I lost a drive without warning once. It wasn't old — just over two years in. One morning it was there, the next it wasn't recognized by the BIOS. The data was recoverable, barely, at significant cost and inconvenience. The thing that frustrated me most wasn't the failure — SSDs fail, that's accepted — it was that the warning signs were probably there for weeks and I hadn't been looking.

That was the last time I ignored SSD health monitoring. Here's everything I've learned about reading the numbers, understanding what they mean, and using them to make smarter backup decisions.

What S.M.A.R.T. Data Actually Tells You

S.M.A.R.T. (Self-Monitoring, Analysis and Reporting Technology) is a standardized system built into almost every storage drive manufactured in the last 25 years. Every SSD constantly collects internal diagnostic data and stores it in a log that software can read.

The values most relevant to SSD health are different from those relevant to hard drives. Here are the attributes that matter most:

Wear Leveling Count (S.M.A.R.T. ID 177) This value counts how many times flash cells have been written to on average. SSDs use wear leveling algorithms to distribute writes evenly across cells — this attribute tracks how far that process has progressed. A raw value starting at 100 and counting down toward 0 (or starting at 0 and counting up) depending on manufacturer implementation. Samsung and WD both use a 0–100 scale where 100 is brand new and lower numbers indicate more wear.

Total Bytes Written — TBW Consumed (S.M.A.R.T. ID 241, 247, 248) Tracks the total volume of data written to the drive since manufacture. Comparing this to the manufacturer's rated TBW gives you a percentage of rated endurance consumed. A drive that has written 200TB against a 300TB TBW rating has consumed 67% of its rated life.

Power-On Hours (S.M.A.R.T. ID 9) The total number of hours the drive has been powered on. This matters for context — a drive with 50,000 power-on hours and low write count may have capacitor wear even if the NAND cells look healthy.

Reallocated Sectors Count (S.M.A.R.T. ID 5) When a flash cell fails, the SSD controller marks it as bad and remaps it to a spare cell (reallocated). A small number of reallocated sectors is normal over time. A rapidly increasing count — especially any increase in a short period — indicates accelerating cell failure. On SSDs, this attribute is less common than on HDDs but not meaningless.

Uncorrectable Error Count (S.M.A.R.T. ID 187) Data that was read with errors that ECC (Error Correcting Code) could not fix. Any non-zero value here is a serious warning. On a healthy drive this should always be 0.

Temperature (S.M.A.R.T. ID 194) Current drive temperature. NVMe drives should operate below 70°C under load. Sustained temperatures above 75°C indicate inadequate cooling and accelerate NAND degradation.

Available Spare (NVMe-specific) NVMe drives report a percentage of spare blocks remaining. When this drops below the Available Spare Threshold (also reported), the drive is signaling that it's running low on spare capacity and may become unreliable. Most drives start at 100% and the threshold is 10%.

Setting Up CrystalDiskInfo: Step-by-Step

CrystalDiskInfo is the most widely used free S.M.A.R.T. reader on Windows. Here's how to get it working properly.

Step 1: Download from the official source Go to crystalmark.info and download the Standard Edition installer or the portable ZIP (no installation required). The portable version is useful if you want to run it from a USB drive on any machine.

Step 2: Launch and read the main screen CrystalDiskInfo opens with a colored health status bar for each drive — Good (blue), Caution (yellow), or Bad (red). The color reflects whether any S.M.A.R.T. attribute has a value below its threshold.

Step 3: Check the critical values For each drive, scroll through the attributes list. Look specifically at:

  • Wear Leveling Count — should be high (close to 100 for newer drives)
  • Total Host Writes — divide by your drive's TBW to get percentage used
  • Power-On Hours — context for how old the drive is
  • Reallocated Sectors — should be 0
  • Uncorrectable Error Count — should be 0
  • Temperature — should be under 50°C at idle, under 70°C at load

Step 4: Enable startup monitoring Go to Function → Resident. This puts CrystalDiskInfo in your system tray at startup and monitors drive health passively. You'll see a colored icon that changes if any value drops into Caution territory.

Step 5: Enable health alert notifications Go to Function → Alert. Enable "Alert If Health Status Changes" and "Alert If Temperature Is Too High." This will pop up a notification if any drive develops a warning condition.

