Mechanical hard drives hide early platter failure behind basic Windows status queries. Inspect raw hexadecimal or decimal SMART values in CrystalDiskInfo or smartctl instead: any non-zero count on Reallocated Sectors 05, Current Pending Sectors C5, or Uncorrectable Sectors C6 demands immediate backup before running chkdsk, whereas UDMA CRC C7 points at a faulty SATA cable.
At a glance
- Class
- Articles
Provenance
- Nov 14, 2026Published
Spinning hard drives die differently from solid-state drives. Flash cells wear down on a predictable write curve—covered step by step in our NVMe and SATA SSD health check guide—whereas a mechanical 3.5-inch or 2.5-inch hard drive is a tiny precision air-bearing machine flying read/write heads nanometers above glass or aluminum platters at 5,400 or 7,200 RPM.
When a mechanical drive starts failing, Windows Explorer freezes for twenty seconds opening a folder, transfers crater to 0 KB/s before jumping back up, or Event Viewer fills with disk Event ID 7 (bad block) and storahci Event ID 129 resets. Yet if you follow decades-old forum advice to run a built-in Windows status command, Windows will tell you everything is OK.
Never trust OK. Read raw SMART counters first. Leave every repair tool closed until you see those numbers.
Why Windows built-in checks miss dying hard drives
Two separate blind spots hide mechanical drive trouble in Windows 11.
First, opening Settings, System, Storage, Advanced storage settings, Disks & volumes, and Properties reveals nothing for a hard drive. Microsoft’s storage critical-warning documentation states explicitly that Windows 11 monitors NVMe SSDs only, leaving SATA SSDs and hard disk drives completely unmonitored in Settings.
Second, older tutorials still tell readers to type wmic diskdrive get status. On my own Windows 11 25H2 desktop, wmic already fails as an unrecognized command because Microsoft removed the WMIC command-line tool from current builds (24H2 and 25H2), and its PowerShell equivalent (Get-CimInstance Win32_DiskDrive | Select-Object FriendlyName,Status) relies on the drive’s coarse ATA SMART RETURN STATUS boolean. That boolean stays OK until a normalized 100→1 score drops below the manufacturer’s final warranty-death threshold—often after hundreds of sectors have already gone unreadable.
Read the raw counters in CrystalDiskInfo or smartctl
Ignore the Current, Worst, and Threshold columns on your first pass. Those three columns are vendor-scaled scores from 100 or 200 downward, and every manufacturer scales them differently. Look straight at Raw Values.
- On Windows (CrystalDiskInfo): Before reading anything, click Function → Advanced Feature → Raw Values →
10 [DEC]. Out of the box, CrystalDiskInfo prints raw numbers in 12-digit hexadecimal (00000000001A=26bad sectors), which confuses people into misreading multi-byte Seagate counters or overlooking small hex letters. - On Linux, Proxmox, TrueNAS, or macOS (
smartctl): Runsudo smartctl -a /dev/sdX(orsmartctl -x /dev/sdXfor extended device statistics; for USB enclosures try-d sat /dev/sdXso SAT passthrough bridges the USB-to-SATA chip).
