Someone a while back put a set of very cheap SSDs through a torture test, and after exhausting many of the write cycles, left them alone for months. When powered back on for reads, the drives were slow as the error correction hardware was working overtime to compensate for the loss of trapped charge over time, but mostly recovered their performance after a while.
That said, I have a similar anecdote where one of my very worn test bench SSDs kept complaining about the same bad sectors despite OS reinstalls until I just overwrote it with zeros using dd. Was fine for many months thereafter.
I’ve no idea either if SpinRite has some secret sauce that FOSS utilities have yet to replicate, but it sounds like a non-destructive read-write test with badblocks ought to do the same.
Also, my CF cards, SD cards, or USB drives from the early 2000s and early 2010s almost never give me trouble despite spending years unplugged. More recent flash memory is a different story though and I suspect the shrinking gate sizes and advent of TLC/QLC/PLC haven’t helped. I’ll usually splurge a little these days to get the industrial or high endurance MLC flavors and hopefully avoid the issue.
Good to hear that dd did the trick for you, and I will keep that one in mind the next time I need to re-install an OS on a drive that might be having the issues. SpinRite works by going sector by sector reading and copying the data, erases the sector, and then writes the data back. Not like something you would want to do all the time for sure. The creator of the program also said he will have the code go FOSS when he dies (pretty old but in good health). I am hoping that he is able to complete the full re-build he has in the works to have it work with how modern motherboards and controllers work (the current code-base is extremely old and requires turning off both Secure Boot and UEFI (or at least enable CSM) in order to boot the live USB (I think it still uses FreeDOS as the environment). There are ways to run it in a VM correctly, but I haven’t messed with trying that yet.
I also agree that it seems that really old cards and drives tend to not have the same issues as newer ones. Hell, I have even found that my floppies that are from the mid to late 90s and very early 00s tend to have a better chance of me still getting data off them. More of my mid to late 00s disks have issues with corrupted data. Found that out when going through a bunch of disks I had forgotten about, to get whatever I could last year.
Someone a while back put a set of very cheap SSDs through a torture test, and after exhausting many of the write cycles, left them alone for months. When powered back on for reads, the drives were slow as the error correction hardware was working overtime to compensate for the loss of trapped charge over time, but mostly recovered their performance after a while.
That said, I have a similar anecdote where one of my very worn test bench SSDs kept complaining about the same bad sectors despite OS reinstalls until I just overwrote it with zeros using
dd. Was fine for many months thereafter.I’ve no idea either if SpinRite has some secret sauce that FOSS utilities have yet to replicate, but it sounds like a non-destructive read-write test with
badblocksought to do the same.Also, my CF cards, SD cards, or USB drives from the early 2000s and early 2010s almost never give me trouble despite spending years unplugged. More recent flash memory is a different story though and I suspect the shrinking gate sizes and advent of TLC/QLC/PLC haven’t helped. I’ll usually splurge a little these days to get the industrial or high endurance MLC flavors and hopefully avoid the issue.
Good to hear that dd did the trick for you, and I will keep that one in mind the next time I need to re-install an OS on a drive that might be having the issues. SpinRite works by going sector by sector reading and copying the data, erases the sector, and then writes the data back. Not like something you would want to do all the time for sure. The creator of the program also said he will have the code go FOSS when he dies (pretty old but in good health). I am hoping that he is able to complete the full re-build he has in the works to have it work with how modern motherboards and controllers work (the current code-base is extremely old and requires turning off both Secure Boot and UEFI (or at least enable CSM) in order to boot the live USB (I think it still uses FreeDOS as the environment). There are ways to run it in a VM correctly, but I haven’t messed with trying that yet.
I also agree that it seems that really old cards and drives tend to not have the same issues as newer ones. Hell, I have even found that my floppies that are from the mid to late 90s and very early 00s tend to have a better chance of me still getting data off them. More of my mid to late 00s disks have issues with corrupted data. Found that out when going through a bunch of disks I had forgotten about, to get whatever I could last year.