Step 6: Check periodically, not obsessively Monthly checks are sufficient for most setups. If you're running a heavy workstation, weekly is reasonable. Daily checking doesn't provide useful additional information — S.M.A.R.T. values change slowly except in acute failure scenarios.

Manufacturer Tools: Samsung Magician and WD Dashboard

For drives from Samsung and Western Digital, the manufacturer's own software provides deeper integration than CrystalDiskInfo.

Samsung Magician Available at samsung.com/semiconductor, Samsung Magician connects to Samsung drives (860 EVO, 870 EVO, 990 Pro, 990 EVO) and provides:

  • A clean health percentage dashboard
  • Drive temperature with historical graph
  • Total bytes written vs rated endurance bar
  • Secure Erase function (useful before selling a drive)
  • Firmware update notifications
  • SMART data with Samsung-specific attribute explanations

The health percentage is the most readable indicator for non-technical users. 100% means excellent. Below 90% in the first two years warrants attention. Below 70% means the drive is aging and backup urgency increases.

WD Dashboard Available at support.wdc.com for WD and SanDisk drives (WD Black SN850X, WD Blue SN580, SanDisk Ultra). Provides similar functionality to Samsung Magician:

  • Drive health and endurance tracking
  • Temperature monitoring
  • Drive performance test
  • Secure Erase

Both tools require the drive to be connected as the primary storage device (not external) for full SMART access. NVMe drives show more attributes than SATA drives in these tools due to richer NVMe protocol reporting.

What the Health Percentage Actually Means

The health percentage shown in CrystalDiskInfo, Samsung Magician, and WD Dashboard is a composite calculation based on multiple S.M.A.R.T. values. Different tools calculate it differently, but they all weight the most critical attributes (Wear Leveling Count, Available Spare, Uncorrectable Errors) heavily.

  • 100–90%: Drive is healthy. No action needed beyond normal backup habits.
  • 89–70%: Drive is aging normally. Increase backup frequency. Begin planning for replacement in the next 1–2 years.
  • 69–50%: Drive has consumed a significant portion of its rated life. Immediate backup review is important. Replacement within the year is advisable for critical-use drives.
  • Below 50%: Drive is at elevated risk. Back up critical data immediately. Replace before the next drive failure window.

The health percentage is not a countdown timer. Many drives operate reliably at 30–40% health for years. But below 50%, the margin for unexpected failure narrows, and backup systems become critical rather than optional.

SSD Failure Modes Unique to Flash Storage

SSDs fail differently from HDDs, and understanding those failure modes helps you know which S.M.A.R.T. values to prioritize.

Sudden death: The most common and most dangerous SSD failure mode. The drive is recognized one day, completely unresponsive the next. No gradual degradation, no warning. This typically results from firmware bugs, controller failures, or capacitor failure — none of which S.M.A.R.T. reliably detects in advance. Backup discipline is the only real protection against this mode.

Gradual wear: Flash cells can only endure a finite number of program-erase cycles. As cells reach their limit, error rates increase and the controller remaps them to spare blocks. S.M.A.R.T. tracks this through Wear Leveling Count and Available Spare. This is the failure mode that health monitoring is actually good at predicting.

Thermal shutdown: Overheated NVMe drives can throttle dramatically and in rare cases trigger protective shutdowns. CrystalDiskInfo's temperature tracking and alerts catch this.

Write cliff: Some drives — particularly QLC drives — experience a sharp write speed decrease when the SLC cache is exhausted or when the drive reaches a high fill level (above 85–90% capacity). This isn't a failure, but it can feel like one. Keeping drives below 80% full avoids this.

Retention failure: Flash storage needs periodic refreshing. Data written and left unread for many years on a powered-off drive can be lost through charge leakage — especially on QLC drives. This is extremely rare in normal use but matters for archive applications.

