Data Recovery · October 19, 2026 · ComputerKing Team

SSD Failure Symptoms: 7 Warning Signs, SMART & Data Recovery

SSD not detected or turned read-only? Learn the 7 warning signs, how to read SMART and TBW, why to avoid chkdsk, and how 3-2-1 backups protect your data.

Key Summary

An SSD can die with almost no warning, and the old hard-drive instincts — listening for noises, tapping it, running chkdsk — do not apply. This article teaches you to recognise the warning signs, read SMART and TBW properly, know what to do (and never do) when a drive disappears, and protect your data with 3-2-1 backups.

---

1. Why SSDs Play by Completely Different Rules from HDDs

Many people have spent years building up instincts for judging whether a hard drive is failing: listen for strange noises, give it a tap, run chkdsk to fix it. Applied to an SSD, those instincts are not just unhelpful — they can turn data that was still recoverable into data that is gone forever.

The reason is simple: an SSD contains no spinning platters, no read/write heads, and no motor. It is NAND flash chips plus a controller. It will never go "click-click", it will never be fixed by a tap, and when it fails it may degrade gradually with warning signs — or it may work perfectly today and simply not appear at boot tomorrow. That is the most frustrating thing about SSDs: HDD degradation is usually gradual, but an SSD can disappear with almost no warning.

So the only reliable strategy for SSDs is twofold: first, learn to recognise the warning signs before failure; second, never wait for a warning sign before backing up. Everything below is based on public sources, with the full list of source links at the end. Any Hong Kong local pricing is stated as "quotation on request" — we do not estimate prices, and we do not invent customer cases. For HDD noises, see what HDD clicking sounds mean.

2. Seven Warning Signs Before an SSD Dies

1. Files suddenly won't open, or get randomly corrupted

Files that opened fine yesterday suddenly report as corrupted or unreadable, or show "unreadable" errors — and it is not one file, but different files failing at random. That is a common early sign of SSD degradation. A single corrupted file can be a software problem; multiple files failing at random points to a problem at the storage layer.

2. Freezes during read/write, boot failures, repeated blue screens

The system suddenly freezes during read/write operations, fails to boot, or repeatedly shows BSODs (blue screens). Many people's first reaction is "Windows is broken again". But if the blue screens are tied to read/write activity, or booting keeps stalling, the SSD belongs on the suspect list — rather than just reinstalling Windows.

3. Noticeably slower speeds

Read/write speeds that are clearly slower than before are worth investigating immediately. One important qualification: "slower" alone does not mean the SSD is failing. Slowness can come from a nearly full drive, garbage collection pressure, thermal throttling, and more. The right approach is to read it together with SMART data and other symptoms, not to judge by feel alone.

4. The drive disappears in BIOS/Windows, or appears intermittently

An SSD that vanishes completely from BIOS or Windows, or appears and disappears intermittently, is a serious alarm. This "drive drop" behaviour is especially worth noting on NVMe SSDs: public reports describe one typical NVMe failure mode as the drive suddenly disappearing from File Explorer, Device Manager, and Disk Management in the middle of a write (for example, the pattern in the 2025 KB5063878 dispute reports, see §8.4; a general NVMe drive drop can also present as a sudden disappearance), with SMART/controller telemetry becoming unreadable; a reboot can sometimes temporarily restore visibility, but does not guarantee that data written during the failure window is intact.

5. The SSD becomes read-only — the most important "last chance" signal

When an SSD detects a serious firmware, NAND, or wear-levelling error, the controller locks the SSD into read-only protection mode: reading allowed, writing blocked. The purpose is to preserve whatever data remains before the drive dies completely. This can be described as the "best of the bad outcomes": the controller actively stops further writes to protect data, and the user can usually still copy files straight off the drive.

The only correct response in read-only mode is: immediately copy the most important files to another drive, and deal with everything else afterwards. Do not reboot, do not attempt formatting or a firmware update (do not run chkdsk — see Section 5 for why). Read-only locking is usually a hardware-level fail-safe; ordinary software methods (DiskPart, formatting, registry changes) cannot reliably clear it, and should not be attempted.

One more point must be stated clearly: read-only does not mean safe. A read-only drive can keep degrading. Back up immediately; do not wait.

6. Wrong capacity shown

If the capacity shown by the SSD in BIOS or the operating system does not match reality — for example a 1TB drive suddenly showing only a few dozen GB — do not treat it as ordinary "lost space" and format it. Possible causes include a damaged partition table (MBR/GPT), a firmware bug, a damaged controller translation table (LBA→NAND mapping), or NAND exhaustion causing the controller to lock. Formatting will not fix these problems, and may trigger TRIM and permanently erase data.

7. Overheating

NVMe drives running at high speed generate noticeable heat. Sustained high temperatures accelerate NAND electron leakage, trigger thermal throttling (slowing down to protect itself), and over the long term can cause the controller to fail early. Different sources give slightly different figures for "dangerous temperature": some say sustained temperatures above 70°C accelerate wear, some recommend keeping everyday operation below 60–65°C, and some say 30–50°C is ideal. Do not present any single figure as an absolute standard here. The point is: sustained high temperatures are harmful, and cooling must be addressed.

3. How to Read SMART: Don't Just Look at the Health Percentage

3.1 The health percentage is not a fuel gauge

The "health percentage" shown by tools such as CrystalDiskInfo often maps directly to TBW (used write endurance) consumption, not to a precise measure of physical lifespan. Different manufacturers and tools interpret and scale SMART attributes differently, so values can differ. Kingston's engineering team confirmed to the media that there is no uniform definition of SSD "health". The same drive can show different health percentages in different tools.

So the practical rule is: don't look only at the percentage; look at the key attributes below — especially whether they are increasing. A useful rule of thumb: "A drive at 50% wear with 0 errors is better than a drive at 75% wear with errors still accumulating." Trends matter more than snapshots: a single bad reading does not mean the drive is about to die; check every few months and watch the trend.

3.2 The most important attributes for SATA SSDs

SATA SSD attribute definitions vary by manufacturer; the table below is a general reference — when in doubt, follow that manufacturer's documentation / tool.

