How Does an SSD Work? Flash, Controllers, and Wear Leveling Explained

Open up almost any laptop bought in the last few years and you will not find a spinning disk. You will find a solid state drive, a slab of silicon with no moving parts that boots your machine in seconds. So how does an SSD work, and why did it make the mechanical hard drive look like a rotary phone almost overnight? The short version is that an SSD stores your data in flash memory cells and uses a small dedicated computer, called a controller, to decide where every byte lives. The longer version is far more interesting, and it explains everything from why SSDs get slower when full to why they quietly wear out.

Table of Contents

• No Moving Parts: The Big Idea
• NAND Flash: Where the Bits Actually Live
• SLC, MLC, TLC, QLC: Cramming More Bits Per Cell
• The Controller: The Tiny Brain Running the Show
• Wear Leveling, Garbage Collection, and TRIM
• Why SSDs Slow Down and Eventually Die
• SATA vs NVMe: The Highway Matters
• Frequently Asked Questions

No Moving Parts: The Big Idea

A traditional hard disk drive (HDD) is a tiny record player. A magnetic platter spins thousands of times per minute while a read head floats nanometers above it, flipping the magnetic orientation of microscopic regions to write ones and zeros. It works, but it is mechanical, which means it is slow to seek, fragile when dropped, and audible when busy.

An SSD throws all of that out. There is no platter, no head, no motor. Data is held in flash memory chips, the same family of memory in your phone and USB sticks. Because there is nothing physical to move, the drive can read any byte anywhere in roughly the same amount of time. That single property, random access without seek time, is why an SSD feels instant and an HDD feels like it is thinking. If you want a sense of how much speed depends on getting data close to the processor quickly, our piece on what a CPU cache is covers the layer that sits even closer than your drive.

NAND Flash: Where the Bits Actually Live

To understand how an SSD works you have to meet the floating gate transistor, the building block of NAND flash. Picture a regular transistor with an extra electrically isolated layer trapped inside it, the floating gate. By forcing electrons onto that gate or pulling them off, the cell holds a charge state even when the power is cut. That trapped charge is your stored bit. No electricity needed to keep it, which is why your data survives when you shut the laptop.

The catch is the way flash is organized. Cells are grouped into pages, typically a few kilobytes each, and pages are grouped into blocks. You can read and write a single page, but you cannot erase a single page. Erasing only happens at the block level, wiping thousands of cells at once. This asymmetry, fine-grained writes but coarse-grained erases, is the source of nearly every quirk an SSD has. Hold onto it, because it explains the slowdowns later.

Why It Is Called NAND

The name comes from the way the memory cells are wired together, in series like a logical NAND gate. The rival arrangement, NOR flash, wires cells in parallel and allows faster random reads but is far more expensive per gigabyte. NAND won the storage war because density and cost matter more than raw read latency for bulk storage. NOR still hangs around in places that need to run code directly from flash, like the boot chip on some embedded devices.

SLC, MLC, TLC, QLC: Cramming More Bits Per Cell

Here is the clever, slightly terrifying trick that made SSDs cheap. A single flash cell does not have to store just one bit. By measuring not whether a charge is present but how much charge is present, a single cell can encode multiple bits at once.

• SLC (Single Level Cell): one bit per cell, two voltage states. Fastest, most durable, most expensive. Reserved for enterprise and caches.
• MLC (Multi Level Cell): two bits per cell, four voltage states. The original consumer sweet spot.
• TLC (Triple Level Cell): three bits per cell, eight voltage states. What most consumer drives use today.
• QLC (Quad Level Cell): four bits per cell, sixteen voltage states. Cheapest per gigabyte, slowest, and wears out fastest.

The pattern is a clean tradeoff. Every extra bit you pack into a cell means the controller has to distinguish between twice as many voltage levels squeezed into the same physical range. The margins get tiny, errors get more likely, and the cell tolerates fewer rewrites before the insulation degrades. That is why a budget QLC drive can feel snappy when empty and then crawl during a big file transfer. It is also why manufacturers stack cells vertically now, in 3D NAND, building skyscrapers of cells hundreds of layers tall to gain density without shrinking each cell further.

The Controller: The Tiny Brain Running the Show

If the flash is the warehouse, the controller is the warehouse manager, and it is the part nobody talks about even though it is the difference between a great SSD and a bad one. The controller is a small multi-core processor with its own firmware. Every read and write your operating system issues goes through it.

Its most important job is maintaining the Flash Translation Layer, or FTL. Your operating system thinks it is writing to fixed logical addresses, the same way it would address an old hard drive. The controller secretly maps those logical addresses to constantly changing physical locations in the flash. When you overwrite a file, the controller does not overwrite the old cells in place. It writes the new data to a fresh page and updates the map, marking the old page as stale. This indirection is what makes everything else possible, and it is why an SSD is really a small computer pretending to be a dumb disk.

