SSD read-only myth?

[Elixer]

I have yet to see any SSD enter the 'read-only' state, which supposedly it does after the NAND writes are used up, but, all evidence shows that this is far from normal, and I would say that less than 1% of SSDs will actually achieve that state.
One of the well known tests on SSD endurance clearly showed that all SSDs basically died out, instead of entering the 'read-only' state, and are more or less bricks. http://techreport.com/review/27909/the-ssd-endurance-experiment-theyre-all-dead

@Billy Tallis, since you do SSD reviews for AT now, have you ever seen a SSD actually get into a stable 'read-only' state, and not brick itself? By stable, I mean that the data is still valid, and the drive isn't giving back corrupted results.
Heck, has anyone ever seen a SSD do that? (If so, can you post a image from CrystalDiskinfo showing the stats of said drive?)

 

[DaveSimmons]

Given those results I'd never expect to see it happen in real-world use with drives from reputable companies. 200 TB for the weakest (Samsung with TLC) is 10+ years of hard daily use. The boot drive on my work PC where I do Visual Studio full rebuilds from source all day long (plus the OS pagefile and temp folders) was under 10 TB after 3 years.

 

[BonzaiDuck]

Everyone has tried to be thoughtful and cautious about how they deploy their SSDs. I'd been using small ones for caching slower devices, now caching SATA SSD to NVMe as well as from an HDD source.

There seems to be an opinion -- from very knowledgeable people here, no less -- that caching will hammer a caching SSD. No less -- I myself thought that it would accelerate the TBWs accumulating.

But it doesn't necessarily do that. Once all the things you habitually use are cached to the SSD, only changes to those things require more writes.

So I think all the chicken-little worries about "using up" SSDs are more myth than fact.

 

[mikeymikec]

I think the "read only" state largely depends on a very fortunate circumstance being that when the SSD used up its last writes, it tidied the file system sufficiently for it to be read without any form of maintenance required. NTFS is pretty resilient, but I don't think it was designed with a device in mind that "runs out of writes". However, a data recovery program should be able to read the entirety of the drive, dump its contents into a file on another storage device, mounts that as a file system in its own right, runs chkdsk on it, voila.

This is also assuming that an SSD actually survives to use up all its writes. How many of us have seen the electronics of a HDD die rather than a purely mechanical issue?

 

[coercitiv]

I have yet to see any SSD enter the 'read-only' state, which supposedly it does after the NAND writes are used up, but, all evidence shows that this is far from normal, and I would say that less than 1% of SSDs will actually achieve that state.

I have yet to see a consumer exhaust the NAND writes in a SSD.

Going back to the Techreport SSD endurance experiment, such a test does present us with a problem: it overwhelms the unit with writes, possibly preventing the drive from ever having a chance to lock up. However, the reviewers did note this:

The Corsair, Intel, and Kingston SSDs all issued SMART warnings before their deaths, giving users plenty of time to preserve their data. The HyperX's warnings ended up being particularly premature, but that's better than no warning at all.

Issuing repeated SMART warnings seems like an even smarter solution than waiting for the drive to give it's last breath and then giving you the opportunity to stare at it's read-only corpse

Techreport also noted that the Intel drive did enter read-only state, and their description seems to indicate it followed a precise per-determined behavior:

Intel doesn't have confidence in the drive at that point, so the 335 Series is designed to shift into read-only mode and then to brick itself when the power is cycled. Despite suffering just one reallocated sector, our sample dutifully followed the script. Data was accessible until a reboot prompted the drive to swallow its virtual cyanide pill.

 

[Elixer]

Techreport also noted that the Intel drive did enter read-only state, and their description seems to indicate it followed a precise per-determined behavior:

Sure, it entered the read-only state, then died on reboot.
How does that even make any sense, the SSD should have gone into this state, AND people should still be able to read data from it.

Coincidentally, Tom's hardware did a new test, http://www.tomshardware.com/reviews/intel-ssd-600p-nvme-endurance-testing,4826-2.html
Though, it seems Intel did fix issues in the firmware to prevent bricking on reboot.

