PM893 MZ7L3 · PM9A3 MZQL2 · PM1643 MZILT · 870 EVO MZ-77E · 980 Pro · 3B2QGXA7
Samsung SSD members drive recovery. The data-centre drives, and the consumer ones that got into the rack.
Samsung's data-centre SSDs are the PM series: PM883 and PM893 on SATA, PM9A3 and PM1733 on NVMe, PM1643 on SAS, under Samsung's name and under Dell's, HPE's and Lenovo's. Its consumer drives get into server sets too, and bring their faults with them: the 870 EVO drives from January and February 2021 batches that developed uncorrectable errors, and the 980 Pro on firmware 3B2QGXA7 that went read-only until 5B2QGXA7 fixed it. Controllers that answer are imaged at once; controllers that do not are worked at chip level where the model allows. One Samsung SSD imaged is £300 + VAT after the free look, fixed in writing, 3–4 days at the bench.
Rather talk it through? An engineer answers the bench line
0800 6890668
Models we see, family by family.
Which kind of drive is it →| Family | Model strings | Interface, capacity and notes |
|---|---|---|
| PM SATA | MZ7LH… (PM883), MZ7L3… (PM893) | 960GB–7.68TB |
| PM NVMe | MZQL2… (PM9A3), MZWLJ… (PM1733) | U.2 and E1.S |
| PM SAS | MZILT… (PM1643, PM1653) | 12Gb/s SAS |
| Consumer | 870 EVO MZ-77E…, 980 Pro MZ-V8P… | In cheap sets |
What tends to go wrong on a Samsung SSD.
Samsung SSD members symptoms, and how long each gives you.
Not listed? Describe it on the form →What the message means on a Samsung SSD.
Describe yours to us →| What you see | The usual reason | Where that leaves you |
|---|---|---|
| Several failed within minutes | A firmware clock | Stop; check the survivors' hours; assessed at controller level |
| Shows 0 GB | Controller halted | Assessed at controller level; model permitting |
| Predictive failure; wear at 100 per cent | Endurance spent | Image now |
| Failed after a power event | Mapping table inconsistent | Controller-level work where supported |
From the parcel arriving to your files going back.
Work we have closed →Logged the day it lands, and the first look costs nothing Free
A case number goes on the parcel and a number on every drive the day it is opened, matched to the slot you wrote on it. An engineer settles what has actually happened before anything spins: the interface, the sector format, the firmware, the security state, and what the controller was saying when it stopped. Back to you come two things together: a straight note of what is liftable and what is not, plus one figure, fixed and written down. Accept it, or decline and owe us nothing.
The drive on its own bench
A member drive is read on the equipment its interface needs: a SAS imager for SAS and NL-SAS, a SATA imager for enterprise SATA, a terminated SCSI or Fibre Channel controller for the legacy drives, a PCIe adaptor for NVMe. It is never put back in a server, and never in a controller that will try to rebuild. Drives that answer at full speed are imaged; drives with weak or failed heads go to the clean bench, where matched donors are fitted and the service area repaired before a sector is read.
Imaged once, at its native sector size
Every drive is imaged sector by sector, head by head where it needs it, with the bad areas last. Drives with 520- or 528-byte sectors are imaged as they are and the extra bytes stripped from the image, never the drive. Self-encrypting drives are unlocked on the imager with the key you supply. From then on the originals are not touched again.
The image, and the set
Where the set is healthy and one member was the job, the image goes back to you on a fresh drive, sector-identical, ready to rebuild from. Where the set is the job, every member is imaged and the array reassembled in software from the images: order, chunk size, parity rotation and the reshape point read from the drives' own metadata, the file system repaired on the virtual volume, never on the originals. The array work is covered in full on our RAID array site, and it happens on the same bench.
You see the file list before you pay
What was recovered is listed for you first, and only then does a bill exist. Approve the list and it is invoiced; turn it down and it is not — and where nothing has come back, most jobs carry no charge at all. Recovered data travels home on fresh media bought in for your job, with the postage at our end. Your case is not closed until you have opened the files on a machine of your own.
From the bench
- An SSD that reads should be imaged now. Controllers fail without warning and without a second chance.
- Do not update firmware on failed drives. The fix prevents the failure; it does not undo it.
- The free look says what is possible, and on some firmware failures the honest answer is nothing. It costs nothing to hear it.
What helps, and what harms.
Do this much first
- Stop, and read the power-on hours on every survivor
- Label the slots and send the failed drives
- Tell us the model strings and firmware revisions
- Send the survivors too if the set has failed
What sets us back
- Updating firmware on failed drives
- Re-adding a failed drive to see
- Rebuilding onto survivors on the same clock
- Assuming a drive that reads today will read tomorrow
Questions answered before you commit.
My 870 EVOs in a RAID are throwing uncorrectable errors. Can the set be recovered?
Usually, if the controllers still answer: each drive is imaged with its bad blocks last and the set reassembled from the images. Replace them with data-centre drives afterwards.
Do I send one drive or the set?
One drive where the set is healthy and you want it imaged for the rebuild; every member, labelled by slot, where a second drive is flagged or the set has failed.
What does it cost?
£300 + VAT for one drive after the free look; a set of two to four is £500 + VAT upwards, fixed in writing.
How long does it take?
3–4 days at the bench for one drive; 5–10 days at the bench for a set.
Begin here if yours is doing the same thing.
Nothing gets worse while it is powered down.
Looking at it is free. Back comes a list of what opened and what did not, together with a single price to finish, set down in writing while you are still free to say no. On most jobs an invoice only follows the data. Until that list reaches you, leave the server off and the drives in their slots.