How long do SSD last and how reliable are SSD?

[TheLostSwede]

It's hard to say, because most research is old, and/or hidden behind stupid academic paywall.
The only solid thing we have, is what's in JEDEC classification, e.g. 1 year @ 30C for Client SSD, and 3mo @ 40C for Enterprise.
I'm sure there's big difference between SLC/MLC/TLC/QLC in terms of storage, but no way to verify due to lack of info (except JEDEC class, of course).

One obvious thing is that a QLC cell holds four times as many charges as an SLC cell. This is going to end up having an affect over time.
That said, no consumer uses SLC any more, so that part is a bit moot.
However, we do know that the cells lose current over time, so as such, NAND should be less reliable than a mechanical drives as a long terms storage medium, be it an SSD or a USB flash drive. This is why there has been so much work on alternative technologies such as MRAM and FeRAM. However, there's still a huge capacity deficiency here, plus the fact that MRAM is crazy expensive compared to NAND. I guess 3D XPoint ends up in the group of alternatives to NAND as well, as it's supposed to be non-volatile memory.

 

[xtreemchaos]

my boot drive ssd is 9 years old its a kingston, im not going to temp providence and do a error check but touch wood its still running fine.

 

[arbiter]

Both can last a long time or be dead in a month. Thing is you can build a HDD raid to store that data far cheaper then cost of SSD's. So really question of how important is that data and how much you willing to spend to not lose it.

 

[EarthDog]

Yikes...

For long term high volume consumer storage... a HDD.

Whoever said magnetic tape should be taken out back and shot. Unless you go with older LTO tapes and drives, that's likely financially prohibitive. Then storing the tapes and having a working drive to read them... nkt good for a consumer. A large business with a dozen or more drives in a silo that stores thousands of tapes sure. For the record, I moved my company off tape to disk for backups.

 

[Bubble99]

Yikes...

For long term high volume consumer storage... a HDD.

Whoever said magnetic tape should be taken out back and shot. Unless you go with older LTO tapes and drives, that's likely financially prohibitive. Then storing the tapes and having a working drive to read them... nkt good for a consumer. A large business with a dozen or more drives in a silo that stores thousands of tapes sure. For the record, I moved my company off tape to disk for backups.

I wonder if there is difference of typical consumer tape drives vs tape drives companies use.

I read a articles that Hollywood still uses tape drives over digital storage because it can last 100s of years.

And has for hard drive I wonder if file system also plays part with how long it last.

It's hard to say, because most research is old, and/or hidden behind stupid academic paywall.
The only solid thing we have, is what's in JEDEC classification, e.g. 1 year @ 30C for Client SSD, and 3mo @ 40C for Enterprise.
I'm sure there's big difference between SLC/MLC/TLC/QLC in terms of storage, but no way to verify due to lack of info (except JEDEC class, of course).

If there is no studies than we can only really go by other members here with their SSD going bad and how long they gone with out powering it up before going bad.

 

[Rei]

I read articles that Hollywood still uses tape drives over digital storage because it last 100s of years.

Depends on which film studios in Hollywood still uses tape drives & dunno if they really last for a century but back in the day that was the only backup solution to many enterprise.

I wonder if there is difference of typical consumer tape drives vs tape drives companies use.

Nowadays, there shouldn't be any difference between consumer tape drives Vs enterprise tape drives if there were any to begin with.
I still wouldn't count on tape drives as a backup drive in this day & age as they are an obsolete technology, not to mention being costly for the capacity they offer.
As I noted in post #22, use mechanical HDD instead or an online cloud backup storage solution. They are a cheaper & longer lasting solution.

And has for hard drive I wonder if file system also plays part with how long it last.

By "file system" do you mean terms like NTFS, ExFAT, FAT32, etc?
Then the long answer is: No, they don't.
For the short one-word answer: No, they do not play any part in the longevity of a drive. File systems are just a means of reading & write data as effectively & efficiently as possible on certain types of storage drives with a certain type of operating system.

 

[Bones]

Been using HDD's for long term data backup and no problems to report at all, stuff that's over 10 years old I can still access anytime I want.
SDD's have been a different story, all total out of 10 SSD's I've bought within the last 7 years 5 are dead with 3 dying within 6 months of being new. Only one out of the 4 I bought at once still works fine to this day, the other 2 failures were drives that lasted longer, one was a really old one when it finally went so that was expected at least.

Out of HDD's I've had across 20+ years of computing only 3 have died, one was 12 years old and a daily that got reused over and over again, one was outright defective and the other just died within 2 years from being new.