Attribute	Why it matters	Warning sign
Reallocated Sectors Count	Bad cells have been remapped to spare capacity	A steadily rising number = real trouble; a small, long-stable number can be normal NAND behaviour
Current Pending Sector Count	Sectors flagged as unstable, waiting to be reallocated	Any appearance / rise is an alarm
Uncorrectable Sector Count / Uncorrectable Error Count	Read/write failures that even error correction could not save	Any non-zero value deserves attention; steady growth = back up immediately
Wear Leveling Count / Media Wearout Indicator	How much write endurance has been used; the normalised value usually counts down from 100, while the raw value varies by manufacturer (some are cumulative P/E cycle counts, counting up)	Near 0 means the rated endurance is used up (not instant death, but beyond warranty coverage)
Total Host Writes / NAND Writes	Lifetime write volume, comparable against the manufacturer's TBW rating	Read together with TBW, never alone
Available Reserved Space (ID 232, varies by manufacturer) / Endurance Remaining	How many spare blocks remain to replace failed blocks	Falling below the threshold = back up immediately
Unsafe Shutdown Count	How many times power was cut without a clean shutdown; frequent occurrences can damage the SSD mapping table	A high number = review power supply / shutdown habits
UDMA CRC Error Count (ID 199, SATA)	Communication errors between host and SSD	Usually points to cable/connector problems, not NAND wear; reseat or replace the cable, then check whether it keeps growing

3.3 The most important fields for NVMe SSDs (different names from SATA)

NVMe uses a standardised health report format, which is more consistent across manufacturers. Key fields:

Critical Warning: 0x00 is normal; bit 2 (0x04, reliability degraded) or bit 3 (0x08, read-only) set = back up immediately. Any non-zero value is urgent.
Available Spare: falling below the threshold (Critical Warning bit 0) = back up immediately. If remaining spare NAND blocks fall below 10%, plan a replacement.
Percentage Used: percentage of rated life used; use as background reference, do not panic on this number alone. Reaching 100% only means the warranty TBW/DWPD threshold has been reached; the counter can keep climbing to 255 and the drive can keep working. It does not mean "100% = instant death".
Media and Data Integrity Errors: must be 0; any non-zero value, especially one that keeps growing = back up immediately.
Temperature: see the "Overheating" section above.

One common misreading to avoid: Error Information Log Entries are not bad sectors. One documented case showed 25,546 Error Information Log Entries while Media and Data Integrity Errors remained 0; most entries were Invalid Log Page / Invalid Field in Command (harmless management messages between Windows and the controller). Do not treat that count as a bad-sector count.

3.4 Back-up-now trigger list

If any of the following appears, back up first and investigate afterwards:

NVMe Critical Warning any non-zero value (especially bit 0 spare capacity, bit 2 reliability, bit 3 read-only)
Available Spare fallen below threshold
Media and Data Integrity Errors non-zero and growing
Uncorrectable Error count growing on SATA
Reallocated blocks / program fail / erase fail rising together within a short period
A pre-fail attribute's normalised value fallen to or below its threshold (shown in smartctl's WHEN_FAILED column)
The manufacturer's tool or Windows reporting a critical warning for the drive

3.5 Values you should not over-interpret

The following do not necessarily indicate failure; they are listed here to avoid scaring readers:

Huge raw values for error-rate attributes (e.g. Raw Read Error Rate, Hardware ECC Recovered): often a packed field or a count of corrected errors.
A small number of reallocated blocks unchanged for months: normal NAND behaviour being handled by firmware.
Percentage Used ≥100% or wear indicator at its limit, with integrity errors at 0: means the rated endurance has been exceeded, not that the drive will die instantly.
Tens of Unsafe Shutdowns (especially on laptops): not necessarily a sign of failure.
Non-zero UDMA CRC Error Count: usually a cable/connector problem.

3.6 The limits of SMART

SMART is good at catching wear-related degradation, but it cannot predict a sudden controller failure, and it will not show a firmware bug that makes an SSD instantly unreadable. Windows Settings does not show full SMART values; but for NVMe SSDs, Windows does pop up notifications for three key warnings (low spare capacity, degraded reliability, read-only) — see §3.7. Third-party tools are needed to see full SMART values. SMART is a monitoring tool, not a crystal ball.

3.7 How to check SMART: tools in practice

CrystalDiskInfo (Windows, free and open source): reads SMART data, temperature, and health status, and warns before potential failure. Practical tip: go to Function > Advanced Features > Raw Values and switch to DEC (decimal) so the numbers are readable; the default HEX is hard to read. Health status: Good (blue) / Caution (yellow) / Bad (red). It supports NVMe, but the attribute names differ from SATA.

smartmontools / smartctl (Windows / Linux / macOS, free and open source): commonly used commands (always verify the device name before use; do not copy blindly):

`smartctl --scan-open` (Windows PowerShell, administrator): find the device name (usually `/dev/pd1` or `/dev/nvme0`)
`smartctl -a /dev/nvme0n1` (Linux NVMe) / `smartctl -a /dev/pd1` (SATA) / `smartctl -a /dev/nvme0` (NVMe): full report. Use the device name from `--scan-open` output; do not copy blindly.
`smartctl -H /dev/sda`: quick health verdict (PASSED / FAILED)
`smartctl -A /dev/sda`: detailed SMART attribute table
`smartctl -l error /dev/sda`: error log; NVMe also has a hidden NVMe Error Information log that Windows does not display — smartctl is needed to retrieve it
`smartctl -t short /dev/sda`: short self-test (about 2 minutes); `smartctl -t long /dev/sda`: full self-test (about 10–60 minutes). Note: a self-test puts load on the drive; if it already shows errors, is unstable, or the data is not backed up, do not run a self-test — back up / clone first.
Linux installation: `sudo apt update && sudo apt install smartmontools` (Ubuntu/Debian); `sudo dnf install smartmontools` (Fedora); `sudo pacman -S smartmontools` (Arch)

Note: if an NVMe SSD sits in a USB enclosure or behind some RAID layers, smartctl may not reach the NVMe admin commands; the drive needs to be plugged directly into an M.2 slot for the full log.

GSmartControl (Windows / Linux, free and open source): a graphical front end for smartctl — SMART attributes and self-test options without typing commands.

Manufacturer tools: Samsung Magician, WD/SanDisk Dashboard, Crucial Storage Executive, Kingston SSD Manager, etc., can show proprietary attributes and update firmware. Generic third-party tools can mislabel attributes or apply the wrong thresholds, so when in doubt, cross-check with the manufacturer's tool.

macOS built-in check: Apple menu → System Information → NVMExpress (Apple Silicon) / SATA (Intel Mac) → check SMART Status: Verified = normal; Failing / Fatal = trouble starting. If you cannot find that field, use the manufacturer's tool or smartctl.