The five SMART attributes that decide a mechanical drive’s fate
CrystalDiskInfo’s official health criteria flip a hard drive from blue Good to yellow Caution the instant 05, C5, or C6 climbs off zero, matching Backblaze’s fleet SMART failure study where attributes 5, 187, 188, 197, and 198 showed the sharpest link to imminent replacement:
| Hex / Dec ID | Attribute Name | Healthy Raw Value | What a Non-Zero Raw Value Means |
|---|---|---|---|
05 / 5 | Reallocated Sectors Count | 0 | The drive hit unreadable/damaged magnetic spots, retired them permanently, and remapped their logical addresses to spare factory tracks. |
BB / 187 | Reported Uncorrectable Errors | 0 | Reads (common on Seagate drives) that hardware error-correcting codes failed to reconstruct and returned an I/O error to the OS. |
C5 / 197 | Current Pending Sector Count | 0 | Weak sectors that failed a read and sit on a watchlist; on the next overwrite, the drive either recovers the sector or increments 05. |
C6 / 198 | Uncorrectable Sector Count | 0 | Sectors proven unreadable during an offline surface scan with no valid copy left to remap cleanly. |
C7 / 199 | UDMA CRC Error Count | Stable (0 or flat) | SATA cable / backplane noise—not platter death! Checksum corruption occurred on the wire between controller and drive PCB. |
Two vendor quirks save healthy drives from the trash bin:
C7(UDMA CRC Error Count) is a cable counter, and it never goes back down. IfC7reads412while05,BB,C5, andC6all read0, your platters are fine. Swap the SATA data cable (avoiding tight 90-degree folds or bundled coils pressed against noisy power leads), reseat the plug until the metal clip clicks, record today’s412in our diagnostic workbench, and check back in a week. As long asC7stays frozen at412, the fault is gone.- Seagate raw
01(Read Error Rate) and07(Seek Error Rate) pack two numbers into 48 bits. On Seagate Barracuda, IronWolf, and Exos drives, a huge decimal number in01or07is normal—the lower 32 bits count total millions of seeks or hardware reads, while only the upper 16 bits count actual errors! Switch CrystalDiskInfo back to16 [HEX]briefly if you own a Seagate:0000084F2A10means0000errors over0x084F2A10operations.
One honest fleet statistic keeps backups non-negotiable. In Backblaze’s published drive-failure analysis, 76.7% of failed drives incremented at least one of their five tracked warning stats before dying, meaning 23.3% failed with clean SMART numbers the day before. Zero pending sectors means healthy today. Tomorrow still requires a backup.
Acoustic symptoms: normal enterprise thumps vs head-crash clicks
SMART counters only update if the drive firmware can still read its System Area tracks on the platter. Pull the side panel off. Listen closely for ten seconds:
- Periodic dull hollow thump every 5 seconds at idle: Normal Preventive Wear Leveling (
PWL) on high-capacity Western Digital/HGST helium drives and modern NAS drives (best-nas-hard-drives-2026covers idle dBA differences across current NAS hard drives). The actuator sweeps briefly so lubricant does not pool under stationary heads. - Crisp rapid chatter during active file copies: Normal
7,200 RPMvoice-coil seek noise. - Rhythmic tick… tick… tick… spin-down right at power-on (and drive missing from BIOS): Classic actuator head failure or stuck/damaged servo track. Unplug power immediately—every extra power-on cycle scrapes magnetic oxide dust inside the sealed chamber and ruins cleanroom recovery odds.
- Faint buzzing beep every two seconds with zero spindle spin: Seized motor bearing or heads parked on the platter instead of the ramp (
stiction). Power off immediately; software cannot touch a seized spindle.
Why chkdsk /r kills wounded drives (and what to run instead)
I see this mistake destroy recoverable drives every month. A secondary hard drive starts stuttering with yellow C5 (Current Pending) or 05 (Reallocated) warnings, so the owner opens an admin prompt and runs chkdsk D: /r.
Read Microsoft’s chkdsk command documentation: /r implies /f (in-place NTFS metadata writes) and forces a full physical surface scan across every sector on the volume. On a 4TB to 20TB hard drive whose heads are already struggling over degrading platter coating, chkdsk /r hammers those failing tracks for ten to thirty straight hours while shuffling fragmented orphan clusters into found.000. Countless drives enter chkdsk /r at 80% readable and die completely at Stage 4 with zero files backed up.
Follow this exact triage order instead:
Case A: Drive still mounts in Explorer with small 05 or C5 counts (1–20)
- Do not run
chkdsk, defrag, or an extended SMART self-test yet. - Plug in a healthy replacement destination drive with enough free space.
- Copy your irreplaceable folders first in priority order (documents, tax records, family photos, project repos—skip re-downloadable Steam libraries and Linux ISOs!) using a non-hanging copy command that skips stuck files after 2 retries instead of freezing Explorer:(Keep
robocopy "D:\ImportantFolder" "E:\Rescue\ImportantFolder" /E /R:1 /W:2 /MT:1 /LOG:"E:\rescue-log.txt"/MT:1single-threaded! Multi-threaded copies thrash mechanical heads wildly back and forth across a wounded platter.)