The DRAM Cache Question

Better drives include a small chip of DRAM to hold that mapping table for instant lookups. Cheaper drives drop the DRAM to save money and either borrow a slice of your system RAM (a feature called Host Memory Buffer) or look the map up from the flash itself, which is slower. If you ever wonder why two drives with identical capacity differ in price and sustained performance, the presence of a DRAM cache is often the hidden reason.

Wear Leveling, Garbage Collection, and TRIM

Remember that flash cells wear out after a finite number of erase cycles. If the controller kept writing to the same physical block, that block would die while the rest of the drive sat fresh. So the controller practices wear leveling, deliberately spreading writes across every block so they all age evenly. A block you have never touched logically might be physically rewritten many times behind your back, purely to keep the wear balanced.

Then there is garbage collection. Because erases happen at the block level, blocks fill up with a messy mix of valid and stale pages over time. Garbage collection reads the still-valid pages out of a mostly-stale block, rewrites them somewhere clean, then erases the whole block to recycle it. This housekeeping happens in the background, and it creates extra writes the user never asked for, an effect called write amplification.

TRIM is the command that keeps this efficient. When you delete a file, the operating system normally just forgets about it without telling the drive, so the SSD still thinks those pages hold valuable data and dutifully copies them during garbage collection. The TRIM command tells the drive which pages the OS no longer cares about, so the controller can skip copying them and erase them sooner. Modern operating systems issue TRIM automatically, which is why an SSD that supports it stays fast over its life.

Why SSDs Slow Down and Eventually Die

Two phenomena explain SSD aging, and both follow directly from the physics above. The first is the full-drive slowdown. When a drive is nearly empty, the controller always has clean blocks ready and can write instantly. When it is nearly full, there are no clean blocks, so every write has to wait for garbage collection to clear space first. This is why keeping ten to fifteen percent free space is the single best thing you can do for sustained speed. Many drives even reserve hidden capacity for exactly this, called over-provisioning.

The second is endurance. Each cell tolerates a limited number of program-erase cycles before its insulation degrades and it can no longer hold charge reliably. Manufacturers rate this as Terabytes Written (TBW), and for typical consumers it is a number you will likely never reach in a decade of normal use. When cells do start failing, the controller retires them and pulls from spare capacity, and only when the spares run out does the drive switch to read-only mode to protect your data. So an SSD does not die suddenly like a dropped HDD. It fades, predictably. This same idea of designed-in failure modes shows up across hardware, much like the deliberate aging in the OLED panels we covered in our dark mode battery piece.

SATA vs NVMe: The Highway Matters

A fast drive is useless if the road to the processor is narrow. Early consumer SSDs used the SATA interface, the same one designed for hard drives, which caps out around 550 megabytes per second. The flash was ready to go faster, but the highway could not carry it. NVMe drives fixed this by connecting straight to the PCIe lanes, the same fast pathway your graphics card uses, reaching several gigabytes per second on modern generations.

For everyday tasks like booting, launching apps, and loading games, the difference between SATA and NVMe is smaller than the marketing suggests, because those workloads are bursty and limited more by latency than raw throughput. NVMe shines when you move enormous files or run workloads that hammer the drive continuously. If you care about how data moves quickly between distant machines rather than inside one, our explainer on how DNS resolution works tackles the networking side of the same speed obsession, and our look at why AI is eating your RAM covers the memory squeeze sitting one tier above your storage. For the tactile side of computing hardware, the guide to mechanical keyboard switches is the input counterpart to all this storage talk.

Frequently Asked Questions

Do SSDs lose data if left unplugged for a long time?

Yes, eventually. The charge trapped in flash cells slowly leaks over months to years without power. Consumer drives are rated to retain data for around a year unpowered when new, and less as they age. For long-term cold archival, an SSD is not the right tool. Plug it in occasionally and the controller refreshes the cells.

Should I defragment an SSD?

No. Defragmentation was about reducing physical head seeks on a spinning disk, and an SSD has no head and no seek penalty. Defragmenting just generates pointless writes that burn through endurance for zero benefit. Modern operating systems know this and disable it automatically for SSDs, running TRIM instead.

Why does my SSD show less capacity than advertised?

Two reasons. Manufacturers count a gigabyte as a billion bytes while your operating system counts it as 1,073,741,824 bytes, which shaves off roughly seven percent on paper. On top of that, drives reserve hidden over-provisioning space for garbage collection and spare cells, which you never see.

Is a more expensive SSD always faster?

Not for everyday use. The biggest real-world differences come from the controller quality, the presence of a DRAM cache, and whether the flash is TLC or QLC. Two drives with the same advertised sequential speed can behave very differently under sustained load, so look past the headline number.

The Bottom Line

An SSD is not just a faster hard drive. It is a small, opinionated computer that hides a swarm of constant background work behind the illusion of a simple disk. Flash cells trap electrons to remember bits, a controller juggles wear leveling and garbage collection so those cells age gracefully, and TRIM keeps the whole dance efficient. Understanding how an SSD works turns its odd behaviors, the full-drive slowdown, the slow death, the capacity gap, from mysteries into things that make perfect sense. Keep some free space, let TRIM do its job, and your drive will outlast the laptop around it.