What happens when an SSD wears out?

Since SSDs are still fairly new, there isn't a history of posts saying "I reached my writes limit on 10% of my SSD... – and here is what happened (I.e. I lost 3 directories of work files. The entire drive just died...). Maybe it starts churning as the available sectors are reduced way down - like a PC with low memory.

They tell us that these devices will only last a few years. That is frightening - THEN WHAT HAPPENS? No one seems to know. Is it "instant paperweight" or a few files here and there vanish? Or do the SSD utilities constantly monitor and warn you long before the drive dies?

Ken Roberts

Posted 2011-10-13T00:34:31.917

Reputation: 381

Answers

It's a bit of a myth that SSDs will wear out, especially for typical desktop use. (At least, this applies to SSDs from the past few years that have proper wear-levelling.) Even if you write 7GB/day to a decent 256GB 25nm-process SSD, the flash memory should last for tens, if not hundreds, of years. It's far more likely that the controller hardware or software will fail, given the number of failures that have been reported by manufacturers and users.

In theory, it is possible to read data even after all program/erase (p/e) cycles have been used up. In fact, the JEDEC specifies that data on consumer-grade SSDs should be readable for one year after all p/e cycles have been exhausted. So the likelihood of losing data due to the drive reaching the end of its lifetime is small; it's more likely that you'll have replaced or upgraded your system by then.

sblair

Posted 2011-10-13T00:34:31.917

Reputation: 12 231

What happens when an SSD fails? The same thing that happens when a mechanical hard drive fails: You recover from your backups onto a new drive, and continue on.

Sean Reifschneider

Posted 2011-10-13T00:34:31.917

Reputation: 1 387

I read that whereas Hard drives often fail completely, losing all of your data . . . SSDs are more likley to fail, losing "some" of your data. The author was predicting what happens when the drive begins maxing out on write cycles for a high number of sectors. Of course, a sudden hardware / interface failure is a different story. – Ken Roberts – 2011-10-13T14:48:26.633

1Sure, I understand what you're saying @Ken, but how recoverable your data is highly depends on what state the software is in when it can no longer write to the drive. I've heard people saying "SSDs are great because when they reach the end of their life they just become read-only, so you can mount them up and copy data off." That seems a bit optimistic... But also, it seems likely to lead someone to saying "I don't have to do backups now", which is absolutely wrong. IMHO, forget what happens when an SSD dies, make sure you have good backups. – Sean Reifschneider – 2011-10-13T16:22:16.630

1Oh, and FYI: The only SSD I've had fail so far started showing up with 8MB capacity. Note that this was NOT the Intel 320 series drive that has this as a well known failure mode with firmware fixes. – Sean Reifschneider – 2011-10-13T16:24:24.663

This article published July 2011 talks about SSD failure modes: http://www.zdnet.com/blog/bott/ssds-are-fast-but-do-they-last/3621 . He has found that they are roughly as reliable as 1TB drives. Talking to sites that use many SSDs, they found that SSDs tend to fail catastrophically, compared to HD's that degrade gracefully over time.

ron

Posted 2011-10-13T00:34:31.917

Reputation: 41

4I object to the assertion that HDDs "degrade gracefully over time". In my experience, by the time they reach the failure mode that most people notice, they've been thrashed by the OS enough that data recovery is often quite difficult. The most common time I see HDD failures noticed is after running software updates, when the corruption affects the OS and in essence bricks the computer from the user's POV. – SplinterReality – 2011-10-18T02:15:01.610

So far, both SSD and magnetic HDDs I've seen dying do the same: When you try to access a broken spot, they go to la-la land and won't respond to anything until they're power cycled. – XTL – 2012-03-27T13:39:18.103

@Splinter the point made in the article is that SSDs fail to give advance warning via SMART. May not be significant for the users who ignore it, but it's an important difference for everyone else. – Tobu – 2013-10-27T11:27:35.463

@Tobu, it would be great if all OSes popped up warnings and really alerted the user when a SMART failure was noticed, but even that is only about 50% according to Google and their massive experience with drive failure. I've seen both Windows and Mac OSX fail spectacularly during Software Updates with a failing drive. In neither case did SMART help or predict the failure.

