Nothing else in a computer stops this cleanly. An NVMe module gives no warning rattle and no slow decline, so the first sign of trouble is a machine that booted on Tuesday putting up an error on Wednesday where the drive used to be. M.2 sticks and PCIe cards of every make arrive from Liverpool households, gaming rigs, and the engineering and research offices dotted around the city.
Every nvme job is diagnosed free. One fixed figure follows in writing, agreed before a screwdriver comes out of the drawer.
No fix, no fee all jobs except electronic and mechanical failures, chip level work, DVR and Forensic jobs. Every band is listed on the data recovery cost page.
Matching the symptom to the fault underneath is the first job on any nvme, and after twenty-odd years these thirty account for very nearly everything that arrives.
Controller or firmware trouble emptied the bus overnight. The data behind that silence is complete. What it lacks is anything capable of managing it properly again.
Four candidates: a dead controller, a failed power stage, damaged boot code, or dead memory. Dead memory is much the least likely of them. The odds sit with recovery, and the module has no way of telling you so.
Translator collapse makes a module quote a figure no manufacturer ever printed on a box. Service-mode repair, and the signposts are clear.
The system module stopped mid-sentence. Lifted out of the machine and put on the bench, it usually hands over its files with very little argument.
Bare modules throttle first, then drop out under load, then scramble their tables after a run of hot afternoons. Imaged cold, a surprising number behave.
Windows Update or a vendor tool, it makes no difference. What is left is a module unable to say what it is. It gets reflashed here, or read around entirely.
Flash keeps no diary. Faultless service right up to the second it stops is the standard life story, which is why backups matter more on this page than on any other.
A bare module is a thin strip of glass fibre and it loses every argument with an upgrade lever or a stiff envelope. Small breaks are a track repair. Everything else goes to chip-off.
Surge damage to the power circuitry gets rebuilt at board level first, and only then does anyone attempt a read. Doing it in that order is most of the job.
Freed blocks start being erased the moment files are deleted, and the process only pauses when the power is off. Pull the module out and the clock stops where it stands.
Early SN850X firmware would stall under certain loads and then take the module off the bus altogether. WD published a patch. Plenty of these never received it, and they keep the booking sheet lively.
The 980 and 990 generation burned through NAND life faster than intended until an update corrected it. Modules that were never updated arrive with most of their endurance already spent.
Kingston's NV2 has shipped with whichever controller and flash were going that quarter, so two modules with identical labels can need completely different handling. What is under the sticker chooses the method.
Steam Decks and similar handhelds seal small modules into crowded shells with nowhere for the heat to escape. A tiny format, and an entirely conventional recovery.
DRAM-less modules keep their mapping in borrowed host memory, and one bad interaction is enough to corrupt what was borrowed. Service mode puts the ledger back in order.
A fast module with no heatsink runs hot and ages for it. Dropouts come first and disappearance comes later, and the imaging window sits in between.
Shader compilation and constant cache churn age a module by about a year each season. Treat the stutters as the notice period and the empty slot as the deadline.
Several budget designs borrow system memory to hold their mapping, and after sleep or a crash that borrowed copy comes back corrupt. The failure class is documented, and so is the way out of it.
Cheap modules hammer their fast cache region until it gives up, then stall or brick. The slower flash sitting behind it usually still answers without complaint.
Over-tightened heatsink screws and squashed thermal pads bow the board until joints crack. Rework under the scope comes first and the image comes after.
After a BIOS update, or in a marginal riser, module and slot stop agreeing terms and the drive appears to have vanished. That gets proved out in known-good lab sockets instead of by swapping cables at home.
A single failed die shrinks the reported capacity, throws errors across the stripes, or forces the module read-only. Imaging routes around the dead die and the surviving ones carry the data out.
Garbage collection interrupted halfway leaves the map partly rewritten, and the module then mounts, loops and sulks by turns. Service-mode reconstruction settles it.
Counterfeit modules get recovered for whatever they genuinely are rather than whatever the label claims. The sticker convinces nobody on this side of the bench.
A cracked socket latch costs the module lanes. It limps along at a quarter of the width it should have, then stops appearing at all. An intact lab socket is where it gets imaged.
One NVMe command reformats a drive to a different sector size and throws the whole mapping away in the process. It is fast, it is silent, and it leaves what looks like a blank module. Recovery starts at the raw memory and rebuilds the arrangement from there.
Both commands are meant to be irreversible and both are very good at it. A sanitize that ran to completion leaves nothing for anyone. One that was interrupted often leaves a great deal. The free diagnostic establishes which of those happened rather than guessing.
