Solid state drives stop without a sound, and owners find that harder to accept than any amount of ticking. Working on Thursday, invisible on Friday, no warning in between. SSD hard drive recovery is a separate trade from disk work, with no platters, no heads and no clean bench involved, and it runs here daily for households, for the software and engineering firms along the A329M, and for a town that has been building computers since the 1960s and still has the business parks to prove it.
Anything sent in for SSD recovery is diagnosed free. The written figure that follows is settled before a screwdriver leaves the drawer: £300 + VAT for any single hard drive or SSD, NVMe included.
No fix, no fee covers logical recoveries. Outside it sit electronic and mechanical failures, chip-level work, DVR jobs and forensic jobs, and physical work is 50% up front. Every band is published on the data recovery cost page.
Reading a symptom back to the failure underneath it is where the work genuinely starts, and this set of thirty accounts for all but a handful of the parcels that reach the Guildford bench. If yours is not on the list, it will still be recognised — describe it on the phone and you will get a straight answer about the odds before you post anything.
This one deserves plain speech. On a working SSD with TRIM enabled, deleted files are usually gone permanently. The operating system tells the drive which blocks are no longer needed, the controller erases them in the background, and reads of those blocks come back as zeros afterwards. No laboratory anywhere restores an erased NAND page. If the deletion happened in the last few minutes, shutting the machine down immediately is the only thing that helps. If the computer has been running since, expect an honest answer rather than a hopeful one.
The management chip goes quiet and the drive drops off the bus, while every file sits precisely where it was written. What has failed is the part that speaks on the flash's behalf. Supplying that role from outside is what a firmware bench is for.
No manufacturer ever sold a product called that. It is a Phison controller stuck in its own boot loader with no working firmware above it. The label is wrong and the data behind the label is untouched. Service-mode work walks it back out.
Or 2MB, or nothing at all. Those are the sizes a controller reports when its translator has collapsed and it has fallen back on a default identity. Rebuilding the translator in service mode restores the real capacity, and the first job afterwards is a complete image.
End-of-life protection has stopped all writing in order to preserve what is already stored. On a desk that is a nuisance. On an imaging rig it is close to a gift, since the drive has lost the ability to make its own position any worse.
There is no warning noise, no vibration and very rarely a gradual decline. The drive simply stops between one session and the next. That is why the advice on this page never changes: stop using the machine before you ring anybody, including us.
Interrupt a write and the tables mapping logical addresses onto physical blocks can tear halfway through an update. A genuine mains spike takes out the power stages as well as the tables. Both arrive here regularly and neither is fixable in software.
Worn NAND returns one answer, then another, while the SMART attributes grumble quietly underneath. Only the first complete image is worth trusting, which is why it gets planned properly and taken once rather than attempted repeatedly.
Thin boards lose every argument with a screwdriver, an upgrade lever or a padded envelope with something heavy on top of it. A clean break out at the edge is a track repair under the microscope. Anything worse goes to chip-off, which is slower and reaches the same files.
The plastic connector breaks during an otherwise routine swap and lifts a few pads on its way. That is a board-level repair, after which the drive behaves as though nothing had happened. Board work is 50% upfront and sits outside no fix, no fee.
Early MZ-77E1T0 units reallocated sectors fast enough that Samsung published a firmware fix for them. Unpatched examples still turn up here years later. The V-NAND underneath reads perfectly well; it is the mapping above it that has to be rebuilt before anything can be copied off.
Those drives went into an enormous number of British machines during the years when everybody was swapping a hard disk for something quicker, so the whole wave is expiring together. The ending is abrupt: the drive leaves the bus mid-session and comes back as a fraction of its capacity or not at all. The controller has mislaid its translator while the memory is sound. Service-mode repair, then a full image.
Crucial's MX500 is well known for reporting a pending-sector figure that alarms owners more than the drive deserves, right up until the day that figure arrives with real read failures attached to it. Telling the harmless version from the genuine one happens on a bench rather than by guesswork.
A BX500 or an A400 has no cache of its own for the mapping tables, so the wear levelling is coarse and heavy use falls unevenly across the flash. Put one in a machine that writes all day and it ages in a fraction of the expected time.
When the translator on an SA400S37 or a similar low-cost SATA drive corrupts, the drive shrinks to a few megabytes or vanishes entirely. Service mode is where that gets rebuilt, and the memory sitting behind it is normally blameless.
These stop partway through a write and then decline to say anything further. A proportion of them wake up properly in service mode. The rest are read at chip level, which takes longer and arrives at the same place.
