Solid-State Drive (SSD)

An SSD reads and writes data entirely electronically: there are no platters, no heads, no spindle. Your operating system sends a write request to the SSD’s controller, which decides where to store the data on the NAND flash chips. To read it back, the controller looks up its mapping table and returns the data to the host. Every part of an SSD exists to make those two operations as fast and as durable as possible.¹

NAND flash: where SSDs store data

Data lives on NAND flash: arrays of floating-gate transistors that hold electrical charge to represent bits. Each cell stores 1, 2, 3, 4, or 5 bits depending on the technology, with higher density coming at the cost of speed and write endurance.²

• SLC (1 bit/cell): fastest and most durable, used in enterprise and SLC cache regions.
• MLC (2 bits): the original consumer standard, mostly retired.
• TLC (3 bits): the dominant consumer NAND in 2026, a balance of capacity and endurance.
• QLC (4 bits): cheaper, denser, but slower and shorter-lived. Common on budget high-capacity drives.
• PLC (5 bits): emerging, mostly enterprise read-heavy workloads.

The SSD controller

The controller is the embedded processor that runs the SSD’s firmware. Modern controllers have multiple ARM or RISC-V cores, multiple NAND channels (typically 4 to 10), and dedicated hardware for encryption and error correction. The controller is the brain that turns flash chips into a working drive, and when it fails the drive becomes a bag of NAND with no way to talk to it.³

Flash Translation Layer (FTL)

The FTL is the firmware abstraction that lets an SSD pretend to be a hard drive. Operating systems still address storage as logical blocks (LBAs), but NAND can only be erased in large blocks, not modified in place. The FTL maps logical addresses to physical NAND pages, handles wear leveling, runs garbage collection, and processes the TRIM command. When the FTL is healthy your SSD is fast and durable; when its mapping tables get corrupted the drive can report 0 MB capacity or wrong model strings (the infamous “SATAFIRM S11” Sandforce failure mode).⁴

SSD DRAM cache and SLC cache

Higher-end SSDs include a small DRAM cache chip to hold the FTL mapping table, which dramatically speeds up random reads. DRAM-less SSDs use a portion of host RAM (Host Memory Buffer, HMB) instead. Most consumer drives also reserve part of the TLC or QLC NAND as an SLC cache, writing one bit per cell for burst speed before migrating data to denser storage in the background. When the SLC cache fills, sustained write speeds drop sharply, sometimes to 100 MB/s or less on QLC drives.

SSD endurance: TBW, DWPD, and what they mean for you

Every NAND cell wears out after enough program/erase cycles, so SSDs ship with two endurance ratings standardized by JEDEC.¹ Terabytes Written (TBW) is the total volume of writes a drive is rated to handle within its warranty. A typical 1 TB consumer TLC drive is rated for 600 TBW, meaning the manufacturer guarantees the drive will accept 600 TB of writes before warranty expires. Drive Writes Per Day (DWPD) expresses the same idea differently: it’s how many times you can fill the entire drive every day for the warranty period without exceeding TBW. A 1 TB drive rated for 0.3 DWPD over 5 years can take 300 GB/day on average for 5 years.

For most users these numbers are theoretical. Real-world consumer write workloads average around 10 to 40 GB per day, which means a 600 TBW drive would last 40 to 160 years at that rate. SSDs that fail in the first 5 years almost never fail from wear: data-center studies consistently show controller defects, firmware bugs, and electronic failures arrive long before NAND endurance becomes the bottleneck. The exceptions are heavy creative workloads (video editing scratch disks, database servers, write-heavy logging) where TBW can matter, and those use cases call for enterprise SSDs with much higher endurance ratings (1 to 10 DWPD versus 0.3 typical for consumer).

What endurance metrics actually predict for recovery: a drive nearing its TBW limit will often switch to read-only mode as a final safety measure rather than fail catastrophically. That’s the most recoverable failure mode SSDs have. If your drive’s SMART data shows percentage used above 95% or unusual ECC error counts, copy your data off immediately, before the controller escalates further.

SSDs fail in fundamentally different ways than hard drives, and the failure pattern almost always determines whether recovery is possible at all. Unlike HDDs that warn you with clicking or grinding, SSDs typically fail without warning. Here are the failures that show up most often in real recovery cases.⁵