When to Act on Warning Signs vs Ignore Them

Not every alert requires immediate action. Here's how to interpret common scenarios:

Warning Condition Action Required Urgency
Wear Leveling Count drops below 90 Increase backup frequency Low — normal aging
Reallocated Sectors Count goes from 0 to 1–5 Back up immediately, monitor closely Medium — investigate
Reallocated Sectors Count increases weekly Replace the drive High
Uncorrectable Error Count > 0 Back up now, replace soon High
Available Spare drops below threshold Replace within weeks Critical
Temperature consistently above 75°C Improve cooling, monitor Medium
Health drops below 70% Schedule replacement Medium-High
Sudden "Caution" status on any attribute Full backup immediately High

The one value that should never be non-zero is Uncorrectable Error Count. Any data that the drive couldn't read correctly represents a real risk to your files. Treat this as a "back up today" signal regardless of the drive's age.

TBW Ratings: How Close Do Real Drives Get?

TBW (Terabytes Written) is the manufacturer's endurance rating — the total amount of data the drive is designed to handle over its life. These ratings are calculated from NAND cycle limits, controller efficiency, and typical workload assumptions.

In practice, most consumer drives never approach their TBW limits:

Drive Rated TBW (1TB) Typical Annual Writes (Desktop) Years to TBW
Samsung 870 EVO 1TB 600 TBW 15–30 TB 20–40 years
Samsung 990 Pro 1TB 600 TBW 15–30 TB 20–40 years
WD Black SN850X 1TB 600 TBW 15–30 TB 20–40 years
Crucial MX500 1TB 360 TBW 15–30 TB 12–24 years
WD Blue SN580 1TB 300 TBW 15–30 TB 10–20 years

For typical desktop use, TBW is not a meaningful limiting factor. You're likely to replace the drive due to capacity needs or new technology adoption long before the NAND wears out.

Where TBW matters more: video editing workstations, database servers, and systems doing continuous logging or cache-heavy workloads. At 100GB+ of writes per day, a 600 TBW drive lasts less than 20 years — still very long — but TBW becomes worth monitoring rather than ignoring.

Backup Schedule Recommendations Tied to Health Data

Your backup frequency should scale with drive health. Here's a practical framework:

Drive health 90–100%: Weekly automated backup to a second drive or NAS. Monthly cloud backup or offsite rotation. This is the minimum baseline for any important data.

Drive health 70–89%: Increase to twice-weekly automated backup. Verify that backup jobs are completing successfully — don't just set and forget. Check backup restore capability once per quarter.

Drive health 50–69%: Daily backup of critical files. Cloud backup active. Plan replacement purchase. Test restore immediately to confirm backup integrity.

Drive health below 50%: Immediate full backup to two separate destinations. Replace the drive before the next quarter ends. Do not rely solely on this drive for any data you cannot afford to lose.

Drive showing Caution or Bad status: Stop using the drive for primary storage. Make a full backup with EVERYTHING before the next restart. Replace immediately.

Frequently Asked Questions

Does S.M.A.R.T. predict all SSD failures? No. S.M.A.R.T. is good at detecting gradual wear and thermal issues, but sudden controller failures and firmware bugs — which cause many SSD deaths — don't produce S.M.A.R.T. warnings. It's a helpful layer, not a guarantee.

How do I check SSD health on a Mac? DriveDx (paid) and Smart Utility (paid trial) provide SMART monitoring on macOS. For NVMe drives in Apple Silicon Macs, Apple's own diagnostics in System Information show drive health. Third-party NVMe drives in external enclosures may have limited SMART access on macOS due to driver limitations.

Is a drive with "Caution" status still safe to use? Temporarily, yes. The Caution status means a value has crossed a threshold — it doesn't mean the drive is about to die. Immediately back up everything, evaluate which attribute triggered the warning, and decide on replacement timeline based on the severity of the flagged value.

Does filling an SSD to 100% damage it? Very briefly, no. But keeping a drive consistently above 90% full degrades performance and accelerates wear by reducing the controller's ability to manage spare blocks efficiently. Keep drives below 80% full for best longevity and performance.

How do I monitor SSD health on Linux? smartctl -a /dev/nvme0 (for NVMe) or smartctl -a /dev/sda (for SATA) using the smartmontools package. GSmartControl provides a graphical interface on Linux.

The combination of CrystalDiskInfo running in the tray and a drive-specific tool like Samsung Magician takes about five minutes to set up. That five minutes is the difference between a planned drive replacement and an emergency data recovery.