Windows built-in check (PowerShell): PowerShell (administrator) → `Get-CimInstance Win32_DiskDrive | Select-Object DeviceID, Status`: OK = normal; Pred Fail = may fail at any time. Older Windows versions had wmic; current Windows 11 (24H2 / 25H2) has removed wmic — use PowerShell instead. In addition, Windows built-in monitoring only covers NVMe SSDs (not SATA SSDs or HDDs); its three key warnings are: spare capacity is low, reliability is degraded, and drive is set to read-only. Microsoft advises backing up data immediately when a notification appears.

4. SSD Suddenly Not Detected: Common Causes and First Response

4.1 Common causes

Cable / slot problems: internal break in a SATA cable (invisible from outside), M.2 not fully seated, a damaged motherboard SATA port, USB enclosure / adapter problems.
BIOS / UEFI setting problems: SATA mode (AHCI / RAID / IDE) mismatch, individual SATA ports disabled by default, outdated BIOS firmware. Secure Boot restrictions belong to the "drive detected but system won't boot" scenario — they do not make an SSD disappear from BIOS, so do not change Secure Boot just to "detect the drive".
Controller / firmware failure: damaged SSD controller or firmware, so the SSD is not detected at all.
Power / power-loss damage: a sudden power cut can damage the SSD's internal mapping table (FTL mapping table), leaving the controller unable to operate.

A practical distinction (not an absolute rule): nothing detected in BIOS/UEFI = more likely a controller / power / physical-layer problem; detected by the system but partition shows RAW / read-write errors = more likely a logical / file-system / mapping-table problem.

4.2 Safe initial checks (no writes, no damage)

1. Power off completely, disconnect power, and reseat the SSD (replug both SATA cables; for M.2, remove and reseat firmly, refit the retaining screw).

2. Try another SATA cable, another motherboard port, another M.2 slot, or another computer — to isolate whether the problem is the cable / slot / motherboard or the SSD itself.

3. Enter BIOS/UEFI and check whether it is detected; note the existing settings, and do not randomly change storage mode, boot mode, or Secure Boot.

4. If BIOS sees it but Windows does not: check Disk Management; a brand-new blank SSD needs initialisation and a drive letter first. But — if the drive contains data, absolutely do not initialise / format it.

5. If the SSD is overheating or repeatedly disconnecting, stop immediately.

4.3 Things you must absolutely not do (when data needs rescuing)

Do not repeatedly power on / reboot: each failed handshake / power cycle can worsen mapping-table damage and make professional recovery more complex.
Do not run chkdsk (see Section 5 for why).
Do not initialise / format: even if Windows prompts you to initialise, do not do it on a drive with data; formatting can trigger TRIM and permanently erase NAND data within seconds.
Do not run recovery software / install software on the failing drive: it writes to the failing drive and overwrites the files you want to rescue.
Do not change BIOS settings as a "generic fix": changing storage mode / boot mode / Secure Boot can make things worse; check your own computer model's documentation before changing anything.
Do not format / rewrite / update firmware (read-only drive): see Section 2, point 5.
Do not put it in the fridge / hit it: see Section 12, myth-busting.

4.4 When to stop and seek professional help

If the SSD is not detected anywhere and the data matters, stop and seek professional SSD data recovery; do not keep trying. Detected does not mean the data is healthy; data inside the NAND may not be directly readable (for example, if encrypted). On pricing: there is no public Hong Kong local pricing source, so this is quotation on request. Before engaging professional recovery, ask about the assessment fee, whether the quote is result-based, and how a failed recovery is charged. One more important point: keep the order of warranty claim and data recovery separate — if the data matters, do data recovery first, then handle the warranty claim, otherwise the data will disappear together with the replaced drive. MacBook users, also see what to do first after MacBook liquid damage.

5. Why You Should Not Run chkdsk on a Failing SSD

chkdsk is a file-system consistency tool, designed for healthy drives after an abnormal shutdown. It is not a data recovery tool, and it certainly cannot fix hardware failure.

For SSDs, the danger is especially large, for three reasons:

1. chkdsk marks clusters as free → Windows issues TRIM → the SSD controller erases the corresponding NAND pages during garbage collection. Once erased, the data is permanently gone; no lab, no chip-off procedure can bring it back.

2. chkdsk rewrites the file system: unreadable MFT records get cut off, files become nameless .chk fragments placed in a hidden FOUND.000 folder; original filenames, paths, timestamps, and permissions get overwritten, making later recovery harder.

3. chkdsk /r reads every sector of the whole drive, putting extra load on an already degraded drive; /r can also rewrite the file system and, after freeing clusters, trigger TRIM. TRIM is issued by the NTFS driver after clusters are freed, not directly by the chkdsk program itself.

The only situations where chkdsk is appropriate: you already have a complete backup; or the problem is only a minor logical issue caused by an abnormal shutdown; or you only want a health overview of the drive (only the parameter-free read-only `chkdsk`; never `/f` / `/r`). For a failing drive whose data you need, do not run it. Some sources describe chkdsk /f as "safe and useful" for SSD file-system logical repair, but the same sources also warn to use /r with care and to rescue data first on a failing drive. So this article's position is: if data needs rescuing, do not run it; consider it only with a backup in place and a purely logical problem. Do not present chkdsk as the first-choice repair.

6. The Relationship Between TRIM and Data Recovery

6.1 What TRIM is

TRIM is the command by which the operating system tells the SSD which logical blocks are no longer needed; NVMe has an equivalent deallocation operation. The controller can then stop returning the old content and erase the relevant NAND pages during garbage collection. The command and the actual erasure are two separate events — TRIM having been issued does not mean everything was erased that very second.

6.2 Why deleted files are so hard to rescue once TRIM is on

When you delete a file on an HDD, only the "pointer" is deleted; the data itself stays on the disk until overwritten by new data, so recovery software has a chance to find it. When you delete a file on an SSD, TRIM marks the relevant blocks as free, and the controller actively erases / zeroes them in the background, making the recovery window very short. An SSD must erase a block before it can write new data to it; TRIM exists so the SSD can prepare clean blocks in advance and keep write speeds up. The cost is that TRIMmed data becomes unrecoverable — a side effect of how TRIM works, not its design purpose.