HDD's have an advantage it's still possible to recover data after a failure and HDD's normally give a warning of imminent failure before it actually happens. The really old one I had (WD 30GB) gave such warnings with the noise it began to make and knew right away it was going Bye-Bye soon. Got a new one, cloned the OS over and all was well.
SDD's normally just fail without much if any warning and that's it.

For my use I've gone with what's already been proven as a potentially reliable storage medium and in fact I have no less than 3 external HDD drives (Fantom Drives) to archive stuff with. All of mine interface with the system by a USB connection and can be switched on or off via it's own powerswitch whenever desired instead of them powering up with the system everytime it's started.
That saves wear and tear on them and of course makes them last much longer than a drive that comes on whenever the system itself does.

I do suggest an external drive that can be switched on or off as desired for long-term data storage for the reasons I gave whether it's HDD or SDD based.

 

[R0H1T]

SSDs can lose data in as little as 7 days without power

New data suggests that enterprise SSD data retention can fall below a week, while client data can reach 1-2 weeks in extreme temperatures.

www.extremetech.com

Yeah well HDDs can "lose data" just as easily, I've had Plextor SSDs (2014/15 MLC) untouched in over 2 years that didn't lose anything. So unless there's scientific research (including stats for types of NAND used) proving how reliable or unreliable SSDs can be for long term storage, I'll say BS to any or all of the claims stating otherwise. SSDs have been more reliable than any spinner I've encountered over the last decade or so!

 

[Bones]

Anything you can name is always subject to failure, that's the long and short of it regardless of opinion or experience.

 

[P4-630]

Debunked: Your SSD won’t lose data if left unplugged after all

“People have misunderstood the data,” said Seagate sources at the heart of an Internet kerfuffle. Find out the truth here.

www.pcworld.com

 

[Bubble99]

Yeah well HDDs can "lose data" just as easily, I've had Plextor SSDs (2014/15 MLC) untouched in over 2 years that didn't lose anything. So unless there's scientific research (including stats for types of NAND used) proving how reliable or unreliable SSDs can be for long term storage, I'll say BS to any or all of the claims stating otherwise. SSDs have been more reliable than any spinner I've encountered over the last decade or so!

I don’t know how often people here power up their SSD. But like the posters above say hard drives can be power off for years and years in storage room than after 5 years or 10 years or more be power up and the data is there to read and write.

Well he is saying SSD that not power up will lose data or get errors. Not sure how often people here in this forum power up there SSD.

 

[R0H1T]

Well he is saying SSD that not power up will lose data or get errors. Not sure how often people here in this forum power up there SSD.

And my experience (of 2 years) says otherwise, so we really need some trusted reliable scientific research to present facts ~ not assumptions or anecdotal evidence.

As far as I know there's none in recent years, especially after the advent of 3d NAND. Heck the single biggest point of failure still remains (arguably) the controller, while on a spinner there's probably half a dozen of them!

 

[Kapone33]

my boot drive ssd is 9 years old its a kingston, im not going to temp providence and do a error check but touch wood its still running fine.

Just install HWinfo64 it will give you percentage of drive life left. In regards to the thread I will give my thoughts. I have had every type of storage and remember when OCZ and Corsair drives were very flaky and Sandisk turned out to be the best. I remember when HDDs started coming in 1 TB platters. In al of that I have had 3 drives fail. 1 was a Seagate 1TB HDD another was a 1TB Sandisk SSD (bought on Ebay) and the other was an Adata Sx8200 256 GB. The problem is that even though consumer NAND has been with us since the Atari 2600 and while Capacity has certainly increased, there has not been enough time for anyone to assess the longevity of any of these new technologies in terms of reliability. Maybe this thread should be revisited in 2031 when the OCZ Revodrive 3 will be 20 years old.

 

[Rei]

Yeah well HDDs can "lose data" just as easily, I've had Plextor SSDs (2014/15 MLC) untouched in over 2 years that didn't lose anything. So unless there's scientific research (including stats for types of NAND used) proving how reliable or unreliable SSDs can be for long term storage, I'll say BS to any or all of the claims stating otherwise. SSDs have been more reliable than any spinner I've encountered over the last decade or so!

And my experience (of 2 years) says otherwise, so we really need some trusted reliable scientific research to present facts ~ not assumptions or anecdotal evidence.

As far as I know there's none in recent years, especially after the advent of 3d NAND. Heck the single biggest point of failure still remains (arguably) the controller, while on a spinner there's probably half a dozen of them!

Yes well, reliability isn't exclusive to one type of storage. There are many number of factors to be considered reliable enough for basically any storage condition. It seems like in you operating environment, SSD would last much longer than HDD, but for others with different operating conditions, it could have the opposite outcome.

