Why did my SSD just die?

Samsung 850 250 GB are great and reasonably priced.

This is old and SSD are getting better. But it's worth knowing SSDs die. A lot. Particularly within the first 2-3 years.

OCZ Vertex was the one with a firmware update that enabled higher performance at the cost of randomly busting the drive permanently. I hope you were not running on that firmware version.

I'm sorry for your loss.

Hardware fails. Sometimes it fails early.

It serves as a reminder to the very important 3-2-1 rule of data. Always sure you have a minimum of: 3 copies, 2 media types, 1 off site. For me and many others, that means a local disk, a NAS disk, and an online backup service. As a recommendation for backup, I use FastGlacier with the data stored on Amazon for a few bucks a month, cheaper than any other service I've found and no pesky AUP violations when their 'unlimited' plan really means 'under 10 GB of user data'.

Most of my SSDs are Samsung, though my first was a 256GB Crucial MX which I believe is old enough that it's only SATA II, and its still going strong. I've also got a Crucial mSATA 128GB SSD that's also been fine. I prefer the Samsung drives now for their consistent high performance, and among those I prefer the PROs rather than EVOs for my boot drive -- since they use higher-quality flash and hence have a longer warranty (I care less about cashing in the warranty some day than I do about it being a statement of Samsung's own confidence in these drives). Never had any problems whatsoever with any of these.

That said, for certain manufacturers and especially in earlier days, the failure mode of SSDs seemed to be a propensity for sudden, total failure. Of course, spinning disks can fail hard too, and newer quality SSDs have this problem less and less, but that's always stuck with me, so I have a mirrored RAID volume in my desktop where my important work is. Currently that's a pair of 750GB laptop hard drives. I also have a home NAS running Windows Storage Spaces with 4 drives providing unmirrored (POD), mirrored (RAID 1), and parity (RAID 5) volumes (I use different ones depending on the resiliency I want for the data), and most of my irrecoverable work is also in GIT somewhere online (though, I do need to create a more thorough, less ad-hoc online backup plan).

I personally would avoid anything based on a Sandforce controller, or on bargain-basement controllers. Sandforce was impressive when they regularly outperformed all but the most expensive competitors, but they don't stand out so much these days and the hoops they jumped through to get there is just one more thing that could go wrong. The cheaper controllers often have inconsistent performance, don't do TRIM well, lack encryption, etc. I know the lower price is appealing (in fact, in my new build next year I'll be using large Samsung EVOs for my local RAID, rather than spinners) but for your primary or boot disk I just don't think the saving a few 10s of dollars on a $100+ product is a smart move. Swallow the extra cost and consider it as a one-time insurance premium.

Thanks! It's indeed a reminder for backing up stuff more often since I lost a few days of work.

"OCZ Vertex was the one with a firmware update that enabled higher performance at the cost of randomly busting the drive permanently. I hope you were not running on that firmware version."

Probably not, I didn't install any update on it. Now that I'm reading into it they don't handle power loss very well and recently I had a lot of crashes where I had to reset the machine. ...I don't know but I guess it doesnt matter now.

Money-wise it came at the worst time, I want to update my PC - new HDD, more RAM and a few small things - and this "update" was totally not planned. I have to go with something cheap. Im looking at the 120gig Intel 535 now. It comes with 5 years warranty which sounds OK but I can't really find any reviews on it.

Have you tried plugging it on another pc as a secondary disk or something like that? Just to see if it at least gets recognized?

Anyway...I had a very similar problem with an OCZ Vector 256GB. It all started with a freeze like the one you described, then the disk was nowhere to be found after 1 or 2 minutes upon booting. After contacting their support, they suggested I would create a bootable media of their OCZ ssd guru toolkit and boot in it. Suprisingly, it worked! I was able to see the ssd without it disappearing.

Well, I couldn't save any data, but it managed to save the disk so I was happy with the outcome. I performed a firmware update, a secure erase of the disk (so it goes back to its factory default state), and then clean installed everything from scratch. This happened like, 4 months ago, and I haven't seen any signs of malfunction since. I have this ssd for a total of 3 years now and that was its only problem, so as far as market recommendations go, I would easily buy an OCZ disk again.

I definitely recommend contacting their support and posting on their forum, you have nothing to lose. They are fast, friendly, and did the job for my case. I hope you can save your disk as well.

Have you tried plugging it on another pc as a secondary disk or something like that? Just to see if it at least gets recognized?

Anyway...I had a very similar problem with an OCZ Vector 256GB. It all started with a freeze like the one you described, then the disk was nowhere to be found after 1 or 2 minutes upon booting. After contacting their support, they suggested I would create a bootable media of their OCZ ssd guru toolkit and boot in it. Suprisingly, it worked! I was able to see the ssd without it disappearing.

Well, I couldn't save any data, but it managed to save the disk so I was happy with the outcome. I performed a firmware update, a secure erase of the disk (so it goes back to its factory default state), and then clean installed everything from scratch. This happened like, 4 months ago, and I haven't seen any signs of malfunction since. I have this ssd for a total of 3 years now and that was its only problem, so as far as market recommendations go, I would easily buy an OCZ disk again.

I definitely recommend contacting their support and posting on their forum, you have nothing to lose. They are fast, friendly, and did the job for my case. I hope you can save your disk as well.

Thanks! I've tried to boot with it but it can't find the SSD. I tried the same on an older laptop, nothing. On my PC I took the battery out from the motherboard for a few hours as to reset the bios but same thing, it's not seen. I will try it on an other PC soon, if still nothing then I will post to their forum maybe they know a magic trick but probably not:)

I wouldn't say 3 years is the "normal" lifetime for recent SSD's

My ssd's have all died within a year or 2, except the newest ones I got have lasted 2 years so far (woo progress!).