Intel is one of the only companies that will lock the SSD before allowing the user to get into a dangerous state where data loss can occur.
Reputable endurance tests from media outlets show that some SSD manufacturers will allow you to write massive amounts of data beyond the endurance threshold--sometimes into the low petabyte range. However, the testers don't follow JEDEC testing standards and usually fail to factor in data retention during extended power-off periods. Consumer SSDs don't just have to keep writing data to meet JEDEC's requirements; they also have to retain the data for a period of time without power.

So, in the Techreport's testing, NONE of them met the JEDEC requirements of making the drive read-only and still be able to access it, and survive without power for at least 30 days.

 

[coercitiv]

Sure, it entered the read-only state, then died on reboot.
How does that even make any sense, the SSD should have gone into this state, AND people should still be able to read data from it.

It doesn't really make sense to me, but Techreport talked as if that was the default behavior.

 

[PliotronX]

Toms tortured a 600p and it bricked at 110TBW IIRC.

 

[Billy Tallis]

I have not yet worn out the endurance on any SSD. It's on my to-do list for this year.

My first sample of the 250GB Samsung 960 EVO died early in the testing process, and appeared to enter a persistent read-only state. The only file on the drive was the IOmeter test file, so I can't attest to just how much of the data was accurately preserved, but it was enough for Windows to recognize the filesystem and list the one file.

 

[BonzaiDuck]

I have not yet worn out the endurance on any SSD. It's on my to-do list for this year.

My first sample of the 250GB Samsung 960 EVO died early in the testing process, and appeared to enter a persistent read-only state. The only file on the drive was the IOmeter test file, so I can't attest to just how much of the data was accurately preserved, but it was enough for Windows to recognize the filesystem and list the one file.

I wouldn't doubt your diagnosis, and the fact the Windows recognized the file-system is something different than what happened to me.

But for some reason -- a system freeze with a game uncorrected for a known bug -- the 250GB Sammy 960 EVO threw a very brief "reset" message box on the screen, and wasn't recognized at all. I pulled it from the system briefly to test other things like the PCIE slot, etc. When I put it back in and booted up, it was there, tests "Good" by Magician and continues working.

I'm interested in reports such as yours, because I'm coming to a crossroads for spending some Franklins on either a Pro or EVO 1TB, and I'm taking my time with it to see how things work out with my EVO SSD-caching. "Work out" is a longer-term prospect, but "working" it is.

 

[dave_the_nerd]

I don't care.

I have had enough bad experiences with things that were supposed to fail gracefully or incrementally or failover and/or recover... (but didn't - big surprise!)

I just assume that at any given moment, lightning will strike and your computer will explode and burn your house down and your dog will pee on the ashes. Plan (and budget) accordingly.

My SSDs can fail any darn way they want, whenever they want, and it doesn't matter to me. I will be back up and running in a few minutes on another computer, and have the original back in proper form within 24 hours. Done.

 

[Elixer]

My main argument is, that lots of SSDs are not following JEDEC specs, and they should.
We do have standards for a reason, and the JEDEC specs clearly outline what should happen when the writes are used up.

 

[Donaldiniosan]

I have yet to see any SSD enter the 'read-only' state, which supposedly it does after the NAND writes are used up, but, all evidence shows that this is far from normal, and I would say that less than 1% of SSDs will actually achieve that state.
One of the well known tests on SSD endurance clearly showed that all SSDs basically died out, instead of entering the 'read-only' state, and are more or less bricks. http://techreport.com/review/27909/the-ssd-endurance-experiment-theyre-all-dead

@Billy Tallis, since you do SSD reviews for AT now, have you ever seen a SSD actually get into a stable 'read-only' state, and not brick itself? By stable, I mean that the data is still valid, and the drive isn't giving back corrupted results.
Heck, has anyone ever seen a SSD do that? (If so, can you post a image from CrystalDiskinfo showing the stats of said drive?)

Here you are: