There is an old joke among storage engineers: open an SSD and look closely, and you can almost hear the controller struggling not to wear out your cells. And it is no joke. Behind that 2.5-inch enclosure there is a meticulous bookkeeping system deciding, cell by cell, where every bit goes so the drive lasts you years instead of weeks.
To understand it you have to start with the raw material: NAND flash memory. Each cell holds an electrical charge that represents one or several bits, and that “storing” is not free. Every time you write to and erase a cell, the oxide insulating its transistor degrades a little. At some point the cell can no longer hold charge reliably and dies. That is the famous program/erase (P/E) cycle, and it is the currency of your disk’s entire endurance.
The asymmetry that complicates everything
NAND has a quirk that conditions everything else: data is read and written in pages (typically 4 KB), but it can only be erased in whole blocks (several MB). You cannot “overwrite” data in place the way a magnetic disk does. If you want to change half a page, the controller must copy the entire block elsewhere, mark the old one for erasure and, once enough “dirty” blocks pile up, erase them in bulk.
That erase operation is precisely what wears out the cells. The more times you have to relocate data for a simple change, the more P/E cycles you burn for the same amount of useful writing. That excess is called write amplification: your disk ends up physically writing more than you logically ask it to.
Wear leveling: so no cell burns out alone
If the controller always wrote to the same first cells, those would die quickly while the rest of the chip stayed like new. That is why wear leveling exists. The controller maintains a logical-to-physical map (the flash translation layer, FTL) that decides where to place each write so that P/E cycles are spread evenly across the whole chip.
There are two flavors. Dynamic leveling picks cold cells for each new write. Static leveling goes further: it relocates even data that has been sitting still for a long time, just to “break in” zones that have barely been used. That constant rearrangement of blocks is one reason your SSD writes more than you imagine even when you barely touch the disk.
The invisible helpers: over-provisioning and TRIM
So the controller has room to maneuver, the manufacturer reserves part of the chip you never even see: over-provisioning (typically 7-28 % of the capacity). Those extra cells act as a cushion: they allow data to be relocated without waiting to erase dirty blocks and reduce write amplification. That is why two otherwise identical SSDs, one with more over-provisioning, can last noticeably longer.
The other key piece is the TRIM command. When you delete a file, the operating system tells the SSD which blocks no longer hold useful data via TRIM (deallocate in NVMe). Without that signal, the controller would not know which pages are dead and would keep copying them on every relocation. With TRIM, the drive’s garbage collector can skip those pages and erase whole blocks with far less writing.
When the bit degrades: error correction
Over time, a cell can be “read” differently from how it was written, especially in multi-level cells. To catch that, every page carries redundancy (ECC codes, and in modern NAND, LDPC algorithms), and the controller rewrites damaged data before the error can spread. That proactive correction is part of the internal maintenance that, again, shows up as extra writes you never see.
Degradation also shows up in the cell type. SLC cells hold one bit per cell and survive tens of thousands of cycles; TLC hold three bits and last in the thousands; QLC hold four bits and last in the hundreds. That is why high-end SSDs use simulated SLC caches for fast writes, and why endurance is advertised in TBW (terabytes written guaranteed) or DWPD (drive writes per day).
Write less to live longer
All this machinery exists because NAND has a finite budget of P/E cycles and no battery to replenish it. The practical takeaway is simple: the less you write and the less you fragment the controller’s work, the longer your drive lives. Keeping TRIM on, not filling the drive to 100 % (so effective over-provisioning remains), and not using the SSD as a dump for downloads are gestures the controller quietly appreciates.
So the next time you see the drive light blinking without having saved anything, now you know: it is not failing. It is putting in overtime so that, a few years from now, it is still there.