It's also pretty weak sauce to say "users who ignore it". USERS WILL! It's the SO or SD's responsibility to ensure that doesn't happen. Vista's incessant popups didn't help with the ignoring it problem either. – SplinterReality – 2013-10-28T00:20:55.600

@SplinterReality In my experience, SSDs have a single (and wery common) type of degradation: SSD is healthy, then it suddenly dies and 100% data is unrecoverably lost. On the other hand, HDDs have these modes: "1 bad sector" (common; 99.9% data recovery), "burnt controller" (easily replaced; 100% data recovery), "many bad sectors" (not common; ~50% data recovery) and "drive platters destroyed" (very rare; 0% recovery). – Ark-kun – 2014-06-06T15:28:57.700

@Ark-kun I have said nothing to the contrary, however, in my experience, end users tend to put up with the finickiness of their computers while the HDD fails, up until an OS Update is required. When that happens, the multiple write errors brick the OS and make the computer fail to boot.

HDDs handle bad sector relocation internally by now. (1 bad sector) If you're seeing data errors, the HDD has already run out of spare relocation sectors, and you're already staring down the "many bad sector" scenario. That's where SMART can help, because it warns when relocation sectors are running low. – SplinterReality – 2014-06-09T01:11:27.900

1@SplinterReality In my anecdotal experience, the OS is most often the canary. Heavily-written files like swap-file or registry fail first, making the system unusable and prompting for drive repair. More rigorous SMART monitoring would be good. But they won't help SSDs in their current state. HDDs slowly "burn", SSDs spontaneously "explode". So, compared to SSDs, I'd call the HDD's degradation rather graceful. – Ark-kun – 2014-06-09T19:37:59.333

@Ark-kun Fair enough, but I don't consider the failure mode of either to be "graceful" personally. Having the OS brick for a power user, isn't as big a deal as it is for your standard consumer. This is why a good, and ideally mostly transparent to the user, backup system is an absolute must, and why I refuse to acknowledge "graceful" failure modes. Without a backup, there's no way to be sure you're not going to lose data. – SplinterReality – 2014-06-10T02:12:18.053

I can only share my experience specific to my Samsung 840 EVO, it lasted 4 years but I admit that I used it intensively, I ran Debian on it and it was almost full the entire time.

It died without any warning, no SMART alert nothing, one evening it displayed 0bytes of space left, so I started deleting files but it stayed at 0bytes left, I noticed that when I deleted something the SSD capacity shrank.

So I shutdown the computer and was still able one day later to recover all data.

Jan Moritz

Posted 2011-10-13T00:34:31.917

Reputation: 131

Like (hard) disk drives they are made from multiple sub-systems and components, and can fail in multiple ways. Some instant, some partial/incremental.

SSD drives use the S.M.A.R.T. (Self-Monitoring, Analysis and Reporting Technology) systems that disk drives use to keep track of problems, and (if queried by the BIOS/OS) alert you of potential failures.

Ƭᴇcʜιᴇ007

Posted 2011-10-13T00:34:31.917

Reputation: 103 763

It really depends on the drive in question or to be more specific it's controller/firmware. Some will just become undetectable or show that their size is 0 and leave you unable to access all data on them, despite the fact that the data already written should still be ok if the reason for the failure is running out of writable flash cells.

Others will still be accessible, but you will only be able to read the data from them and not write to them.

Check this article for more info on how various SSDs died.

orange_juice6000

Posted 2011-10-13T00:34:31.917

Reputation: 115

To really know how these things fail, maybe you can get Ian over on Dangerous Prototypes to build an SSD destroyer. We've already seen what happens to a flash EEPROM subjected to the Flash Destroyer after 11.49 million writes.

Fiasco Labs

Posted 2011-10-13T00:34:31.917

Reputation: 6 368

LULZ, I thought SSD destroyer meant . . . http://www.swrebellion.com/cards/expandeduniverse/enh%20ssd%203_s.jpg

– surfasb – 2011-10-13T09:33:38.463

I think all newer SSD drives (anything beyond 1st generation) are designed to still be readable but you will start to have problems writing to some blocks which I presume SMART will tell you.

Sounds much better than using mechanical drives which (in my experience) tend to fail mechanically which results in catastrophic failure.

fred

Posted 2011-10-13T00:34:31.917

Reputation: 11

What I have heard is that ssd's fail on write and not on read. Therefore, it is much more likely that you will not loose data but, if you have unsaved documents you won't be able to save them (in the worst case scenerio).

alexyorke

Posted 2011-10-13T00:34:31.917

Reputation: 129