Differently
Flash storage dies quietly, often suddenly, and almost always without the mechanical warnings that HDDs give you.
01The Physics of Flash
A spinning hard drive is a mechanical system. It degrades in ways you can hear — a bearing softens, a head drifts, the read errors climb. SMART data reflects that accumulation: reallocated sectors tick upward over weeks, and you have time to act. A flash-based SSD is an electronic system with no moving parts, no clicks, and a fundamentally different failure mechanism. It does not wear in; it wears out, one program-erase cycle at a time.
Every NAND cell can only be written so many times before the insulating oxide layer that traps electrons becomes permanently leaky. The cell stops holding a reliable charge, and its stored value becomes ambiguous. In SLC flash, manufacturers typically rate cells for tens of thousands of program-erase cycles. In the TLC and QLC chips that now dominate consumer and small-business storage, that figure is far lower — sometimes a few hundred cycles per cell under real workloads. The controller compensates through wear levelling, distributing writes across the entire flash pool so no single block is hammered while others stay fresh. But wear levelling cannot create cycles that do not exist. It only spreads the budget evenly.
Write amplification compounds the problem. Whenever you write a small file, the controller cannot simply overwrite an old page — NAND can only be erased at the block level, which is far larger than a page. So the controller reads the surrounding valid data, erases the entire block, writes the valid data back, then adds your new data. One logical write becomes multiple physical writes. The ratio between physical writes and the data you actually intended to write is the write amplification factor, and on a fragmented or nearly full drive it can climb well above 1:1. A drive carrying heavy small-file workloads — a database write-ahead log, a browser cache hammered by dozens of tabs — may consume its rated endurance significantly faster than the headline terabytes-written figure suggests.
02How SSDs Actually Die
The failure mode that catches people out is the read-only transition. When a consumer SSD's firmware detects that the drive is approaching its endurance limit, it may flip the entire device into a read-only state to preserve what remains on it. The intention is protective: your data survives, locked in place, while the drive refuses further writes. In practice, this is still an emergency. The read-only state is not guaranteed — some drives simply stop responding, and some lose their mapping tables, which leaves the data physically present but logically inaccessible.
Power loss at the wrong moment is the other critical SSD vulnerability. HDDs are vulnerable to sudden power loss too, but an SSD caught mid-block-erase-and-rewrite can corrupt not just the block being written, but the mapping tables that track where every logical address lives on the physical flash. Without an intact mapping table, the drive may present as unrecognised, empty, or entirely absent. Some enterprise SSDs carry capacitors specifically to complete in-flight operations through a brief power failure; consumer drives almost never do.
Retention is a third issue that has no equivalent in mechanical drives. Flash cells that are not regularly refreshed gradually lose their charge, especially at elevated temperatures. Data stored on a powered-off SSD does not sit perfectly still. Industry specifications acknowledge that unpowered retention decreases significantly as cells age and temperature rises. An SSD used as a cold-storage backup and left in a drawer for several years may be less reliable than one that has been running continuously. The received wisdom — that SSDs are ideal for archival storage — is largely false for anything beyond a couple of years.
| Term | What it is | What it means for you |
|---|---|---|
| Wear levelling | The controller spreading writes evenly across all blocks | You cannot wear out one file; you wear out the drive |
| Write amplification | One logical write causing several physical ones | Small random writes spend the budget faster than large ones |
| Endurance rating (TBW / DWPD) | The total writes the manufacturer will stand behind | A budget to plan against, not a cliff edge |
| Over-provisioning | Spare blocks the drive keeps back for itself | Why a full SSD gets slower and wears faster |
| Read-only fallback | The state a drive enters when it will no longer accept writes | The last cooperative state: read everything off now |
Without an intact mapping table, the drive may present as unrecognised, empty, or entirely absent.
03What This Means for Monitoring and Recovery
SMART data for SSDs does exist, and a handful of attributes genuinely matter: total bytes written, percentage of rated lifetime used, and spare block count. These are worth watching. But SMART is not tuned for flash failure modes the way it evolved to describe mechanical degradation. An SSD can report zero reallocated sectors and a healthy percentage-of-lifetime figure right up until the moment it transitions to read-only or stops responding. The warning window is far shorter than with a spinning drive.
The practical response is not complicated, but it is unforgiving. If an SSD shows any sign of trouble — sluggish reads, mounting errors, an unexpected read-only mount — stop writing to it and take a full image immediately. Do not run a file-system repair on the only copy. Do not power-cycle it repeatedly hoping it will come back. Flash in distress can transition from degraded to unresponsive in a single write cycle, and every attempt that touches the drive uses a cycle. If imaging fails and the data is important, this is a job for a professional recovery service equipped to work on the controller and flash chips directly.