It’s the common SATA SSD failure, and what's confusing about it is that all health indicators showed the drive as being in good condition. In reality, wear reporting and the actual failure of something on the drive were happening in two totally separate systems.
The Samsung 860 EVO will spread out your data over V-NAND chips, and use a translation layer to keep track of where each block physically resides on those V-NAND chips. If you lose that layer, everything in the physical storage remains intact but is now completely inaccessible via the interface.
The Samsung 860 EVO was (and remains) one of the top-selling SATA SSD drives in history; hence we've seen an enormous number of these units, and when they do fail, their failure mode(s) are consistent enough to document. In a typical instance where a unit failed, it worked perfectly well without any performance degradation, had zero SMART warnings, showed zero signs of slowdown, and then simply wasn't visible on the system.
What has happened is the MJX controller has ceased to function. It wasn’t the flash: the NAND itself is typically healthy with ample life left. The controller has either locked up or failed and taken the drive's ability to enumerate to the system with it.
That difference is important, since the data is nearly always physically still present on the drive itself. It's an issue with getting at it, not losing it.
Stop power-cycling it. The immediate reaction will be to try multiple ports and systems. Although a controller that occasionally comes online is far preferable to one that is dead and gone, repeatedly cycling power to your device can cause the controller to go into a state from which it cannot recover.
Do not run a firmware update. Samsung Magician will provide an option to do so. Attempting to flash firmware to a controller that is starting to fail can sometimes partially succeed and put it into a worse situation than it was in prior to flashing.
Do not initialise it. If it does show up in Disk Management as uninitialised, that prompt will destroy the existing partition layout. Do not accept it.
Check whether encryption was on. Both Microsoft's BitLocker hardware offload and Samsung's own encryption also fall under this category. Without a valid decryption key, simply reading out the NAND accomplishes nothing.
First, there is vendor mode: a Samsung engineering interface allowing examination and repair of both the drive’s firmware and translation tables; this will likely be the best option when the controller responds in some manner, because the drive reassembles its own data.
When there is no response from the controller, however, the NAND is accessed and the raw flash is extracted directly from the memory. After extracting the raw flash, we reverse what the controller did originally (wear levelling, interleaving, scrambling and error correction, all specific to the MJX controller and its firmware revision) to reassemble the data.
Reversing the controller’s original assembly process is the most difficult portion of the entire process. If a user attempts to read raw NAND without properly reversing the controller’s arrangement, they will receive nothing but “noise”. This is why simply inserting NAND chip(s) into a reader does not constitute recovery.
The issue appears consistently enough that we are able to identify this issue at first glance: complete loss of function without any warning signs whatsoever, on a drive whose flash memory is otherwise healthy. Samsung has not characterised this as defective.
Possibly, if the drive is still under warranty. Replacing it will give you a new drive and will not provide you with access to your data. Since warranty replacements consist of receiving a new drive in exchange for your failed unit, it is recommended to recover your data before sending your drive back for warranty replacement if the data matters to you.
No. Updates should not be attempted on a failing controller, as this can make things worse. The update process is intended for use with functioning controllers only.
Infrequently. When these units are received, they have considerable endurance remaining. It is the controller that has failed, not the memory.
From £300 +VAT, fixed in writing after the free diagnostic.