Drives keep their own statistics. DiskSalvor reads them directly (the same way CrystalDiskInfo does): NVMe SSDs always, SATA drives directly, and USB drives when the bridge allows it. Memory cards and most sticks have no health log.
Each item has a count (the real number of events) and a score (the drive's own 1–253 rating; lower is worse). The drive only calls itself "failing" when a score drops to its fail line, which is often too late. The counts are more useful:
| Item | Healthy | Meaning when it's above zero |
|---|---|---|
| 197 Pending sectors | 0 | Sectors unreadable right now. The key number for recovery: copy first, gently. |
| 198 Uncorrectable (offline) | 0 | Sectors that failed the drive's own scan. |
| 5 Reallocated sectors | 0 (a few can be stable) | Bad spots already swapped for spares. Growing = deteriorating. |
| 187 Reported uncorrectable | 0 | Reads the drive couldn't correct. |
| 188 Command timeouts | 0 | The drive stopped answering: weak power, cable, or the drive. |
| 199 Interface CRC errors | 0 | Transfer errors: almost always the cable, port or USB adapter, not the disk. |
| 10 Spin retries | 0 | The motor struggled to spin up. |
| 9 Power-on hours, 194 Temperature | – | Age, and heat (above 50–55 °C speeds failure). |
| NVMe: Percentage used | under 100% | Rated write endurance used up. |
| NVMe: Available spare / Media errors / Critical warning | high / 0 / none | Spare flash left; unreadable data; serious condition. |
Self-tests: the drive can test itself (short ≈ 2 min, extended = the whole surface). Its log of past results is shown too. An extended test on a failing drive takes hours and adds wear, so copy your data first.