Remember when booting a computer meant making tea? You pressed the power button, the drive whirred and clicked, and Windows slowly crawled to life. Then came the solid-state drive (SSD) — and everything changed. Laptops now boot in seconds, games load in a blink, and nothing inside moves at all.
So what is inside that little stick of silicon, and why is it so much faster than the hard drives it replaced? The answer involves trapped electrons, a rule that says you can never overwrite anything, and a housekeeping routine named after a gardening tool.
How a hard drive stored data (the old way)
A traditional hard disk drive (HDD) stores data magnetically on spinning platters, like a vinyl record. A tiny read/write head floats above the platter and physically moves to the right spot to read or write each chunk of data.
That physical movement is the whole problem:
- Reading random files scattered across the platter means the head must dance back and forth — each seek takes several milliseconds
- The platters spin at 5,400 or 7,200 RPM and are easily damaged by a drop
- Even the fastest HDD tops out around 80–160 MB/s in real-world transfers
An SSD throws all of that away. There are no moving parts — no platters, no heads, nothing that spins or clicks. Data lives in NAND flash memory chips, and reading any byte takes microseconds, not milliseconds.
NAND flash: storing data as trapped electrons
At the heart of an SSD sit NAND flash cells — microscopic transistors that store data by trapping electrons inside them. Think of each cell as a tiny cup: an empty cup means one value, a cup holding electrons means another.
The cells are organised in a strict hierarchy:
- Cells store individual bits of data as charge levels
- Pages group many cells together — this is the smallest unit you can write to
- Blocks group many pages together — and this is the smallest unit you can erase
That last point creates the SSD's most important rule: you cannot overwrite data in place. To change a single page, the drive must erase the entire block it sits in first — a process called erase-before-write. Erasing wears the cell down slightly, and each cell survives only a limited number of program/erase (P/E) cycles before it wears out.
Because the cells are built from the same transistor technology as processors and memory, they benefit from the same chip-making advances covered in our semiconductors explainer.
The erase-before-write problem (and the TRIM solution)
When you delete a file, your operating system normally just marks the space as free in its own records. On a hard drive that is fine — the next write simply overwrites the old sectors. On an SSD, it creates a silent mess:
- The SSD's controller still believes the old data is valid
- When new data arrives for those "free" pages, the drive must first read the whole block, erase it, then write the new data — a slow read-modify-erase-write dance
- Over time, the drive spends more and more effort shuffling stale data around, getting slower and wearing out faster
TRIM is the command that fixes this. When you delete a file, the OS sends TRIM to the SSD (on NVMe drives the equivalent is called Deallocate), telling the controller exactly which pages are no longer needed. The controller then quietly erases them during idle time — like a cleaning crew tidying empty rooms before the next guest arrives.
This is also why you should never defragment an SSD. Defragmentation rewrites the entire drive, burning through precious P/E cycles for zero speed gain — flash has no seek time, so file placement barely matters. When Windows "optimises" an SSD, it actually just runs a batch retrim, not a defrag.
Wear levelling: how SSDs stay alive for years
Since every erase wears cells down, an SSD's controller — a small dedicated processor on the drive — works hard to spread the wear evenly. This technique is called wear levelling: the controller tracks how many times each block has been erased and deliberately rotates writes toward the least-used blocks, so no single area dies early.
Not all flash is equally durable, though. Manufacturers squeeze more bits into each cell to make drives cheaper and bigger, and every extra bit costs some endurance:
| Flash type | Bits per cell | Endurance | Typical use |
|---|---|---|---|
| SLC (Single-Level Cell) | 1 | Highest — longest lifespan | Enterprise servers, industrial use |
| MLC (Multi-Level Cell) | 2 | High | Older consumer drives, some prosumer models |
| TLC (Triple-Level Cell) | 3 | Good — lasts many years of normal use | Most consumer SSDs today |
| QLC (Quad-Level Cell) | 4 | Lower, but fine for light workloads | Budget and high-capacity drives |
The more bits packed into a cell, the finer the charge distinctions the controller must make — QLC cells distinguish between 16 different charge levels, which is why they wear out faster. Still, for everyday use a TLC or even QLC drive lasts many years; endurance is measured in terabytes written (TBW), and a typical 1 TB TLC drive is rated for several hundred TBW.
SATA vs NVMe: why the plug matters
Early SSDs used the SATA interface — the same cable designed for hard drives back in 2003 — which caps out at about 550 MB/s. That was already a huge leap over HDDs, but it left enormous flash performance on the table.
NVMe (Non-Volatile Memory Express) drives connect through PCIe, the same high-speed lanes used by graphics cards. The difference is dramatic:
- SATA III: ~550 MB/s read — fast, but the interface is the ceiling
- NVMe PCIe 3.0: ~3,500 MB/s — over six times faster
- NVMe PCIe 4.0: ~7,000 MB/s — roughly thirteen times faster than SATA
- NVMe PCIe 5.0: ~12,000 MB/s and beyond — for enthusiasts and professionals
Part of the secret is parallelism. The old SATA protocol allows a single queue of just 32 commands at a time. NVMe supports up to 65,535 queues, each holding 65,536 commands. That is why a modern laptop feels instant: booting an OS means fetching thousands of tiny files, and NVMe handles them all in parallel.
One caution: M.2 is only a physical shape, not a speed guarantee — an M.2 slot can hold a SATA or an NVMe drive, so check which one you are buying.
Takeaway checklist
- SSDs have no moving parts — data lives in NAND flash cells as trapped electrons, so any read takes microseconds, not milliseconds
- Erase-before-write is the golden rule — flash blocks must be erased before reuse, which is why TRIM (or NVMe's Deallocate) matters so much for speed and lifespan
- Never defragment an SSD — it wastes write cycles with no benefit; let the OS run TRIM/retrim instead
- Wear levelling spreads writes across cells; TLC and QLC drives easily last years of normal use
- NVMe over PCIe is the real upgrade — up to ~7,000 MB/s on PCIe 4.0 vs ~550 MB/s on SATA, thanks to massive command parallelism
- When shopping: check the interface (NVMe vs SATA), the NAND type (TLC for the best balance), and the TBW endurance rating — not just the headline speed