 

[R0H1T]

Except having no moving parts & basically one or two (major) points of failures SSDs will always be more reliable than HDD, archival or long term storage isn't any metric for gauging reliability though it is one of the factors to consider if you aren't going to use an SSD in a long time for any number of reasons. NAND usage isn't just exploding because they're convenient, reliability has always been a hallmark of top SSDs & SSD makers. AFAIK between firmware &/or controller you have narrowed down the two biggest culprits for SSD failures, endurance is again something which needs further scrutiny but the last test that I saw on the subject was half a decade back & we've had SSD's with 10x-100x (rated) endurance since then & of course 3D NAND.

 

[Rei]

reliability has always been a hallmark of top SSDs & SSD makers.

Reliability has always been the hallmark of any product by serious product makers really. Doesn't mean they always succeed in achieving that effort.

The worst critical point of failure for SSDs as archival storage medium is actually the electrical charge on the NAND flash. They could potentially last for about a decade if properly stored.
While HDD has more points of failure & the magnetic charge could also decay overtime, considering their built for robustness, an archival-grade HDD could last for way much longer. Also I doubt there exist an archival-grade SSD but correct me if I'm wrong.

 

[R0H1T]

Yeah I'm not sure where you're going with this? For true long term storage you already have ~

https://www.sportsvideo.org/2020/06...n-3-petasite-optical-disc-archive-technology/

https://sony.incom.de/fileadmin/design_sub/hersteller/sony/pdf/Sony-Optical-Disc-Archive-Broschuere_E.pdf

By archival data I assume you mean WORM because if not then we're talking past each other?

electrical charge on the NAND flash

Yes & no long term studies to prove the theory that SSDs cannot reliably hold charge over say 10 years, or more.

archival-grade HDD could last for way much longer

Same issue, any recent studies which prove this theory?

 

[Rei]

Yeah I'm not sure where you're going with this? For true long term storage you already have ~
https://www.sportsvideo.org/2020/06...n-3-petasite-optical-disc-archive-technology/

https://sony.incom.de/fileadmin/design_sub/hersteller/sony/pdf/Sony-Optical-Disc-Archive-Broschuere_E.pdf

By archival data I assume you mean WORM because if not then we're talking past each other?

Well, we are talking about comparing SSD & HDD as a back up medium as per this thread's topic... Aren't we???

Same issue, any studies which prove this theory?

No specific studies in particular. Mostly & merely anecdotal evidence.

 

[_UV_]

For backup SSDs don't make sense, as they lose charge over time. In other words, if they're not powered on from time to time, the data will get corrupted.
Obviously this can happen to hard drives as well, but with regards to them losing the magnetic charge. However, it would take much longer for this to happen on a hard drive than an SSD.

Beyond that, one is not really more or less reliable than the other. Hard drives are mechanical and mechanical parts fail. SSDs rely on flash which can fail for no reason and then you're screwed.

That looks like urban legends, except maybe if you are talking about very cheap first gen dramless tlc or modern qlc.

IMO intel x25 /ocz arc 100/toshiba (sold by kingston)/samsung 850-860 evo&pro/crucial m4, capacities 80-250GB, easy shelf life with data without any signs of corruption within 2-5 years. If you really care about preserving charge just use it once a half year to store new data.

The only corruption i encountered is always tied to incompatibility with controller, some ssd can't work with certain controllers. My experience with that is: OCZ ARC100 lose data with anything older intel sandy bridge or ryzen, and that is real disaster, installing OS and after 3-4 reboots you have corrupted data. Same goes to samsung and intel SATA 600 drives on SATA300/150, but it takes several hundreds of hours of work before either data corrupted or drive locks due too to many sata ecc errors.

 

[TheLostSwede]

That looks like urban legends, except maybe if you are talking about very cheap first gen dramless tlc or modern qlc.

IMO intel x25 /ocz arc 100/toshiba (sold by kingston)/samsung 850-860 evo&pro/crucial m4, capacities 80-250GB, easy shelf life with data without any signs of corruption within 2-5 years. If you really care about preserving charge just use it once a half year to store new data.

The only corruption i encountered is always tied to incompatibility with controller, some ssd can't work with certain controllers. My experience with that is: OCZ ARC100 lose data with anything older intel sandy bridge or ryzen, and that is real disaster, installing OS and after 3-4 reboots you have corrupted data. Same goes to samsung and intel SATA 600 drives on SATA300/150, but it takes several hundreds of hours of work before either data corrupted or drive locks due too to many sata ecc errors.

I guess you don't understand how NAND flash works then. But yeah, you go on believing it's an urban legend and then we'll see how it goes with your data.
Personally I wouldn't use an SSD as archival media.
Also, I never said it would happen over night, but the data will degrade over time, more so than on other types of media stored in the same way.
Maybe Russia uses some special kind of NAND flash that works differently, what do I know?