6.3 "Deleted" and "drive broken" are two opposite problems

This is a point this article must make crystal clear, because many people confuse the two:

Deleted file (SSD itself healthy, TRIM active): recovery is close to impossible, because the SSD physically erases the flash cells soon after deletion.
SSD failure (not detected, disappears from BIOS, becomes read-only, dies mid-write): TRIM is irrelevant; the data is often still inside the NAND, waiting for chip-level recovery.

Do not throw away a dead drive because you "heard SSD deleted files can't be recovered"; and do not pay for a hopeless scan for a "deleted file". The two situations are opposite problems with opposite odds.

6.4 Immediate response after accidental deletion

Power off and disconnect immediately: TRIM erasure happens during idle time; a powered-off SSD cannot keep "taking out the rubbish". The chance is still not high, but it is the only chance there is.
Do not keep using the drive: even background operating-system writes can trigger TRIM or overwrite the blocks you want to rescue.
Do not install recovery software on the same drive, and do not save rescued files back to the same drive.

6.5 Exception: TRIM does not always reach the SSD

For an external SSD connected via USB Type-A / Type-C, depending on the operating system, enclosure, controller, and connection method, the TRIM / UNMAP command may not reach the SSD, so deleted files may survive longer and recovery chances may be higher. This is a "possibility", not a guarantee; do not read it as "deleted files on an external SSD can always be recovered".

6.6 Formatting and TRIM

Even a quick format can permanently delete files on an SSD because of TRIM, since TRIM can zero the NAND blocks within seconds. Never format an SSD that has files to rescue.

7. NVMe vs SATA: How Failure Looks Different

Comparison	NVMe SSD	SATA SSD
Interface / protocol	PCIe + NVMe protocol, throughput far above SATA	SATA / AHCI ecosystem, SATA III nominal link speed 6 Gb/s, actual throughput affected by protocol overhead
Health report format	NVMe SMART / Health Information Log defined by the NVM Express specification, more consistent across manufacturers	Older ATA SMART framework, attribute IDs and meanings vary by manufacturer (Samsung, Crucial, Western Digital, etc.); check SMART against that manufacturer's documentation
Drive-drop behaviour	More often tied to controller hang, HMB (Host Memory Buffer) / host behaviour interaction, and heat; disappearing mid-write is one typical symptom	Also disappears / not detected, but causes are more often cable, slot, or power
Common causes of slowdown	Thermal throttling, garbage collection pressure, nearly full drive; consumer M.2 NVMe active power is typically around 5–10 W; Gen4/Gen5 drives still run noticeably hot under heavy load and need a heatsink / airflow (15–25 W / peaks around 40 W are enterprise U.2/EDSFF figures, not consumer M.2)	Lower power (SATA SSDs around 3–7 W), generally milder temperatures; slowdown more often tied to full drive, ageing, TRIM/garbage collection — but do not write "SATA never overheats"
Read-only warning mechanism	NVMe specification has explicit Critical Warning bits; Windows built-in monitoring also only targets NVMe SSDs	Also has read-only / write-protect states, but no unified NVMe-style Critical Warning bits; behaviour and diagnosis vary by manufacturer

One more enterprise-vs-consumer reminder: a consumer SSD placed in a server/array under sustained writes may not fail as an instant dead drive; instead it gradually slows down, with intermittent latency spikes (looking like a network or application problem), and only much later is the drive identified as the cause.

8. Public Firmware Cases: Some Admitted, Some Disputed — Both Sides Must Be Told

Firmware problems can break an otherwise healthy SSD. All cases below have public sources. Editorial red line: do not write "community reports" as "manufacturer admitted a recall", and do not write "a firmware update has been confirmed" as "all models are affected". What is admitted and what is disputed is marked separately below.

8.1 Samsung 990 Pro / 980 Pro rapid health drop (2023) — firmware fix confirmed, but damage already done is not reversible

From December 2022, users reported on forums that the 990 Pro's health dropped by more than 10% after only a few days/weeks of use, with write volumes far below the rated value. The Samsung 990 Pro 2TB model is rated at 1,200 TBW — in theory, writing 12 TB should cost 1% of health (1TB / 4TB models are rated at 600 / 2,400 TBW; figures here assume the 2TB model) — yet complainants had lost several percentage points after writing far less than 12 TB. Multiple health-monitoring tools produced the same results, pointing to a hardware or firmware problem.

Samsung released firmware in succession: 5B2QGXA7 for the 980 Pro, 1B2QJXD7 for the 990 Pro. A Samsung Europe community forum moderator confirmed the firmware caused "anomalies". For the 980 Pro, system integrator Puget Systems reported an "unusually high failure rate" in the field for the 2TB model and worked with Samsung for months; Samsung also released a "critical" firmware update for the 980 Pro.

There are two key limitations: first, the firmware fix only stops further degradation; it does not reverse degradation that has already happened — health values that have already dropped will not bounce back. Second, when updating, if the 990 Pro is the boot drive, you must fully power off and disconnect (not merely reboot), otherwise the drive may become unresponsive after the new firmware is applied, causing a blue screen.

Lesson: back up before updating firmware; a firmware update is not a cure-all — do not say "once updated, everything is fine".

8.2 SanDisk Extreme portable SSD data loss (2023) — official line says fixed, but an independent side disputes it

From spring 2023, users reported that some SanDisk Extreme portable SSDs suddenly lost data or became unreadable. Western Digital (then SanDisk's parent company) acknowledged problems with the 4TB SanDisk Extreme and 4TB Extreme Pro and released a firmware update in May 2023. Models named in the class action included the 4TB Extreme, 1TB/2TB/4TB Extreme Pro, and 4TB WD My Passport.

Official position: Western Digital said the firmware update resolved the problem and that hardware was not the cause. The disputing side: data recovery company Attingo said it was a hardware problem — soldering process, solder bubbles, component sizes mismatched with PCB layout, causing solder joints to come loose easily; Rossmann Group reported that some replaced/updated drives still failed; the class action continued into 2024. Independent verification is limited, so the honest wording is "officially stated as fixed, but disputed".

Another recovery vendor described the affected drives as not merely failing, but "actively destroying data that had been successfully written"; the root cause was described as a defect in how the controller handled wear-levelling and atomic write operations. If the drive is still detected, do not write to it again — the next write could trigger another wave of corruption.