The Blue and Green SATA families have a documented habit of leaving the bus and staying away. The memory is rarely the culprit. It is a well-trodden route with a predictable ending, provided nobody has spent a fortnight power-cycling it first.
The manufacturer's tool halted partway through and left a drive that can no longer state what it is. Depending on how much the controller will still admit to, it is either reflashed on the bench or bypassed completely and read as raw flash.
QLC flash is fast until the pseudo-SLC cache fills, after which the speed falls through the floor. Owners read that collapse as a crash and pull the power out, and it is pulling the power that causes the actual damage.
Every read nudges the charge in neighbouring cells very slightly. On a drive full of material that gets opened often and rewritten never, those nudges accumulate until the error correction runs out of headroom. Slow repeated passes bring back a great deal more than one confident sweep.
Charge leaks out of unpowered cells, gradually and then not so gradually. A drive last used in 2021 can read as static. Patient imaging, with the error correction given the time it wants, returns far more than a single quick attempt would.
A drive that warms every morning and cools every night eventually fractures the ball joints beneath the controller and the memory packages. Reworking them restores contact for long enough to take one clean image, which is all that is being asked of them.
Enterprise and datacentre SATA drives carry capacitors that hold the drive up long enough to flush its buffers when the power disappears. Capacitors age. Once they have quietly failed, the next power cut takes the mapping tables with it and nobody knew the safety net had gone.
Very nearly every modern SSD encrypts on the way in, whether or not the owner ever asked for it, and the key lives inside the controller. Damage the key store and the memory reads perfectly while meaning absolutely nothing. We say that at the outset rather than billing hope by the hour.
A fault in the security state shuts an OPAL drive to everybody. The PSID printed on the label will unlock it by erasing every block first, which is the exact opposite of what anyone wants. It stays unused on this bench and it should stay unused on yours.
Two separate problems that people tend to merge into one. The drive is a hardware question and the key is not. With the recovery key to hand, an encrypted SSD is recovered like any other and unlocked at the end. Without it there is nothing anybody can do, so check the Microsoft account the machine was signed into before assuming the worst. Saving the key onto the drive it protects is a trap a lot of people fall into.
Several manufacturers have shipped firmware carrying a counter bug, the best-known example halting the drive dead at 32,768 hours of use. The hour arrives, the drive does not come back, and the flash inside is in perfect health. Service-mode repair, then an image straight afterwards.
Counterfeit drives wrap their writing around one small piece of real memory, so whatever fits inside the genuine capacity survives and everything after that was discarded as it was written. The recoverable portion comes back in full. The rest never existed in the first place, and no technique invents it.
An SSD with no free blocks cannot tidy itself, because shuffling data requires somewhere to shuffle it to. Write speed collapses, errors gather, and some drives switch themselves to read-only rather than carry on. Awkward to live with, and rather convenient at imaging time.
The ATA secure erase and sanitize commands are designed to be irreversible and they are extremely good at it. One that ran to completion leaves nothing for anybody. One interrupted partway often leaves a great deal. The free diagnostic establishes which of the two happened instead of guessing at it.
There is no fixed relationship on an SSD between the address your computer asks for and the physical location the data occupies. Wear levelling keeps moving blocks around the flash packages according to whatever the controller judges sensible at the time, and one record exists of where things ended up, held inside the controller's own firmware. Should the controller stop responding, or should that record become corrupt, the drive goes silent on the bus while every byte remains present in the memory, correctly written, with nothing left capable of pointing at any of it. Two honest ways forward exist. One approaches through service mode, a diagnostic channel the manufacturer built in, and persuades the controller to reconstruct what it mislaid. The other detaches the flash packages from the board, reads them on dedicated hardware, and rebuilds the translation in software, which is a slower business and is entirely indifferent to whether the drive ever responds again. Not one downloaded utility performs either operation, for the straightforward reason that they all require a drive Windows can already see before they will do anything.
All of it arrives here in the end: Samsung, Crucial, Kingston, SanDisk, WD, Intel, SK hynix, Kioxia, and the Apple blades that unplug, in 2.5-inch SATA, mSATA and M.2 formats. Which ones come round most frequently is basically a record of what builders and repair shops across Berkshire were fitting in a given year, so at any point the shelf tends to hold a Samsung 860 EVO or its 870 EVO successor, a Crucial BX500 or MX500, a Kingston KC600 or A400, a SanDisk Ultra 3D, a WD Blue SA510. One drive on that list earns a paragraph to itself, because a single characteristic failure on it brings more people to this page than every other model combined. An MZ-76E500 that had shown no symptoms whatsoever falls off the SATA bus during use and comes back reporting a lost or corrupt signature, a capacity measured in megabytes, or nothing readable at all. The cause is the MJX controller losing hold of its translator. Nothing has happened to the V-NAND underneath, and that difference is the entire point, since a drive with no working map has nothing coherent to offer even while every file sits exactly where it was originally written. Samsung published firmware revisions across that generation specifically to address it. When a drive has already deteriorated past the point of accepting one, the translator is rebuilt via service mode and a complete image is captured before the drive is asked to do anything further. Two things are turned away here, and saying so before you buy a stamp is fairer than saying so afterwards: flash soldered to a mainboard, and mobile phones or tablets, none of which this service covers. Aggressive TRIM and hardware encryption both make a job more involved than it appears at first, and that is explained at assessment rather than appearing later on a bill.