 

[mechtech]

The SSD Endurance Experiment: They're all dead

I never thought this whole tech journalism gig would turn me into a mass murderer. Yet here I am, with the blood of six SSDs on my hands, and that’s...

techreport.com

If you want permanent chisel it into a rock

 

[_UV_]

I guess you don't understand how NAND flash works then.
Maybe Russia uses some special kind of NAND flash that works differently, what do I know?

I never deep dive into this, and all you can easy find - marketing BS telling you about cost of manufacturing etc...

As i understand how it works from a simple electric principles, cell store a charge of some value, depending it be SLC/MLC/TLC/QLC, it just a value, for SLC it just 2 possible values, MLC 4, TLC 8. So it's more about precision of charging and reading value.

And i have enough samples of HDDs that don't live longer than 200 hours with usage pattern: write data and store it on shelf until you need it.

 

[silentbogo]

If there is no studies than we can only really go by other members here with their SSD going bad and how long they gone with out powering it up before going bad.

Read previous posts. There are some limited studies, but most of them are behind a paywall on academic sites. I'm not a researcher, so I won't be spending my cash to support this greedy and corrupt publishing structure.
And, no, you don't have to go with hearsay on forums to get the ballpark. JEDEC JESD218 has minimum requirements listed, so if you buy a brand-name consumer-class SSD (even QLC), it must comply with these minimum retention requirements. As far as rumors go, BICS3 and Samsung's VNAND had some engineering magic done to improve data retention, but once again - no specific numbers.
Here's a nice presentation. If you aren't interested in boring stuff, scroll down to slide 27 onwards.

https://www.jedec.org/sites/default/files/Alvin_Cox%20[Compatibility%20Mode]_0.pdf

Under certain conditions it is possible to get several years(400+ weeks) out of it without losing a single byte of data, but that's hardly practical.

Yeah well HDDs can "lose data" just as easily, I've had Plextor SSDs (2014/15 MLC) untouched in over 2 years that didn't lose anything. So unless there's scientific research (including stats for types of NAND used) proving how reliable or unreliable SSDs can be for long term storage, I'll say BS to any or all of the claims stating otherwise. SSDs have been more reliable than any spinner I've encountered over the last decade or so!

Don't forget about ECC. If your drive was shelved at room temperature for a couple of years, it's totally possible that you've had a few funky cells, but controller managed to fix it and you didn't even need to know what happened. Or maybe it was fine all along, who knows... Probability is a funny thing...
Still, I wouldn't let the chance to decide whether my 15+ years of photos and videos, or a 2TB server backup survives or not.

I still wouldn't count on tape drives as a backup drive in this day & age as they are an obsolete technology, not to mention being costly for the capacity they offer.

Obsolete and expensive? If you have lots of data - it's the cheapest option. Even previous-gen LTO-8 offers up to 12TB raw (or up to 30TB compressed). At $120/ea it's half the price of the cheapest cloud storage even for raw capacity. I'm not even gonna compare to compressed for the sake of argument. Plus, with a cartridge you get 12TB forever, not 12TB for a month and some extra expenses for using cloud provider's bandwidth and compute. I've already mentioned older LTO-5/6, which is probably the most affordable option ATM. If you buy a new-old-stock Ultrium-5 drive and a 10-pack of 3TB tape, even pessimistically it'll pay for itself after a couple of months comparing to 30TB of cloud storage at 2¢/GB. At larger scale it pays off even faster, regardless of several grand you need to put off for equipment. Even HDDs are still a cheaper option than cloud. The only few "competitive" options are Backblaze and crypto-based Sia (and a shitton of small cloud providers that use SIA as backend).

One of my newer clients uses LTO-5 for raw 4K footage archival (they shoot those crappy local non-prime-time TV shows for grandmas, so they have a lot of it).

I wonder if there is difference of typical consumer tape drives vs tape drives companies use.

Same thing, just different interfaces. Most of the "consumer" drives still meant for SMB and professional use (or crazy enthusiasts), and cost more than their enterprise equivalent. I'd still go for SAS card and enterprise drive.

 

[Bubble99]

Under certain conditions it is possible to get several years(400+ weeks) out of it without losing a single byte of data, but that's hardly practical.


Is that a typo? You say can go years with out being powered up than say 400 weeks with out being powered up.

 

[Bill_Bright]

Yes, years - if properly stored. There will always be extremes and exceptions on both ends of the timelines. But extremes and exceptions must not be used to set the rule.

One rule that MUST be set, regardless the media used, or the timeline involved if we want to ensure our data will always be preserved and that is we must always have a robust backup plan that involves multiple backups in a variety of formats. Period. No exceptions.