Practical tip: if you hold a SanDisk Extreme Portable V2, Extreme PRO Portable V2, or WD My Passport SSD and have not updated the firmware, there is still risk; check the firmware version in SanDisk Dashboard or WD Dashboard and update to 10.03.02.01 or newer, backing up first. If the drive has already failed, an update will not help.

"No Media" failure: if a portable SSD shows "No Media", the underlying hardware (the power-management circuit on the PCB) has failed; the drive cannot even load its firmware, so the firmware updater has nothing to talk to. This kind of case requires hardware repair first, with the goal usually being to get the data back rather than to keep using the drive.

Do not write this portable-SSD incident as "all SanDisk SSDs have problems". The 1TB Extreme was not a model named by Western Digital or the lawsuit.

8.3 Windows 11 24H2 + WD/SanDisk NVMe BSOD (2024–2025) — firmware updates + Microsoft compatibility holds

Under Windows 11 24H2, specific large-capacity WD/SanDisk NVMe models showed repeated BSODs (often storage-driver-related) and system hangs during large file transfers or sustained I/O. Affected SKUs included (not an exhaustive list): WD_BLACK SN770 (specific 2TB variant), WD_BLACK SN770M (specified SKU), WD Blue SN580 (2TB model), WD Blue SN5000 (specified SKU), SanDisk Extreme M.2 NVMe SSD (one 2TB SKU). Exact model numbers must be checked.

Response: Western Digital released firmware updates for the affected models and advised users to update SSD firmware; Microsoft used compatibility holds to block the 24H2 update on affected configurations until the manufacturer released firmware fixes.

Note: the main publicly reported symptom here was "instability (BSOD)", not "total hardware failure"; do not exaggerate it into "a Windows update will kill your drive".

8.4 Windows 11 KB5063878 + NVMe drive-drop dispute (August 2025) — community reproducible vs manufacturer/Microsoft say no causal link

Both sides must be told here; do not settle it one-sidedly.

The incident: Microsoft released Windows 11 24H2 cumulative update KB5063878 (OS Build 26100.4946) on 12 August 2025. In the following days, independent testers and several specialist media outlets reported a reproducible failure pattern: during sustained large sequential writes (often cited as 50 GB or more, with the drive about 60% or more full), some NVMe SSDs stopped responding, disappeared from the system, and SMART/controller telemetry became unreadable; a minority of reports said files written during the failure window were corrupted or lost. In the early samples, Phison-controller drives were over-represented, particularly the PS5012-E12 family.

Manufacturer response: Phison publicly acknowledged awareness of the reports and said it was working with partners to investigate potentially affected controllers; Phison later said it had run more than 2,200 test cycles, about 4,500 hours, without reproducing a systematic failure related to the Windows update, and denied a circulating alleged Phison internal notice (calling the document a forgery).

Microsoft response: After investigation, Microsoft said it found no connection between the August 2025 Windows security update and the types of drive failures reported on social media; telemetry analysis also found no evidence that the update caused a large-scale increase in SSD failures.

Accurate wording: community testing reproduced drive drops under a narrow write pattern; Microsoft and Phison, after extensive testing, said they could not reproduce a systematic failure and found no causal link; the actual losses reported by individual users nevertheless remain real. The practical advice is unchanged: back up, update SSD firmware, roll out system updates in stages, and avoid long large writes on a nearly full drive before firmware + system behaviour have been verified.

8.5 Crucial MX500 (SATA) firmware-related issues

CVE-2024-42642: Micron Crucial MX500 series SSD firmware M3CR046 has a buffer overflow vulnerability, triggerable by the host sending a specially crafted ATA packet to the drive controller; the supplier stated the vulnerability was fully fixed in December 2024, and updated firmware can be obtained via Crucial's official support page. There is also a forum user's recollection that an early Crucial SSD became write-protected after a certain power-on time and needed power cycling to recover an hour of use; Crucial released a firmware fix. This is a community recollection — cite it as "user recollection / forum account", not as an official announcement.

Also state clearly: Crucial discontinued the MX500 (reported December 2024; the reason was product-line adjustment, not a security incident). Do not conflate discontinuation with the vulnerability.

8.6 Vicon server SSD firmware notice (June 2025, enterprise example)

On 11 June 2025, Vicon issued a critical SSD firmware update notice: some Youngsville-family SSDs with firmware older than DL7A could unexpectedly fail and become unresponsive, requiring replacement; the recommended minimum firmware was DL7A, with the latest validated version DL7C. Lesson: firmware updates are not only a consumer matter; enterprise/server environments also have firmware reliability notices, and notices state plainly that "older than a certain version may fail without warning".

8.7 General rules for firmware updates

From the cases above, four rules for firmware updates:

Back up important data before updating firmware.
Only download firmware from the manufacturer's own tools / official support pages (Samsung Magician, WD Dashboard / SanDisk Dashboard, Crucial Storage Executive, etc.); do not use firmware files from unknown sources.
Do not cut power during an update; if it is the boot drive, follow the manufacturer's instructions for a full power-off (the Samsung 990 Pro case proves a simple reboot is not enough).
A firmware update cannot repair an already dead drive; nor will it reverse health degradation that has already occurred.

9. How to Read TBW, DWPD, and Wear Levelling

9.1 TBW (Total Bytes Written / Terabytes Written)

Definition: the total amount of data the manufacturer guarantees the drive can have written to it within its rated lifespan. For example, a 1TB NVMe SSD rated at 600 TBW is guaranteed to withstand 600 TB of host writes, after which warranty no longer covers wear-related failure. The drive can — and often does — keep working beyond its rated TBW, but planning should use the rated value.

Samsung white-paper formula: TBW = (physical capacity × NAND P/E cycles) ÷ (write amplification factor × wear-levelling factor). Physical capacity and P/E cycles are fixed hardware values; write amplification is determined by workload.

How to check: compare the drive's actual total writes (NVMe's Data Units Written; 1 data unit = 512,000 bytes) against that model's published TBW rating — more useful than looking at the health badge alone. Worked example (not a prediction): a drive recording Data Units Written of 98,948,652 means about 50.66 TB of lifetime writes (98,948,652 × 512,000 bytes); if that model is rated at 600 TBW, that is only about 8% used.

Reference ranges (not buying advice): consumer SSDs typically 150–600 TBW; prosumer 600–2,500 TBW; enterprise 3,000–35,000+ TBW. Typical desktop use (about 30–50 GB written per day) converts to about 8–30+ years. Emphasise that this is a "rated write-endurance conversion", not a "guaranteed to last this many years"; actual lifespan is also affected by temperature, controller, firmware, and more.