Remember back in the day IDE hard drives used to fail just as often (A computer store near me sold 512 mb drives for $5 as-is. some even worked!)

Hopefully the same trend follows and we get blazing fast ssd drives that are reliable.

Never had an SSD fail so far. Do you mean that they reach the end of their wear-level as shown by various SSD-tools, for example Intels and Samsungs bundled tools, or that they just die for some malfunction reason?

I have an Intel 510 120 GB that's going on 5 years and has been full or near full at least half of that and have had to endure a lot of source-code storage and temp build files from Visual Studio so at least somewhat well used.

Never had an SSD fail so far. Do you mean that they reach the end of their wear-level as shown by various SSD-tools, for example Intels and Samsungs bundled tools, or that they just die for some malfunction reason?

They typically die for other reasons.

There have been many different writeups, some are quite comprehensive. This one from Tom's Hardware back in 2011 covered data released by Google and SoftLayer, and from technology analysts. They basically had many hundred thousand drives they could watch and monitor over the course of five years.

This graph probably describes the rates better than any other:


That is the annualized failure rate, looking at constantly-used drives in big datacenters like at Google. Some are used continuously, others used more sporadically, some used 24/7, others used only a few hours per week. Thousands of drives in lots of situations across a wide range of usage patterns.

Note that these were corporate server failure rates. You almost certainly use your drives much less, so extend the graph dates by a few years.

The graph describes that a small number of drives arrive just fine, but then about 3% fail within the first three months. Then another roughly 2% fail between 3-6 months. Then another roughly 2% fail between six months to a year. By the end of five years, about 40% of these big drives have failed.


They can die for any number of reasons. When drives warm and cool parts expand and contract, the boards and circuitry may not handle the expansion and contraction; maybe it causes something to crack, maybe something cools too fast or too slow and expands wrong triggering a short. Maybe something overheats. Maybe they are hit by random radiation that kills a circuit. Maybe it was a subtle manufacturing defect that took a small amount of wear and tear to uncover.

Their survey discusses that specific vendors are more likely to have better drives and others are more likely to have worse drives, certain models are known for reliability and others are known for rapid failure, but ALL models from ALL vendors had unpredictable failures. Some were very small, about a quarter of one percent AFR, but that is still not zero; that is one in four hundred that dies young.

No matter the reason, no matter the brand, a small number of drives die early. Some drives fail within weeks. Other drives can run continuously for a decade without problems. But sooner or later every drive will die. You don't know when that is. You need to be prepared for it.

Personally I don't trust a work disk drive at all after three years, and I don't trust my home's disk drives after five years. At that point they are due for replacement.

They typically die for other reasons.

...

Great post, thanks!

Just to drive Frob's point home -- wearing of the flash (not defect) is really a non-issue. The write ratings on flash are firstly a lower-bound, and flash cells commonly go well beyond their rating without issues, some have been recorded going 10s of times longer than their rating. Secondly, even at 'only' 3000 cycles, that's around 8 years of entire replacing the drive's contents 3 times each day, more than 10 years if you give them the weekend's off. Wear-leveling multiplies that by a factor proportional to the contents of your disk that do not change.

I had a Mushkin SSD for 6 months and it died, and reboot didn't show up as well. I got it replaced from warranty, however it has been 6 months and I still haven't plugged it in. I already had to re-install windows on a standard hard drive while waiting to ship there, and back for a replacement, so I didn't want to re-install windows 2 weeks later again.

I already had to re-install windows on a standard hard drive while waiting to ship there, and back for a replacement, so I didn't want to re-install windows 2 weeks later again.

Haha, I will need to install both windows and linux, ...the latter one will take me days to get it back to the way it was. I should have kept more backups:).

In the meantime I tested it on other PCs. Seems like it's totally gone.

I'm sure I remember the marketing spiel for the first solid-state hard disks saying things like "say goodbye to hard disk failure" and that they would be more reliable than a disk with moving parts. I've had ide disks that have lasted ten years in constant use, until the average for a solid state matches up to that I think they were perhaps fibbing a bit...

I've never had an HDD failure (except an expensive NAS/"enterprise" grade one in my company NAS)...
The first SSD that I owned died in 6 months (no-show in BIOS, unrecoverable). But since then I've had 5 more SSDs and no failures.

I'm not sure that's enough data to even pull anecdotal evidence from.
I'd be interested to know the large scale SSD vs HDD fail rates. There's been some SSD scares/controversies, but I don't know about a trend.
IMHO they are an essential component in any new PC now, and I trust them :)
(but still: backup!)

I think that SSD lifespan is much that of RAM like. Pick large width bus, slower frequencies, get it cooled, and you should be set for a lifespan of a DDR3 1333MHz solid brand RAM. Not too long still.

I wonder why there are not rather HDDs with 1GB cache and stuff like that.

I've seen plenty of HDD failures, also in the enterprise/server area, where things are very different when a disk has 10s or 100s of people hitting it constantly and concurrently for an 8 to 10 hour period each day.

My understanding is that the failure characteristics of HDD and SSD are very different. HDD builds up slowly, starting with bad blocks, and it can be quite some time before a failing disk is actually unusable. SSD tends to die in style; it goes out instantly. So it's easy to understand that a hypothetically more robust technology that goes out instantly can actually seem more failure-prone than one that takes a longer time to die: there may actually be a lot of HDDs out there that are in the process of failing, but because it's just damaged blocks and sectors the disk is still (mostly) usable and it doesn't appear to be that way.