• Controller failure. The most common SSD failure. The drive disappears from BIOS, shows zero capacity, or stops responding entirely. NAND chips are usually intact but unreachable. Recovery requires lab-level chip-off extraction or controller bypass with PC-3000 SSD.
• Firmware corruption. The FTL mapping tables become unreadable. The drive may show wrong model strings (the classic “SATAFIRM S11” message), report 0 MB capacity, or boot into a vendor service mode. Software cannot help. Lab tools can sometimes rebuild the FTL without chip-off.
• NAND wear-out. After enough program/erase cycles the NAND cells lose their ability to hold charge reliably. Modern controllers respond by locking the drive into read-only mode, which is actually the friendliest failure mode: data is still readable, just not writable. Copy everything off immediately.
• Sudden power loss without PLP. Without power-loss protection capacitors (rare on consumer drives), an unexpected shutdown during a write can corrupt the FTL or leave NAND blocks in an inconsistent state. The drive may need lab recovery to rebuild the mapping tables.
• Read-only lockdown. Some failure modes (failed firmware updates, ECC errors exceeding thresholds, write count exhaustion) put the drive in permanent read-only mode. Treat this as urgent: copy data to a healthy drive before the SSD escalates to no-detection.
• Logical corruption. The drive is healthy but the partition table or file system is damaged. This is the only failure mode where consumer software like R-Studio, Disk Drill, or EaseUS Data Recovery Wizard can reach the data. Even so, any file deleted before TRIM ran is permanently gone.
• Apple T2 / Apple Silicon SSDs. SSDs in 2018+ MacBooks, T2 Macs, and Apple Silicon Macs are soldered to the motherboard and encrypted with a key tied to the secure enclave. If the logic board or T2 chip fails, the data is unrecoverable by any known method, lab or otherwise.⁶

Warning signs your SSD is failing

Because SSDs rarely give physical clues like clicking or grinding, the warning signs are almost all behavioral. Any one of these symptoms in isolation can be a software issue, but two or three together almost always mean the drive is on its way out:

• Frequent file system errors or files that suddenly become corrupted or unreadable. Bad blocks are accumulating faster than the controller can remap them.
• Sudden, dramatic performance drops on operations that used to be fast. The controller is retrying reads multiple times to get a clean copy, or struggling with a degrading FTL mapping table.
• Boot failures or system freezes during startup. Bad blocks have hit the partition table, system files, or boot loader. If you have to retry the boot more than once, treat it as urgent.
• Drive switches to read-only mode. The controller’s last-resort failsafe when it has run out of healthy spare blocks. Files are still readable but no new writes succeed. Copy everything off, in this order of priority: irreplaceable personal data, then everything else.
• SMART warnings from tools like CrystalDiskInfo (Windows) or smartmontools (macOS/Linux). Pay special attention to Percentage Used over 95%, rising Uncorrectable ECC Errors, and any Reallocated Sectors count.
• Drive disappears from BIOS or Disk Management intermittently. A controller heading toward total failure. Usually fatal within days or weeks.
• Wrong capacity reported (zero MB, eight MB, or a model string like “SATAFIRM S11”). Firmware corruption. Software cannot help; this is a lab job.

SSDs come in several physical and electrical formats. The interface (SATA vs NVMe) matters more than the form factor for performance, and both affect recovery: NVMe drives often have proprietary controllers that lab tools support inconsistently.

Form Factor	Interface	Typical Capacity	Common Uses
2.5-inch SATA	SATA III (6 Gb/s)	250 GB – 16 TB	Desktop, laptop, NAS upgrades
M.2 SATA	SATA III (6 Gb/s)	250 GB – 4 TB	Older laptops, low-cost builds
M.2 NVMe	PCIe 3.0 / 4.0 / 5.0	250 GB – 8 TB	Modern laptops, gaming PCs
U.2 / U.3	PCIe (NVMe)	1 TB – 30 TB	Servers, workstations
EDSFF (E1.S, E3.S)	PCIe (NVMe)	4 TB – 122 TB	Data center, AI workloads
Soldered (BGA)	NVMe or proprietary	128 GB – 4 TB	Ultrabooks, MacBooks, Surface

As of 2026, the largest shipping SSDs are 122.88 TB enterprise drives from Solidigm and Samsung in EDSFF form factors, sold for AI training and data-center workloads. Consumer NVMe SSDs top out at 8 TB, and 2.5-inch SATA at 16 TB.⁷ Capacity continues to climb as NAND vendors stack more layers (200+ now standard) and shift from TLC to QLC.

The clearest way to see what an SSD is good at (and where it falls short) is to compare it directly to the hard drive it replaced. The trade-offs are almost mirror images of each other.

SSD vs HDD at a glance

Property	SSD	HDD
Read/write speed	500 MB/s – 14,000 MB/s	80 MB/s – 250 MB/s
Moving parts	None	Spindle, platters, heads
Shock resistance	Excellent	Poor when powered on
Annualized failure rate	0.5–1%	1–2%
Unpowered data retention	1–2 years	5–10 years
Cost per TB (2026)	~$60–$100	~$15–$25
Maximum capacity	122 TB enterprise / 16 TB consumer	24 TB enterprise / 22 TB consumer
Software recovery success	Low (TRIM wipes deleted data)	High (data persists until overwritten)
Mechanical recovery	N/A (no mechanics)	Lab cleanroom service

SSD advantages and drawbacks