9.2 DWPD (Drive Writes Per Day)

Definition: TBW normalised into a daily write rate over the warranty period (usually 3 or 5 years). Formula: DWPD = TBW ÷ (capacity × 365 days × warranty years). Example: 1TB drive, 600 TBW, 5-year warranty: 600 ÷ (1 × 365 × 5) = 0.33 DWPD, i.e. guaranteed daily writes of about 33% of capacity, sustained for 5 years. 1 DWPD means the whole drive can be written once per day.

Reference ranges: enterprise SSDs commonly 1–3 DWPD; data-centre class can reach 10+ DWPD. How to read them: TBW speaks to cumulative wear, DWPD speaks to a daily usage ceiling; the two complement each other. Monitoring both helps plan backups and replacement and avoid surprise failure.

9.3 Wear levelling

Principle: the controller spreads writes evenly across all NAND blocks, preventing some blocks from wearing out early through repeated writes, thereby extending usable life. Samsung's white paper notes the Wear Leveling Factor is usually very close to 1.00 — meaning that when wear levelling is done well, its effect on the rated value in the TBW calculation is small; what really makes endurance vary by workload is write amplification. Random small I/O writes usually produce higher write amplification than large sequential I/O writes.

Practical steps (consistent across multiple sources): keep 10–20% free space on the SSD to help wear levelling; enable TRIM (modern operating systems usually enable it automatically); avoid sustained high temperatures (e.g. if temperatures stay above 70°C long-term, address cooling); update firmware regularly; use a good-quality power supply.

What not to write: do not say "wear levelling lets an SSD last forever"; it only delays localised wear, it does not remove the P/E cycle limit.

9.4 What happens near the end of rated life

SMART Percentage Used near 100%: the primary replacement-planning signal.
Media / data integrity errors may increase; write performance may drop under heavy garbage collection; some drives switch to read-only mode to protect remaining data; drives without PLP (Power Loss Protection) face increased power-loss risk during heavy writes.
As NAND wear increases, even before the rated TBW is fully used, data-retention margin and uncorrectable-error risk can worsen.

10. DIY vs Stop-and-Seek-Help: The Dividing Line

Situation	DIY is OK	Must stop and seek professional help
SSD still detected by the system; data loss caused by deletion, formatting, RAW partition, or lost partition; no signs of physical failure	Yes: connect the SSD as a secondary (non-boot) drive to another computer; install recovery software on a different drive; save recovered files to a different physical drive; for an unstable drive, make a sector-by-sector clone first, then scan the clone (with ddrescue, work from another computer / a standalone boot USB, and double-check source and target device names — swapping them erases data; if unsure, stop and seek professional help); record which tool, version, and parameters were used	—
Accidentally deleted file (SSD healthy, TRIM active)	Power off immediately — the only chance there is; do not keep using the drive	Do not pay for a hopeless scan — recovery is close to impossible
Read-only mode	Immediately copy important files to another drive (direct copying usually still works)	Do not reboot, run chkdsk, format, or update firmware
SSD not shown in BIOS, Disk Management, or another computer	—	Stop. Software recovery cannot help
Signs of physical damage: overheating, repeated shutdowns, abnormal electronic noise, unusually hot drive	—	Stop
Drive repeatedly disconnecting, repeated scan failures, or scans freezing the system	—	Stop
Firmware-type symptoms with no manufacturer fix, or each attempt making the drive more unstable	—	Stop
Data extremely important and not worth risking (business records, irreplaceable family photos, source code, research material, legal documents)	—	Stop and seek professional help directly; do not try yourself
Drive still not detected after cross-testing (SATA drive: three different cables / two different ports; M.2 drive: two different M.2 slots; both tried in another computer)	—	Stop: the controller may be burnt; each power-on carries a risk of a short reaching the NAND chips (this sentence reflects the cited source's wording, not a law)
Drive half-alive, each operation uncertain how much longer it will survive	—	If the data really matters, stop early rather than gambling with repeated scans

On RAID, note: RAID is not a backup — how RAID recovery works.

What professional recovery does: firmware-level recovery (specialist tools and expert knowledge for firmware corruption; ordinary software cannot do it); chip-off recovery (removing NAND chips from the SSD, reading data directly from the chips, then rebuilding with algorithms; complex, expensive, only in a professional lab); professional labs generally use firmware repair tools, specialist hardware interfaces, and NAND chip readers. (Clean rooms are for HDD platter work; SSD chip-off does not need one.)

Important limits — do not write "professional recovery always gets it back": modern SSD recovery depends on whether the controller and encryption path are viable; standalone raw-NAND chip-off cannot bypass encryption; a lost controller path can make recovery impossible. If the drive or partition is encrypted and the key is not in hand, recovering usable data can be extremely difficult even if the raw data still exists. BitLocker (Windows) / FileVault (Mac) example: many modern laptops have hardware-level encryption on by default; if the SSD is encrypted, the recovery key is required, otherwise recovered data will look like random noise. If the controller chip is dead, the lab must move the NAND chips to a matching healthy board, and the original platform / original TPM must also cooperate for any chance of recovering the key; swapping the board alone is usually not enough. Even professional recovery is not guaranteed, especially in TRIM-related cases.

Why SSD recovery is harder than HDD recovery: TRIM permanently deletes data; data is stored in fragmented form; encryption is often enabled; controllers vary by manufacturer; even two SSDs of the same brand can behave differently internally. SSD recovery can fail when: TRIM has already cleared the relevant blocks; lost data has been overwritten by new files / updates / installations; or the SSD has severe controller damage or is no longer detected by the motherboard / OS.

DIY taboos, restated: do not repeatedly format, do not install recovery software on the same drive, do not open up the SSD yourself, do not keep rebooting a failing drive.

11. How to Back Up Properly: 3-2-1 + SSD-Specific Notes

11.1 The 3-2-1 rule

3 copies of the data: the original + two backups.
2 different storage media: e.g. internal SSD + external hard drive / NAS / cloud.
1 off-site copy: away from home / office, e.g. a cloud backup, protecting against theft, fire, and flooding.

This rule is widely regarded as the gold standard of data protection, originating from photographer Peter Krogh's concept in *The DAM Book*. For the full method, see the full 3-2-1 backup strategy guide.

