This week our Brno workshop had a 2.5” laptop drive on the bench. The owner had previously left it running all night under chkdsk /r. The drive now clicked more than when it arrived — and thousands of pending sectors had meanwhile turned into sectors that no longer read at all. This is an article about why „just run scandisk and see what happens" is, in 2026, advice for drives with no data on them.

What bad sectors are

A hard drive divides the platter surface into sectors (512 B or 4 kB). Each sector holds metadata (ECC codes, an address) and data. A bad sector means that, at this spot on the platter, something has happened that prevents the data from being read or safely rewritten:

  • Mechanical damage — scratches, scoring from a previous head crash
  • Magnetic degradation — a weak signal, retention limits exceeded
  • Electronic error — a read error that surfaced as non-recoverable
  • Firmware — the drive knows about the problem and reports it

Drives have spare sectors (typically several thousand) which they can use to reallocate bad sectors. The process works like this: the drive hits a sector that won’t read on the first try → it retries repeatedly → if it eventually succeeds, it moves the contents into a spare sector and marks the original as reallocated.

Screenshot of a reconstruction tool with a virtual RAID 0 configuration and a hex editor showing the contents of a recovered disk image.
Sector-level analysis does not work with "files" — it works with sector addresses and their raw contents in a hex editor. When a given sector won't read, the clone skips it and returns to it at the end.

How to tell whether you have bad sectors

SMART attributes (the most reliable signal)

Install CrystalDiskInfo (free), open the drive detail, and look for:

  • Reallocated Sectors Count (05) — how many sectors have already been remapped. Any value > 0 is a warning.
  • Current Pending Sector Count (C5) — sectors that won’t read, but the drive is waiting to see whether it succeeds next time. Any value > 0 is a reason to back up.
  • Uncorrectable Sector Count (C6) — sectors that can neither be read nor remapped. Serious.
  • Reported Uncorrectable Errors (BB) — errors that got past ECC

On SSDs the attributes are different (Wear Levelling Count, Total Host Writes, Percentage Used) — and defects develop differently, too.

System symptoms

  • Copying a file stalls at a particular spot and reports a „read error"
  • Windows reports „The file system is corrupt" (NTFS journal errors)
  • BSOD INACCESSIBLE_BOOT_DEVICE on startup
  • The drive is „slow when I play a specific video"
  • chkdsk finds an „unrecovered file record"

What is NOT worth trying (or makes things worse)

1. Repeatedly running chkdsk /r or scandisk on a mechanical fault

If the drive is clicking, grinding, or showing any mechanical symptom, chkdsk /r takes 6–24 hours, and the whole time the heads are clicking across the damaged platter. Platter wear increases. Data that could otherwise have been saved gets damaged.

An open 3.5-inch hard drive on a green cutting mat lit with green light, with the spindle, platter and head parking ramp visible.
When the drive is running, the air gap between the platter surface and the head is thinner than 5 nanometres. When chkdsk, after hours of reading, passes over the same track on a damaged surface again and again, wear grows exponentially.

Rule: a mechanical drive with a physical symptom → power it off and do not fire off any more reads.

2. HDD Regenerator on a physically damaged drive

HDD Regenerator (and similar tools) work on the principle that they rewrite the bad sector with a specific magnetic signal and force it to reallocate. It works for magnetic degradation and logical corruption.

It works when: the drive is physically healthy and the magnetic recording has merely degraded slowly (old, little-used drives).

It harms when: the drive has a mechanical fault. The tool will push the head across the damaged surface, the heads get even more damaged, and the platters get scratched.

3. Low-level format on a drive with important data

Modern drives do not allow a low-level format via the OS (it is only done in the factory). What is occasionally labelled as LLF (e.g. HDD Low Level Format Tool) is in reality a zero-fill — it overwrites the entire drive with zeros. Your data is gone.

Rule: if there is data on the drive, no format, no zero-fill, until it is backed up.

What is worth doing

Scenario A — the drive is physically healthy, it just has pending sectors

  1. Back up — first recover the data to another medium. Use a clone (ddrescue, HDDSuperClone, Macrium Reflect), not copies of individual files, to minimise the load on the drive.
  2. After backing up, you can try overwriting the drive with zeros (Gparted, DBAN) — the drive then reallocates the pending sectors to spares.
  3. Check SMART again. If the Reallocated Count has risen dramatically, the drive is at the end of its life → replace it.
  4. If SMART is clean after the overwrite, the drive can keep being used, but not for primary data without a backup.

A practical note on ddrescue: it works in two phases. First it reads all the good blocks in large jumps and skips the bad ones. Then it returns to the bad ones and tries to read them in smaller blocks, many times over. The logfile remembers where it has already been — so you can interrupt the clone and continue the next day.