Aggressive power-saving settings have caused documented dropouts on particular module and platform pairings, Linux especially, where the module sleeps and never wakes. It looks fatal. Usually the hardware is sound and it is the tables that want rebuilding, because of the crash that followed.
A module with no free space cannot tidy itself, because shifting data around requires somewhere to shift it to. Write speed collapses, errors gather, and some modules go read-only instead of continuing. Awkward to live with, and helpful when it comes to imaging.
M.2 sockets carry different signals from machine to machine, and a module that fits mechanically may not be wired to anything it can talk to. Nothing shows up, nothing is broken, and an afternoon disappears into a fault that does not exist. Rule it out before writing the module off.
A module plugs into the PCIe lanes and speaks to the processor directly, with no motor, no arm and no platter anywhere in the path. That is the reason for the speed, and equally the reason there is no limping phase before the end. Inside, a controller runs a thick layer of firmware over stacked NAND, keeping track of where each block of your data has been moved to. Let either the controller or that firmware lose its place and the module goes quiet on the bus, while the contents sit in the memory behind the silence completely undamaged. Reaching them means one of two routes: the manufacturer's service mode, where a controller can often still be coaxed into a conversation, or reading the flash chips directly and rebuilding the translation layer from scratch where it cannot. Restarting the machine another twenty times helps neither route, natural though the urge is.
Anything that unclips and fits in a jiffy bag is workable: 2280, 2242 and 2230 sticks, add-in cards for the PCIe slot, Gen3 through to Gen5, wearing Samsung, WD, Crucial, Kingston, SK hynix, Kioxia, Sabrent or Solidigm labels. The standing refusal is flash soldered onto a logic board, which falls outside what this bench takes in. Temperature deserves a paragraph of its own. Run a bare module hard with no heatsink over it and it throttles first, then drops off the bus altogether, and enough of those episodes will leave its own mapping tables in a mess. Compact builds and thin laptops send a steady trickle of exactly that through every warm month. Whichever of these started your fault, the free assessment identifies it and puts a figure against it before you spend a penny.
Controller work, firmware work and chip surgery at PCIe speeds. The bench is specified so no job ever waits on equipment:
Silent modules come up in the maker's own service mode. Firmware is corrected, the translator rebuilt, and the image taken without anybody gambling with the memory.
A module that stalls or wanders gets captured behind timeouts enforced in hardware, with retry behaviour set at a level software has no access to.
Short 2230 stubs through to full-length 2280s, add-in cards through to enterprise U.2, every one of them connected natively. Modules that faint once warm get active cooling.
Where service mode cannot get a word out of the controller, the packages come off and are read one by one. The long way round to the same destination.
Interleave, XOR and ECC solved from the raw dumps until the mapping stands again, with the file system built back on top.
Power stages rebuilt component by component, cracked boards reworked under the scope, and nothing imaged until that is done.
Every reputation on this page was earned the hard way. SN850X firmware stalled under sustained load until WD issued the fix. Samsungs of the 980 generation that were never updated still arrive with most of their endurance gone. An NV2 is a lucky dip, and whatever is inside it picks the method for us. Gen4 modules fitted where somebody judged a heatsink optional cook themselves by degrees. Lengths from 2230 to 2280, add-in cards, enterprise U.2, and every generation from Gen3 to Gen5: all workable at £300 + VAT a module, on the one condition that it unplugs and can therefore be posted. The single refusal is NVMe soldered to a logic board. Phones and tablets are the other.
Almost every job on this bench came in by tracked, insured post. It is the calmest way to move a drive that is already struggling, and a parcel handed in anywhere on Merseyside usually reaches the bench the following working day.
Still screwed inside a laptop, tower, MacBook, iMac, server or a CCTV recorder? Take the drive out and post that on its own. We do not strip machines here, and any computer shop will do it in a few minutes. The one job nobody can take on is flash soldered straight to a logic board, as on Apple Silicon Macs and a handful of very slim laptops: if the storage will not unbolt, there is nothing to send.
↓ Print the shipping & booking-in form (PDF)
Address it to Manchester Data Recovery. It is roughly 35 miles from Liverpool along the M62 if you fancy the run, and next working day by tracked post if you do not. We ring or email the moment it is booked in, and the free diagnostic closes 2 working days after that.
Not sure what belongs in the parcel? Call 0800 689 0668 before you tape it shut, or step through the free online diagnostic first.
Diagnosis free, one figure written down, and £300 + VAT covers a single drive or SSD. Start online or ring.