11.2 Sync is not backup

Cloud sync services (Dropbox, Google Drive, OneDrive, etc.) are not real backups: if you delete locally, the cloud copy often gets deleted too; if the account is stolen or locked, the files go with it; version history is limited. Use backups with version history / snapshots to protect against ransomware and accidental deletion sync. See ransomware response: why sync is not backup.

11.3 SSD-specific notes

An SSD can die without warning — do not wait for symptoms before backing up. SMART catches wear-related degradation, but not sudden controller failure.
Read-only mode is a last chance, not a repair. The moment you see read-only, copy your files; do not reboot, do not "repair".
Check SMART regularly, but do not treat the "health percentage" as the only backup trigger.
Keep at least 10–20% free space, so the SSD's internal maintenance (TRIM / garbage collection) has room to work; a stuffed drive hurts performance and maintenance.
Power-loss protection: most consumer SSDs lack the capacitor-based power-loss protection of enterprise drives; a sudden power cut can damage data in the DRAM cache not yet written to NAND, and the critical mapping table. Enterprise SSDs use onboard capacitors to provide brief backup power, writing the cache and mapping table to NAND before shutting down in an orderly way. The practical response for home / SME users is: use a UPS, avoid hard power cuts, and do not unplug during writes.
A backup only counts once it has actually been restore-tested. (A matter of principle; this article cites no restore-test statistics.)
Do not rely on a single SSD as the only copy. The "2 media types" in 3-2-1 exists precisely to stop a single medium / single failure mode from taking all copies at once.

11.4 Why sudden power loss is so dangerous

A consumer SSD that loses power mid-write can corrupt data or the mapping table, sometimes bricking the drive. A consumer SSD without PLP (Power Loss Protection) that loses power while updating its internal mapping table can brick the whole drive, leaving the OS unable to boot and the partition showing as RAW and unreadable.

How PLP works: a row of small capacitors (supercapacitors or tantalum capacitors) provides brief backup power, letting the controller flush the DRAM cache to NAND after power loss. Enterprise SSDs almost always have PLP; most consumer SSDs do not, with only some higher-end models having some degree of PLP.

Practical advice: use a UPS / surge-protected socket; do not do large writes during thunderstorms / unstable voltage; do not hold the power button to force a shutdown (it increases Unsafe Shutdowns).

12. Myth-Busting

"SSDs never fail" — wrong. SSDs are more shock-resistant with no mechanical wear, but still have a write-endurance limit — not immune. SSDs do fail — sometimes gradually with warning signs, sometimes instantly with no warning.

"A strange noise means the SSD is failing" — wrong. An SSD has no spinning platters, no read/write heads, no motor, so physically it cannot make grinding / clicking / buzzing sounds. A strange noise is usually the cooling fan, or another mechanical HDD in the same machine. Judging SSD failure by "listening" will miss the real warning signs or misdiagnose a fan problem.

"Putting it in the fridge will fix it" — wrong, and completely inapplicable to SSDs. The fridge trick comes from HDD-era folklore: low temperature makes metal contract, temporarily reseating shifted mechanical parts. An SSD has no moving mechanical parts that could reseat through contraction, so freezing will not revive an SSD. Extreme temperature differences can instead damage chips and components on the SSD circuit board, and condensation adds further damage.

"Tapping / hitting it will fix it" — wrong. Tapping / hitting likewise comes from folklore about stuck HDD mechanics. An SSD has no moving mechanical parts, so tapping does nothing except add physical damage risk. (Note: this article cites no study specifically on "tapping an SSD"; the above is an inference from "SSDs have no moving parts".)

"SSDs need defragmenting" — wrong; do not manually defrag an SSD. HDD defragmentation rearranges scattered files into contiguous runs to reduce head movement; an SSD uses flash memory, where access speed is barely affected by location, so defragmentation does not help performance — and each defrag is extra writing that consumes limited write endurance. Windows 10/11's "Optimize Drives" tool automatically detects the drive type: for SSDs it runs TRIM, not traditional defragmentation; it runs automatically weekly by default, and users do not need to do it manually. The one thing to avoid is downloading a third-party defrag tool to "defrag an SSD". Supplementary note: Windows sometimes performs limited file-system metadata tidying on SSDs, which is not the same as traditional defragmentation; do not confuse the two.

"Swapping the circuit board will fix it" — a misunderstanding. Every hard drive / SSD has different firmware controlling data flow and operation; casually swapping in another drive's circuit board can make things worse. Professionals do similar procedures in specific cases, but with specialist equipment and matched firmware.

"Health percentage still X%, so it must be safe" / "Low health percentage means instant death" — both wrong. The percentage mostly reflects TBW consumption, not a precise measure of physical lifespan; a drive with a high health percentage but failing on the three key error counts (Reallocated / Pending / Uncorrectable) is the truly dangerous one; conversely, a drive with a low percentage but zero error counts may not need immediate replacement.

"Deleted files on an SSD can never be recovered" / "A dead SSD can never be recovered" — both too absolute. Deleted + TRIM active = close to impossible; broken drive = TRIM irrelevant, data may still be in the NAND. Do not conflate the two situations.

"Running chkdsk means repairing the SSD" — wrong. chkdsk is a file-system consistency tool, not a data recovery tool; running it on a failing SSD can trigger TRIM and permanently erase data.

"Synced to the cloud means backed up" — wrong. Sync is not backup; accidental deletion / ransomware / account problems can follow the sync.

13. Hong Kong User Safety Checklist: The Safe Order of Operations When You Suspect SSD Trouble

Step 1: Stop — no more writes (0–5 minutes)

1. Stop all writes immediately: no more copying, installing, downloading, or opening large programs. If you deleted something by mistake, power off immediately (TRIM erasure happens during idle time).

2. Do not run CHKDSK (see Section 5 for why), do not format, do not defrag / "optimise", do not install recovery software on the same drive.

3. Judge the value of the data: if it is irreplaceable (business records, family photos, legal documents), jump straight to Step 5 (stop and seek professional help); do not try it yourself.

Step 2: Record the symptoms (don't rely on memory)

4. Write down: when it first happened, what you were doing (large write? booting? gaming?), what messages appeared (blue-screen code, read-only prompt, disappearance), whether BIOS detects it, whether it recovers after reboot, any smell / abnormal heat. These records matter a great deal for a professional recovery assessment.

5. If Windows pops up a "storage device may be at risk of failure" notification, click into it to see which drive and which warning.

