SSD's tally for data retention on moving data?

Elixer

Lifer
May 7, 2002
10,371
762
126
Since a SSD is supposed to either refresh a NAND cell, or move the data every X days, does anyone know how that figures into the writes a SSD has?

I assume that this isn't counted by the 'Total Host writes' in the SMART parameter, so, is there another counter someplace that keeps track of this? I don't think the SSD controller would do it that often, but, I would be curious to know just how often that happens. Is it only when the NAND cell is constantly failing ECC checks, or something else?
 

Glaring_Mistake

Senior member
Mar 2, 2015
310
117
126
Not a counter just for rewrites if that is what you're after but some drives do have a counter where rewrites are included, what that counter is called varies however.

I've seen Total Erase Count, Total TLC Write Count and Total Flash Write Count being used for drives using the SM2256 as a controller.
And Total Data Written, Total Flash Write Count and Total GB Written to NAND for drives using the SM2246 as a controller.
So if it is using a controller from Silicon Motion you should probably look out for the word Total.
Also something I've noted is that its Host Writes sees a small increase (like two units) just before starting to perform rewrites.

With Marvell controllers I've seen FTL Program Page Count likely being that and some attribute that probably is something like Total Erase Count.

As for Samsung's drives they may perhaps not include it but instead just have Host Writes and Wear Levelling Count.
Same for drives using Phison S10 as a controller (and at least the Trion 100 doesn't even give you that much).


Regarding how often it is needed to perform rewrites depends on the drive and the conditions.
Remember that I've mostly tested how drives recover so you may get different results if you don't let it get that bad before letting them recover.
But it has varied from the drive just needing a few minutes to do rewrites to needing to do rewrites almost every day.


Don't know what exactly triggers rewrites but it may be that it decides to do it when the ECC reaches a certain point.
Have seen a drive that seems to have improved read speeds for some files using algorithms (or a limited amount of rewrites) up to around 300-400MB/s which it deemed sufficient and then after a few days went down to around 20MB/s to make sure everything was ok and then perform rewrites.

It is also possible that they may instead check for voltage drift rather than checking for if the ECC starts to struggle since I've seen the ECC/algorithms cause read speeds to drop and the drive not seeing it as necessary to rewrite anything since the drop is not really due to voltage drift.
Though since that is kind of an odd issue who knows what is going on?
 

Elixer

Lifer
May 7, 2002
10,371
762
126
Shame there isn't a standard for this stuff, I am guessing there must be some way to talk to the firmware directly and get access to a lot more info than what SMART shows.

I bet that the data recovery guys would know about some custom ATA commands that they use to talk to the SSD at a lower level, but then again, I bet that is all different as well, and most likely under NDA. :(

Had a old small Toshiba SSD that has been sitting unpowered for a few months, and it was really, really slow.
I then just rebooted the machine, and left it on the BIOS screen for around ~5 hours while I did some other stuff, came back and let it continue to work, and this time around, the speed was back to normal.
None of the main SMART stats changed, but, I am guessing it had to go through a refresh and or move things around, but, no way to know what it actually did sitting idle.
 

Glaring_Mistake

Senior member
Mar 2, 2015
310
117
126
Had a old small Toshiba SSD that has been sitting unpowered for a few months, and it was really, really slow.
I then just rebooted the machine, and left it on the BIOS screen for around ~5 hours while I did some other stuff, came back and let it continue to work, and this time around, the speed was back to normal.
None of the main SMART stats changed, but, I am guessing it had to go through a refresh and or move things around, but, no way to know what it actually did sitting idle.

That sounds a bit odd.
In my experience not even their TLC drives would be that affected by voltage drift after a few months unless they either have seen some wear or been exposed to cold/heat during write/storage.