SSD storage and NAND retention?

V

Valnar

Supreme [H]ardness
Joined
Apr 3, 2001
Messages
4,302
I've seen a few articles this like over the years, but it didn't pertain to me until I had a collection of SSD's. I'm now at a point where I have some rarely used computers with an SSD in it, and even some SSD's sitting in a drawer waiting for their next computer build.

The question is not how long the data will stay intact. I think I've asked that before and nobody knows the answer. The question of the day is...how long do I need to fire them up for to refresh "that which can deteriorate"?

In other words, if I wanted to pull out my unused SSD's once a month, or fire up my rarely used computers once a month, for HOW LONG do they need to run sitting there in order to prevent the NAND from decaying? Is there something going on in the background while electricity is flowing through it?

I'm having a hard time wrapping my head around the concept that they need to run in order to save the data written on them. I mean, unless I'm actually writing something new, aren't certain bits just stagnant for years anyway, like my Windows system files?
 
The question is not how long the data will stay intact. I think I've asked that before and nobody knows the answer.
Click to expand...

I expect that this will most likely be the answer. Theoretically retention depends on the nand and its wear level. Also I think JEDEC publishes some specification for minimum retention but I have not looked at that in years.

Edit: Here is some info
https://www.micron.com/about/blogs/2015/may/addressing-data-retention-in-ssds

Is there something going on in the background while electricity is flowing through it?
Click to expand...

Yes. The drive actively detects and prevents this by refreshing the cells when it detects that it is needed.
 
Last edited:
I can not answer that one (without guessing). I would expect that it varies from model to model depending on the firmware and controller used. It may also be dependent on what type of flash is used and other factors like wear.
 
An SSD which isn't close to being worn out should hold its data for years.

You might be able to scrub the drive by simply *reading* the entire drive. This assumes that the drive is designed to rewrite blocks which have excessive (correctable) errors.
 
The quick and dirty is that SSDs will hold their data for years without power.

How many years depends on the drive, the tech used in it's actual chips, and the temperature in the storage area. If it's anything important, plug the drive in once every two years for a while and you should be fine.

That being said...backup, backup, backup.
 
I was told roughly 10 years by an engineer at work. I think it was nand and not nor....
 
Then why do they say you should keep SSD's running and not powered off? Surely there is a reason regarding retention?
 
I do a alphabetical sort defrag once a year to turn the soil on my active SSDs. Basically everything gets lifted and re-written once a year.
 
From my first link JEDEC standard for consumer drives is 1 year if the drive is close to its wear limit and longer if it is not. For enterprise drives the retention is reduced to 3 months after the wear limit. Most likely enterprise customers don't typicality pull there SSDs and put them on a shelf for years.

Similarly, for a client SSD, JEDEC set the data retention spec at one year, which allows the user much more time to go back to get data from an unused device, if needed. Again, keep in mind that this one-year spec only applies to the end of the SSD’s lifetime. When the SSD is new or of modest age, measured in TBW, the data retention is much longer.
Click to expand...
 
Last edited:
The link in my first post for one.

But that's the question. Perhaps I'm being too long winded...

Is there any difference in data retention on an SSD whether it is powered on or not?

It sounds like the answer is yes, otherwise active running SSD's in datacenters would go kaput in the middle of the day. So electricity flowing through them "refreshes" the cells, for lack of a better term. My question is, how long does power need to go through them to reset whatever-the-hell that thing is which causes them to deteriorate? And does it work for static data or only written refreshed data? Nobody knows. And yes, I realize a backup is a good idea.

Here is one tiny example. I built this MAME emulator with an SSD but I don't exactly use it very often. How often do I need to boot it up?...And for how long?...in order to prevent NAND decay?
 
If you're very concerned about long term data retention on a powered off SSD, then you'll want to stay away from TLC drives. I mean, I'm sure a new TLC drive likely will get you years of cold storage, but a new MLC drive should be fine for 10+, and a new SLC drive will surely have data retention periods measured in centuries. I remember reading something quite a few years ago where a company estimated their NAND chips should hold data in an unpowered state for something like 1500+ years.

It's fairly simple to picture. Electrons will slowly escape from the NAND gate as the drive is unpowered, and as the electrons escape, the charge in the cell changes. If enough electrons escape, the charge has changed to a different value, so the data is now corrupt. SLC/MLC/TLC is the number of bits (values) each cell can hold. SLC is single bit/2 values, MLC is 2 bits/4 values, TLC is 3 bits/8 values. As you can imagine, as the number of bits increases, the voltage differential between each value decreases, so it's much easier to get corrupted data.

If you have 2 drives with identical gate sizes made of the same semiconductor, then a TLC drive will in theory have an endurance 8 times worse than a similar SLC drive. In practice, it's not nearly that simple, and the endurance for each type is pretty much an order of magnitude worse than the previous one. So an SLC NAND might have a 100k P/E endurance, MLC 10k and TLC 1k. Those aren't exact, but you get the idea.

One important thing to remember is that the relationship between data retention and P/E endurance is not linear (more like exponential), so this is how you can get 10+ year data retentions on fresh drives, but only a few months or even days once you've worn the drive down to near failure. As the gates become more and more damaged from use, it's becomes easier for the electrons to escape, plus they're not as good at storing the correct charge in the first place, so even a tiny change in the stored charge will corrupt the data.

Personally, I've seen Vertex 2s, the SSD with the worst failure rate in history, last 3+ years in storage and boot up just fine. And these were drives that were beaten to a pulp for years before being chucked in a PC that just sits on a shelf. Would be interesting to try this with an 840 Evo, since they had all those problems with cell charge degradation. Might have to chuck the 840 EVO I'm using now in a closet for a year or so after I upgrade and see what happens.
 
You must log in or register to reply here.
Back
Top