13

I am planning to stock on SSDs but I will not use them right away. How long can they be kept in storage and what factors play in the degradation of SSDs? ( M.2 NVMe, SATA SSD, U.2 SSD).

Clarification: I am not asking about how long does data last on SSD if left unused, but rather the longevity of the SSD (M.2 NVMe/SATA SSD/U.2 SSD) if bought and stored right away to a shelf.

15
  • 1
    Unused storage device do nothing expire
    – Ramhound
    May 8, 2021 at 17:45
  • 14
    To what end are you stocking up? Unless you've found some unbelievably cheap source, then anything you buy this year will be either superseded by next year, or cost half as much. idk details, but much 'flash-style' storage needs periodic charge to avoid losing integrity.
    – Tetsujin
    May 8, 2021 at 17:46
  • 5
  • 1
    @harrymc Keeping them with data is a very different question than keeping them with the expectation of writing new data later.
    – pipe
    May 9, 2021 at 2:53
  • 3
    They can remain unused forever, as long as no one ever plugs them in (or uses them as a paperweight). Presumably you'd like to know how long can they go unused and then be used successfully. May 9, 2021 at 6:31

3 Answers 3

19

If you are talking about just storing them for later with no data on them, they should be fine for a long time.

If you are talking about putting data on them and expecting to get the data back, I believe the JEDEC spec only stipulates that a consumer grade SSD must retain data unpowered for at least a year. So if you expect to get the data back, make sure you power them on for a few hours once a year and run a full set of built in self tests.

3
  • 4
    Is this fully true? Brand new SSDs are still preformatted with the FTL metadata. If you store the SSD "unused" for a few years and that metadata deteriorates, then...? Can you reformat an SSD to rebuild the RTL from scratch?
    – d3jones
    May 9, 2021 at 13:47
  • 4
    @d3jones The good ones have firmware that will repair it’s own FTL if it runs into errors in unused blocks. May 9, 2021 at 15:24
  • 1
    Also note customer versus enterprise here. SSDs designed for customers generally do more best-effort recovery than SSDs designed for data centres. (Why? Best-effort recovery both a) is often slow and b) leads to a higher risk of silent corruption / miscorrection of the data. Data centers are less tolerant of latency spikes or silent data corruption (they are more likely to have a backup 'ready to go', and e.g. missing an SLA can be a major problem) whereas a customer generally prefers best-effort recovery)
    – TLW
    May 10, 2021 at 0:03
7

If you are talking years, not many decades, the SSDs do not deteriorate. This is the same for memory modules, and other electronic components like resistors and non-electrolytic capacitors.

They may not keep data forever (never tried) but they will last for many years. Without data, SSDs will last a very long time.

2
  • This is only true if the SSD has measures in place to recover from a bad FTL. Which is nowhere near guaranteed (among other things, best-effort recovery methods to recover from a bad FTL make it more likely that the SSD inadvertently silently corrupts data instead, which is a big nono, especially in the enterprise space.)
    – TLW
    May 9, 2021 at 23:53
  • However, I would expect an enterprise grade device to be far more conservative regarding multilevel cells.... May 10, 2021 at 9:30
3

There is not much difference between the degradation of an SSD with or without data.

An SSD contains electronic components that will degrade over time and magnetic cells that will lose their magnetism over time. It's not the data that degrades, it's the memory cells that die with time.

Your question: At what state of decay does it cease being a drive altogether and "lock in" its unusedness for all time?

Answer: This is when either:

  • An electric component has failed, or
  • When the number of failed memory cells exceeds the number of spare cells used for remapping dead cells to spare ones.

You will find more data about the statistical failure rate of SSDs in the post Lifespan of an SSD (NAND Flash) for minimal write use archive purposes: Write once, toss in (proverbial or literal) storage closet.

7
  • 3
    This is only true if you ignore data retention. SSDs do lose data over time (the cheap ones can hold data without prophylactic access for about a year, the good ones maybe last 2-3) because the very design of flash memory makes permanent retention impossible, and this gets worse overall the more the drive is written to. May 9, 2021 at 15:23
  • 1
    @AustinHemmelgarn: Some cells can be revived with re-magnetism, which is why the firmware refreshes the cells continuously. The cells that cannot be refreshed get remapped. A disk shelved for a long time will be in an unknown state, where an unknown number of cells have failed. This might keep the firmware quite busy when power is restored. It might take some time for the state of the disk to finally be defined enough to answer the poster's question about its usefulness.
    – harrymc
    May 9, 2021 at 15:56
  • How often would I need to "refresh" data on an archive SSD - perhaps by performing a mass read? I assume running a trim command isn't going to be useful. Furthermore, if I have a live powered up SSD which contains large data areas that are never accessed, then my data is going to gradually die of bitrot? May 9, 2021 at 16:25
  • 1
    Powered up will not gradually die of bitrot, because the firmware keeps on doing house-keeping to prevent it by refresh and remap (if required). An archived SSD that is powered up for a few hours may be refreshed just by that, but a mass read can certainly help as verification. The best bet is to keep a second backup.
    – harrymc
    May 9, 2021 at 17:05
  • 3
    Flash memory is not magnetic, the memory cells store charge in the gate capacitance of a transistor. May 9, 2021 at 20:16

Not the answer you're looking for? Browse other questions tagged .