Three attributes decide the replace question: any non-zero pending sectors (C5) or uncorrectable sectors (C6) means plan replacement now, and a growing reallocated count (05) means the same. CRC errors (C7) usually mean a bad cable, not a bad drive. Back up before anything else, then use the decoder below.
| Attribute | Healthy | Warning | Verdict |
|---|---|---|---|
| 05 Reallocated Sectors Count | 0 | Any non-zero, especially a GROWING count | A few stable historical reallocations on an old drive: monitor across scans. A growing count: the surface is deteriorating — plan replacement now. |
| C5 Current Pending Sectors | 0 | Any non-zero value | Any non-zero C5 = plan replacement now. Pending sectors are reads the drive could not complete — data in those sectors is already at risk. |
| C6 Offline Uncorrectable Sectors | 0 | Any non-zero value | Any non-zero C6 = plan replacement now. These are sectors the drive tried and failed to fix — confirmed damage, not suspicion. |
| C7 UDMA CRC Error Count | 0 (or historical and stable) | Growing count | Usually the CABLE, not the drive. Replace the SATA cable and reseat both ends; if the count stops growing, the drive is fine. |
| 09 Power-On Hours | Context, not verdict | 30,000-40,000+ hours = aging fleet | Not a failure signal by itself — a planning signal. High hours plus ANY of 05/C5/C6 shifts the answer from monitor to replace. |
| 03/0A Spin-Up Time / Retry Count | Stable time, 0 retries | Rising spin-up time or any retries | Rising spin-up metrics point at the motor/bearing assembly — a mechanical wear signal. Pair with your ears: whining or grinding confirms it. Plan replacement. |
When the drive detects a failing sector it remaps it to spare area; this attribute counts those remaps. The remap itself is the drive working as designed — the signal is the trend. A count that increases between scans means the media is actively degrading, and reallocation events cluster: drives that start reallocating tend to keep reallocating.
A pending sector is one the drive failed to read and has flagged for reallocation-on-next-write. Until that write happens the data in it is unreadable, which is why C5 is the attribute most directly tied to data loss in progress. Back up immediately, prioritizing the files that matter; a surface rewrite may clear the count, but the underlying cause rarely stops there.
Where C5 marks sectors under suspicion, C6 counts sectors the drive's own offline scanning confirmed unreadable and uncorrectable. It is the strongest single replace signal in the SMART table: the error-correction machinery has already lost the fight in those locations.
CRC errors count corruption on the wire between drive and controller — bad cable, loose connector, marginal backplane slot. The attribute never resets, so an old stable number is history, not disease. It earns a place in the decoder because it is the scary-looking attribute that most often does NOT mean replace — and misreading it buys a drive nobody needed.
Power-on hours is the odometer. Drives commonly serve 3-5 years (26,000-44,000 hours); Backblaze's long-running fleet statistics show failure rates following a bathtub curve — elevated early (infant mortality), low through mid-life, rising again with age. Hours alone never condemn a healthy drive, but they change how much benefit of the doubt the other attributes get.
Spin-up time measures how long the platters take to reach speed; retries count failed attempts. Both degrade when the spindle motor or bearings wear. Mechanical wear is progressive and not repairable in the field, which is why a clear trend here reads like early clicking: the drive is telling you how it will eventually die.
Live replacement options · prices updated every 4-5 hours · last checked 26 min ago · CMR-family drives only (RAID-safe)
| Drive | Cap | Cond | Warranty | $/TB | Price | |
|---|---|---|---|---|---|---|
| Toshiba MG Series 8TB Enterprise SATA | 8TB | REFURB | 5 yr | $10.63 | $85.00 | Buy → |
| MDD 3TB SAS 6G 7200RPM | 3TB | NEW | 5 yr | $11.33 | $33.99 | Buy → |
| Seagate Constellation ES.2 3TB SAS | 3TB | REFURB | see listing | $11.63 | $34.90 | Buy → |
| Seagate 3TB Enterprise Capacity SAS 6G | 3TB | REFURB | see listing | $11.67 | $35.00 | Buy → |
When the BIOS or OS announces a predicted failure, the sequence is short and the order matters. Back up first — the message means the drive's own firmware has crossed a manufacturer-set threshold, and the remaining useful life is unknowable. Copy irreplaceable data in priority order before running anything else; long diagnostics add hours of sustained load to a drive you have just been told is failing.