Step 3: Do one read-only health check (don't scan repeatedly)

6. Use the manufacturer's tool (Samsung Magician, WD Dashboard / SanDisk Dashboard, Crucial Storage Executive, etc.) or CrystalDiskInfo / GSmartControl to look at SMART once, and copy down these numbers: Percentage Used, Available Spare, Media and Data Integrity Errors, Critical Warning, temperature, Data Units Written, Unsafe Shutdowns. Do not look only at the "health percentage".

7. Judge:

Critical Warning non-zero, Media Errors non-zero, Available Spare at threshold, or read-only → back up immediately (Step 4), then plan replacement; do not wait.
Drive not detected in BIOS at all, repeatedly disconnecting, abnormally hot → jump to Step 5.
Only the health percentage has dropped, but all other indicators are normal → no need to panic, but strengthen backups and track the trend.

Step 4: If the drive is still detected + readable, back up first (don't "fix" first)

8. Copy important files to another physical drive / external drive / cloud; do not save them back to the same SSD. A read-only drive can usually still be copied from directly.

9. If the drive is unstable (copying disconnects, system freezes), prioritise a sector-by-sector clone (e.g. ddrescue — double-check source and target device names first; swapping them erases data), then work on the clone; do not repeatedly scan the original drive.

10. Only after the backup is complete, think about "repair": check cables / slots (SATA cable, M.2 retaining screw, heatsink), check firmware updates (already backed up before updating), check temperature and airflow. Non-zero CRC / SATA errors may point to cable, controller, or interface problems.

Step 5: Stop-and-seek-professional-help situations (stop on any one)

11. Stop conditions are the same as the Section 10 dividing table (not detected, physical damage, repeated disconnection, no manufacturer fix, getting worse with each attempt, data too important to risk) — stop on any one.

12. Before seeking professional recovery, prepare: the symptom record from Step 2, the SMART transcript from Step 3, what attempts have already been made, and encryption status (BitLocker / FileVault on or off, where the recovery key is). If encryption is on and the key is not in hand, say so honestly — it greatly affects feasibility.

13. Pricing: quotation on request. Ask clearly about the assessment fee, whether the quote is result-based, and how failure is charged. Keep the order of warranty claim and data recovery separate: if the data matters, do data recovery first, then handle the warranty claim.

Step 6: Replacement + prevention (after data is safe)

14. If the drive has shown Critical Warning, spare capacity exhausted, Media Errors non-zero, or repeated drive drops, do not "observe a bit longer" — plan a replacement.

15. Prevention: important files in at least two places (external drive + cloud); monitor health + keep 10–20% free space; use a UPS / surge protection; keep cooling; update firmware regularly; do not wait until something breaks to think about backups.

Hypothetical example (not a real case): Suppose your computer keeps freezing during large file copies, and CrystalDiskInfo shows Media and Data Integrity Errors as non-zero. The correct response is not to run chkdsk, but to immediately copy important files to an external drive, then prepare to seek professional help following Step 5 above. This is a hypothetical scenario written to illustrate the process; it is not something that actually happened to a particular customer.

FAQ

Q1: My SSD health is still 95% — does that mean it is definitely safe?

No. The health percentage is mostly a simplified estimate of TBW consumption; it does not cover controller, firmware, circuitry, or data integrity. Look at the full set of numbers — especially whether error counts are growing. See Section 3 for details.

Q2: My SSD suddenly disappeared from Windows. What should I do first?

Power off completely and disconnect, then reseat the SSD; try a different cable, port, slot, or another computer to isolate the source of the problem. Enter BIOS and check whether it is detected. Remember: when data needs rescuing, do not repeatedly power on, do not run chkdsk, do not initialise / format, and do not install software on the failing drive.

Q3: My drive has become read-only. Does that mean it is dead?

No. Read-only is the controller's protection mode, designed to preserve remaining data; files can usually still be copied straight off. The correct response is to immediately copy the most important files; do not reboot, do not run chkdsk, do not attempt formatting or a firmware update. But read-only does not mean safe — back up immediately, do not wait.

Q4: I deleted a file by mistake. Is there any chance of getting it back?

If it is a healthy SSD with TRIM active, recovery is close to impossible. The only chance is to power off and disconnect immediately, because TRIM erasure happens during idle time. Do not keep using the drive, and do not install recovery software on the same drive. For an external SSD connected via USB, the TRIM command may not reach the SSD, so recovery chances may be higher — but that is a "possibility", not a guarantee.

Q5: How do NVMe and SATA SSD failures differ?

NVMe health reports are more consistent across manufacturers, with explicit Critical Warning bits; Windows built-in monitoring also only targets NVMe. SATA uses the older ATA SMART framework, with attribute meanings varying by manufacturer. NVMe drive drops are more often tied to controller hang, HMB, and heat; SATA problems are more often cable, slot, or power. But both types share common symptoms: read-only, disappearance, slowdown.

Q6: Can a firmware update repair an SSD that has already failed?

No. A firmware update cannot repair an already dead drive (e.g. the SanDisk "No Media" failure, where the underlying hardware has failed and the firmware updater has nothing to talk to), nor will it reverse health degradation that has already occurred (Samsung 990 Pro case). Firmware updates prevent and fix known problems; back up before updating.

Q7: How much does professional SSD data recovery cost in Hong Kong?

This article cites no Hong Kong local public pricing source, so pricing is quotation on request. Before engaging professional recovery, ask about the assessment fee, whether the quote is result-based, and how failure is charged. As a note: in international markets, professional SSD data recovery generally ranges from several hundred US dollars to well over US$1,000, depending on failure type and workload — this is an international reference figure, not a Hong Kong price.

Sources

---

Frequently Asked Questions

My SSD health is still 95% — does that mean it is definitely safe?

My SSD suddenly disappeared from Windows. What should I do first?

My drive has become read-only. Does that mean it is dead?

I deleted a file by mistake. Is there any chance of getting it back?

How do NVMe and SATA SSD failures differ?

Can a firmware update repair an SSD that has already failed?

How much does professional SSD data recovery cost in Hong Kong?

This article cites no Hong Kong local public pricing source, so pricing is **quotation on request**. Before engaging professional recovery, ask about the assessment fee, whether the quote is result-based, and how failure is charged. As a note: in international markets, professional SSD data recovery generally ranges from several hundred US dollars to well over US$1,000, depending on failure type and workload — this is an international reference figure, not a Hong Kong price.