Storage SSDs seem to die more often than I expected...

I just want to air this and see if it rings a bell with anyone. I've sold somewhere in the region of 600 SSDs in the past 5 years and have only had about 50 or so failures - so far. I guess 8% isn't too bad? But it's more than I was expecting, and the % is climbing all the time - which mathematically makes sense, but does imply we're nowhere near seeing the true 5-year or 10-year % yet. Over a 10-year interval it's looking like the failure rate might approach 15-20%, which seems lousy to me.

And these failures follow a pattern: they almost all died within about 18-24 months of being installed. Not ancient, clapped out drives understandably retiring after 8 years of loyal service - more like perfectly good, freshly recruited drives going AWOL and turning up dead in a strip joint. No commonalities in terms of how much they're used, what controller or chipset, environment, etc. They just gave up, usually after minimal use. They also tend to fail catastrophically and brick themselves, making data recovery much less likely.

I'm sure out of 600 HDDs I'd have had worse failure rates, but not that much worse, really. The "SSDs are so reliable" patter is wearing a bit thin. I feel like I'm omitting important caveats every time I say it. Yes, on paper they spontaneously die less often than HDDs, but that's like saying your papier maché house is much stronger than the cheese and spittle house you used to live in. They still up and die suddenly, often and at the expense of all your data and software. Saying "this is statistically unlikely" to a customer who's just lost all their data gets real old, real fast.

Of course I am now giving out a standard caution to all SSD recipients to do your backups but I wanted to see if anyone else has had experiences like this or if I'm just choosing crappy SSDs, because I find the failure rates kinda baffling.

 

I've not aware of any in my (medium sized) company or in my personnel life have an SSD fail on them yet but as we know anecdotal evidence is patchy at best.

Some data from google reported here: https://www.zdnet.com/article/ssd-reliability-in-the-real-world-googles-experience/

Seems to suggest a high rate of bad blocks and 2-7% failure rate for a bad chip in 4 years.

 

Much lower use case, but out of the 7 SSDs I've purchased I've had 1 failure. An OCZ Vertex.

Still works, claims it's healthy, reports no errors or bad blocks, just slowly and inexorably corrupts every file on it if you try to use it.

 

It's quite pleasing to see that Google's upper failure rate roughly matches mine. I'm not going crazy or buying crap SSDs, at least.

 

I've had at least 15 SSDs over the last few years. An OCZ Vertex (they were renowned for it iirc) and a SanDisk failed although a power surge through a faulty USB Hub may have killed the SanDisk.

I Over Provision all my larger drives (500GB to 2TB) by at least 20% and the few smaller ones by 10%. About 1/2 of mine are actively cooled and most of the rest passively cooled with heat sinks.

I've only been running my 4 x M.2 and NVMe M.2 drives for less than a year and have no failures yet. Again all but the WD are Over Provisioned. The WD disk tool doesn't recognise an NVMe M.2 in a Thunderbolt caddy.

 

I had a Vertex die, but somehow my Agility3's just keep on going. Also had a Sandisk that was DOA. All others still seem OK though, and that probably covers 10 drives or so (500 & 1TB) that have been re-purposed for other people.

 

From what you describe it sounds like electronic failure or firmware counter (I remember some SATA SSD's used to have this so just after xxx amount of time it'd brick itself) rather than NAND wear, as that should be a more gradual thing. If you're hitting the top end of googles numbers then that makes sense for cheaper, consumer drives and firmware. The super high MTBF and TWB drivess only really appeared in the last few years (the older ones had prohibitive performance, so few people used them) so thats probably why you're not seeing super long lives.

Plus, if its electronics failure, HDD's and SSD's are simular amounts of electronics so thi makes sense why it'd feel they'd be in simular plaves on the bathtub curve.

 

The worst part I've noticed is that unlike HDDs, they're fine one second and gone the next. I've had a 256gig adata SX8200 M.2 (was like 8 months old) drive and an old patriot wildfire 128gig drive die on me in the last 2 years, just poof gone. That reminds me, I haven't done a backup in a few months....

 

Only had a couple of Samsungs, a couple of HyperX's and the majority Crucials.

All still fine (even the 64Gb M225 from 10+ years ago) but then I don't thrash them if that makes a difference.

Tell a lie, running a Corsair nvme at the mo as well.

 

Among the list of all my SATA SSDs, witch is: 6 Samsung EVO (oldest are from 2016), a few Crucial BX, 2 HyperX, 4 Plextor M6, 2 Lexar, 2 Trancend, 1 Kingston and the only one who died, my oldest one, an OCZ vertex 3

 

i have installed almost close to that number of ssds over the years for systems i have built, i only ever had 2 fail one ocz vertex one in the early days and a samsung evo 840 ( the ones that had degrading issues and firmware fix was issued).
i remember in the early days i was really worried about the long term reliability of ssds but compared to mechanical hdds i my self have had no issues bar the 2 drives i have mentioned

 

The failures I'm seeing don't seem to be due to the long-term wearing out of cells that we all worried about in 2010. These failures are more like premature controller/chipset failure; the drives I've had come back dead are simply too young and too underused for cell decay to be the culprit. So something else is going on.

If it was really heavily used drives, or really old drives, I'd shrug and go "ah, well - that's the nature of SSDs I guess." But the fact that it's controller/chipset failures killing them really young makes me wonder if it's just a QC issue that will be ironed out in the future.

 

I especially wonder about M.2 drives. The mounting socket has slight spring tension and every motherboard manufacturer under the sun has a slightly different integrated heatsinks. Super easy to crack a BGA solder joint if the engineering or manufacturing isn't perfect.

 

I can't remember the last time I installed a spinner.... perhaps it was back in PS3 days when Hybrid drives were a thing, well over a decade ago, its been all SSD for me, never heard of a failure in my circles, probably means I'm due a catastrophic one.....

I may have had a failure on a 64Gb Sandisk I used for Rapid storage technology but I was doing some FSB overclocking at the time and not sure I could blame the drive for taking its bus out of spec.

The only things that have spinners are big storage devices.

 

Which is exactly why I wasn't going to reply to this thread.

The old 'strange, i've not had a problem.... hang on, what's going on here, oh fu...'.

Fateus temptus superstitous.

 

Nah, its a proper standard where you buy the socket and standoff as a standardised pair so although you feel like there's pressure its far below what FR4 PCB's need to actually bend. If it was a problem we've hace seen it in Wifi cards, cellular cards, mobile GPU's ect...

 

Laptop RAM slots were the ultimate proof of concept for that, I had the same anxiety about them but I've actually never seen a faulty slot. I have on desktops, oddly. Maybe having the push force directed straight towards the motherboard actually makes it more likely to crack a solder joint? Or maybe it's just that people are instinctively rougher with desktop hardware. Or maybe it's that a desktop motherboard is suspended on risers and so has more surface area to flex and bend than a laptop motherboard...

 

I don't believe in superstition, but tech does have a nasty way of breaking after you talk about it. Probably Gremlins...you got one wet didn't you?

 

Personally I've bought about 20 ssds and not seen one fail, including my ocz vertex 2e that's still going strong.

Professionally I've purchased about 65 to 100 laptops a year with ssds of all types and size for the past 6 years and have yet to have a sad fail.

 

Maybe the problem is the naff Chinese brand I used for the first 3 years, Drevo.I have also had 3 dead Crucials and a Samsung too, though. And speak of the devil - a chap who bought a brand new SanDisk external SSD a month ago has had it totally fail. There's just no rhyme or reason to it...