Verify second, with the manufacturer's tool: SeaTools for Seagate, WD Dashboard for Western Digital, the vendor diagnostic for others. Two reasons for the brand tool specifically — its result is what the warranty process recognizes, and it rules out the rare false positive from a generic utility misreading vendor-specific attributes. Then decide with the decoder above: threshold trips driven by 05/C5/C6 mean replacement (under warranty, that is an RMA — check your serial on the manufacturer's warranty page); a clean manufacturer pass with only C7 noise means check the cable and re-test before spending anything.
What the sequence never includes: silencing the warning in firmware. Disabling SMART notifications in BIOS is offered on many boards, and it is a trap dressed as a setting — the failure proceeds on schedule, minus the reminders. The warning is doing exactly its job: giving you the window that sudden failures never grant.
Honesty about the tool's limits belongs on the same page as the decoder. SMART monitors the failure modes it can measure — media degradation, mechanical wear, interface integrity — and is blind to others: controller-board electronics can die between one poll and the next with a spotless attribute table, and Backblaze's fleet-scale experience has long shown that a meaningful share of drive failures arrive with no SMART warning at all. Fleet-wide annualized failure rates in their published statistics run in the low single digits (roughly 1-2% across recent years), which is both reassuringly low and a reminder that at any scale, some drives simply die unannounced.
The practical conclusions: treat SMART as an early-warning system that catches most — not all — failures in progress; treat backups as the mechanism that makes the unannounced failures survivable; and treat the attributes on this page as replace signals rather than repair projects. A drive that has started failing physically does not return to health, and the market data on this site exists to make the replacement decision a cheap, informed one rather than a panic purchase. For the planning version of this question — replacing drives by age before symptoms appear — that guide is coming in this cluster's next phase. Until then, the working heuristic: a drive past 30,000-40,000 power-on hours has served its statistical tour, and any surface attribute activity on a drive that age is the retirement notice — replace on your schedule, at the live prices above, rather than on the drive's.
No — not the physical kind. chkdsk and fsck repair filesystem structures; they do not repair the magnetic surface or the mechanics that SMART attributes measure. A surface scan can trigger the drive to remap a pending sector (C5 may drop as 05 rises), which is bookkeeping, not healing: the damaged area is retired, not repaired. For non-zero C5/C6 the honest fix is a backup and a replacement.
Treat it as real. That message means the drive itself has tripped a manufacturer-set threshold — it is the drive's own firmware predicting failure, not an over-eager utility. Back up immediately, then confirm with the manufacturer's tool (SeaTools for Seagate, WD Dashboard for WD) whose verdict the warranty process recognizes.
No — that is the trap. Disabling the warning does not slow the failure; it removes the last automatic reminder that one is coming, and the next notification you get may be the drive not mounting. If the message is wrong, the manufacturer tool will say so; if it is right, you want to hear it every boot until the backup is done.
Blue/green 'Good' means no monitored attribute is past its threshold. Yellow 'Caution' typically means non-zero reallocated (05) or pending (C5) sectors — worth exactly the attention this page's verdicts describe. Red 'Bad' means a threshold is exceeded — the drive's own failure prediction. The colors summarize the same attributes decoded above; when in doubt, read the raw values rather than the paint.
A small, historical, stable count on an aging drive can be monitored — re-scan weekly and watch attribute 05. The moment the count grows between scans, or is joined by pending (C5) or uncorrectable (C6) sectors, the honest verdict flips to replace. Trend is everything with reallocations.
No. SMART catches the failure modes it measures — surface degradation, mechanical wear, interface errors — and misses sudden electronic failures entirely. Backblaze's fleet experience is instructive: a meaningful share of failed drives showed no advance SMART warning. SMART is a smoke detector, not a guarantee; backups remain the actual safety net.
The manufacturer's own: SeaTools for Seagate/LaCie, Western Digital Dashboard for WD, and the vendor diagnostics for others. Third-party tools read the same attributes, but RMA processes ask for the manufacturer tool's test result, so run it before filing — details in our RMA guide.
Yes — urgency. In an array, a member drive throwing C5/C6 is a rebuild waiting to happen, and you want to replace it on your schedule (array healthy, backup fresh) rather than the drive's. Our failed-RAID-drive guide covers the replacement rules and the degraded-array risks.