Scenario B — the drive has a mechanical fault (clicking, grinding, not detected)

  1. Power it off and do not plug it in again.
  2. Open a clean chamber? Don’t attempt that at home. Dust 0.3 µm in size on the platter = more scratches.
  3. Bring it to us — diagnostics are free. In a clean box (HEPA particle filtration) we examine the heads, platters and firmware and tell you whether recovery is feasible — and what it would cost.
  4. Bit-level reading (an image clone) is done with specialised hardware (PC-3000, DeepSpar). An ordinary PC would only damage the drive further.
An open 2.5-inch laptop hard drive seen from above, with the platter, spindle, head stack in the park position and the voice coil motor magnet visible.
A 2.5" laptop drive with the top cover removed. The platter is smaller than on a desktop 3.5" drive, but the principle is the same — the head flies over the rotating surface at a recording density of several Tbit/in². A weak head generates pending sectors before the user even notices.

When the head is too weak to clone or is clicking, the drive is opened in a laminar-flow box and the head stack is replaced. That is now hardware work — the donor drive must be the same model and firmware revision, otherwise the heads won’t settle onto the tracks.

A technician in a blue ESD glove holds a four-arm head stack pulled out of a 3.5-inch hard drive, with the flex cable and preamp chip visible.
A removed head assembly from a 3.5" drive, held in an ESD glove. Each arm carries a slider with a read and write head. The replacement is done in a cleanroom, and the donor stack must come from the same model and firmware revision — otherwise the heads won't settle onto the tracks.

Scenario C — an SSD with defects

An SSD has no mechanical faults, but flash cells can „wear out". The SMART attributes are different:

  • Wear Levelling Count / Media Wearout Indicator — how much lifespan remains
  • Uncorrectable Error Count — unreadable blocks
  • Total NAND Writes — total writes

If the SSD is showing blocks that won’t read:

  1. Don’t write anything more to it.
  2. Clone an image onto a healthy drive using tools that can read through errors (ddrescue, HDDSuperClone).
  3. An SSD with exhausted cells sometimes never powers up again once it has been switched off. Don’t power it off until you have an image.
Close-up of the printed circuit board of an ADATA SATA SSD with two NAND flash chips, a Maxio controller and a SATA connector.
The insides of an ADATA SU630: two NAND flash chips in the middle, a Maxio controller bottom-left, the SATA connector at the bottom. With SSDs we don't talk about "bad sectors" but about unreadable NAND blocks — and when the controller fails, the data in the memory is still there, just only reachable via a chip-off on specialised hardware.

Why backing up solves all of this

None of the procedures described above is needed if you have current backups. Bad sector → I replace the drive → I restore from backup → done.

The 3-2-1 rule: 3 copies of the data, 2 different types of media, 1 copy off-site. How we back up data for businesses.

If you have no backup and you are at risk of losing data, stop working with the drive and contact us before you try anything yourself. As a rule: the first recovery attempt is the easiest, and every further one is more expensive and has a lower success rate.

Frequently asked questions

What should I do when a drive has bad sectors?

Use CrystalDiskInfo to find out whether the Pending or Reallocated value is higher than 0. If the drive does not click and connects normally, the first step is a clone via ddrescue onto a healthy medium. If the drive clicks or won’t connect, power it off and contact a recovery service — further read attempts lower the chances of recovery.

Does CHKDSK help with bad sectors?

It helps with logical corruption of the file system (e.g. after a power loss). It does not help and often harms with a physical fault — chkdsk /r reads the entire surface, so on a drive with a mechanical symptom it prolongs the suffering by 6–24 hours and damages more tracks. The rule: only run chkdsk on a drive that has no Reallocated > 0 in SMART and that also does not click.

When is it already too late and not worth recovering the drive?

From a technical standpoint it is almost never „too late" — even from severely damaged drives, something can usually be pulled out in a cleanroom. From an economic standpoint the line is where the cost of recovery exceeds the value of the data. For an ordinary family photo collection that is typically 20 000–35 000 Kč for a standard case; cleanroom openings and SSD chip-offs are more expensive. Our diagnostics are free — then you decide knowing the exact situation.

Can I replace the heads at home following a YouTube tutorial?

Technically yes. Realistically no — without a laminar-flow box, dust settles on the platter within a minute and the heads crush it on the next spin-up. The second problem is firmware pairing: two otherwise identical 2.5” head sets can differ in adaptive parameters that are written in the service area of the original drive. Loaded heads without the matching adaptives cannot follow the track. All of this is handled by the PC-3000 — without it the odds are slim.

Do you suspect bad sectors and have important data? Data recovery · Contact us · +420 775 556 063 (24/7)

See also: What to do when a drive clicks · Overview of CMR and SMR drives · How to send a drive by post