Case B: Drive stalls Explorer, drops offline, or climbs rapidly in C5/C6
Stop mounting the filesystem in Windows (Windows background indexers and Defender keep poking unreadable MFT records). Boot a Linux live USB and clone the raw block device non-destructively onto a larger healthy disk using GNU ddrescue with a mapfile:
# Pass 1: sweep all fast, healthy areas first without scraping bad spots (-n)
sudo ddrescue -f -n /dev/sdX /dev/sdY rescue.map
# Pass 2: retry skipped damaged sectors carefully up to 3 times (-r3)
sudo ddrescue -d -f -r3 /dev/sdX /dev/sdY rescue.map
Because ddrescue logs every finished block in rescue.map, it rescues 99%+ of the healthy surface sequentially first before ever touching damaged tracks, and resumes cleanly from the exact byte if you have to power-cycle a hung drive. Run filesystem repair (chkdsk or ntfsfix/file carvers) only on the cloned target copy /dev/sdY, never on the dying original.
Case C: Drive has 0 across 05/BB/C5/C6 and you want a routine health test
Only when your data is backed up and raw media counters sit at 0 should you run a proactive surface test: sudo smartctl -t short /dev/sdX (~2 minutes) for quick checks, and sudo smartctl -t long /dev/sdX quarterly (or on day-one arrival, as detailed in our refurbished enterprise hard drive guide and Drive Arrival Census).
Frequently asked questions
Why doesn’t Windows 11 Settings show health for my hard drive?
Why does Get-CimInstance or wmic report OK on a failing hard drive?
Get-CimInstance Win32_DiskDrive | Select-Object Model,Status) only flips from OK to Pred Fail after a vendor’s normalized value falls all the way past its factory failure threshold—long after dozens of bad sectors have already corrupted files.What do CrystalDiskInfo Caution codes 05, C5, and C6 mean?
05 (Reallocated Sectors Count) counts damaged sectors the drive retired and remapped to spare tracks. C5 (Current Pending Sector Count) counts unstable sectors waiting to be read cleanly or remapped on the next write. C6 (Uncorrectable Sector Count) counts offline sector reads that ECC could not recover. Any raw value above zero on these attributes means copy your data off immediately.Does a high C7 (UDMA CRC Error Count) mean my hard drive is dying?
C7 (decimal ID 199, UDMA_CRC_Error_Count) counts checksum errors across the SATA data link between the motherboard or HBA and the drive’s PCB—caused by a loose, kinked, or unlatched SATA cable or dirty backplane fingers rather than failing platters. Once incremented, the historical count never resets to zero; replace the SATA cable and verify the raw number stops climbing.Should I run chkdsk /r on a slow or clicking hard drive?
chkdsk /r scans the entire physical disk surface to locate bad sectors alongside repairing filesystem metadata. Forcing hours of full-surface thrashing across degraded heads or shedding magnetic media frequently pushes a wounded drive into total head failure before you recover a single folder.Sources and corrections
- Last updated
- Methodology
- See our methodology for research and review standards. It draws on 7 cited sources, listed below, each checked against the original page on the date above. Windows 11 NVMe-only health monitoring and WMIC removal behavior come from Microsoft Support notices (also verified on our Windows 11 25H2 desktop). Mechanical SMART attribute thresholds and Caution rules come from Crystal Dew World’s official CrystalDiskInfo health-status criteria, smartmontools’
smartctl(8)manual, and Backblaze’s fleet SMART correlation study. Recovery order references Microsoft’schkdskcommand documentation and the GNUddrescuemanual. - Sources
- Microsoft Support: What to do about a critical warning for a storage device accessed 2026-11-14
- Microsoft Support: WMIC removal from Windows 11 accessed 2026-11-14
- Crystal Dew World: CrystalDiskInfo Health Status Criteria accessed 2026-11-14
- Backblaze Storage Pod Blog: What SMART Stats Tell Us About Hard Drive Failures accessed 2026-11-14
- Debian Bookworm Manual: smartctl(8) accessed 2026-11-14
- Microsoft Learn: chkdsk command reference accessed 2026-11-14
- GNU Project: GNU ddrescue Manual accessed 2026-11-14
- Corrections
- Spotted an error? Email a correction with the affected line and what you found. The draft includes this page’s address. Verified changes are recorded here. Contact details.