A mechanical disk tells you it is unwell. It ticks, it slows down, it needs a sector requested twice, and an owner paying attention can normally extract a few days of grace from those signals. A solid state hard drive supplies nothing comparable, since there is no moving component in it able to make a noise. Either the controller replies or it does not, and the transition between those two conditions occurs between one ordinary Tuesday and the next. It explains why solid state hard drive data recovery so often arrives with an owner adamant that the drive worked perfectly the day before. They are usually right. Memory that has genuinely worn out does exist, but it is seldom the explanation. In a consumer drive, a power interruption, a defect in the firmware, or a controller that has simply stopped answering accounts for a missing translator far more frequently than spent write cycles ever do, and spent write cycles are the failure everybody anticipates and the one this bench encounters least. Whatever the cause proves to be, establishing it costs nothing and takes two working days from the parcel landing, and the figure for one SSD is £300 + VAT, which is also the figure for an M.2 or NVMe module.
Solid state work is firmware repair and chip surgery from the first hour, so the bench is equipped for that and nothing else:
Brings a silent drive up in the manufacturer's own service mode, repairs the firmware, rebuilds the translator and takes a complete image off before the drive is asked to do a single thing more.
Where a drive stalls, the timeouts are enforced by hardware and the retry pattern is chosen for the unit on the bench rather than copied out of a manual written for something else.
SATA, mSATA, M.2 in all four lengths and Apple blade modules, each with a proper connector rather than a hopeful USB caddy. Active cooling is fitted for the drives that turn temperamental once they warm up.
Where nothing will wake the controller, the packages come off the board and each one is read on its own. Slower by a distance, and it arrives at the same files. Chip-level work is 50% upfront.
Interleave, XOR and error correction solved from raw dumps until the mapping exists again in software and a file system can be stood back up on top of it.
Cracked boards and burnt power stages are repaired before any imaging starts, because a drive that browns out halfway through a capture has gained nobody anything.
The Samsung EVO line sets the tone of this page. The 850, 860 and 870 went into machines by the thousand along the M4 corridor during the years when replacing a hard disk with something faster was the standard upgrade, and that whole decade of upgrades is reaching its end together. The usual ending is a translator failure sitting above completely healthy V-NAND, and translator failures recover. Around them sit an A400 that has mislaid its map, an MX500 worrying about pending sectors, a BX500 aged well before its time, and a WD Blue that left the bus without a word. Every removable format is workable here, whether that is 2.5-inch, mSATA or M.2, and that includes drives which have already shut themselves into read-only. One SSD is £300 + VAT with a free diagnostic ahead of it and 2 to 4 working days once the work begins. The two refusals are flash soldered onto a logic board, and phones or tablets.
Nearly every job here arrived as a parcel. Tracked, insured post is the calmest way to move a drive that is already struggling, and something handed over in Berkshire, Surrey or London is normally on the Guildford bench the next working day.
Is the storage still bolted into a machine — laptop, tower, iMac, MacBook, rack server, a DVR under the till? Free it first and send the bare unit. Stripping hardware is not something this lab does, though it is ten minutes' work for any repair shop on your high street. There is a single case with no way round it: memory chips soldered flat onto the mainboard, which is how Apple Silicon machines and certain ultrabooks are built. Where the storage cannot be unbolted, there is no parcel to make up.
↓ Print the shipping & booking-in form (PDF)
The name on the parcel wants to be Guildford Data Recovery. Driving it over from Bracknell is roughly forty minutes on the A322 then the A3; posting it costs you a stamp and a day. Either way, a message goes out to you as soon as it is logged onto the system, and two working days later the diagnostic is finished.
Unsure whether something should go in the box? Ring 0800 689 0668 while the lid is still open, or work through the free online diagnostic and let it tell you.
Nothing to pay for the diagnosis, one written figure before any work begins, and the band on this page is £300 + VAT for any single hard drive or SSD, NVMe included.