Samsung 4TB QLC SSDs enter mass production

[CAT-THE-FIFTH]

I don't think you will, it comes down to extended life tests and trusting the manufacturer data as by the time an independent reviewer has spent 3 years testing a product it will be obsolete an no-one will care about the results.

https://www.anandtech.com/show/9248/...data-retention


See the above, it is all about temperature. Storing in a greenhouse is bad, but if you were after an excuse for that wine cellar you always wanted to get built...

Nothing about QLC drives and later tech which is viable enough for large storage capacities - I want figures not "marketing" rubbish. They go from one to 52 weeks,to then some made up figure at the end. An entirely useless figure - three years is utterly rubbish,especially if drives can have warranties for 5 years.

So basically you are saying for storage purposes,SSDs are useless as it says one week to 52 weeks and then some made up figure at the end. It also means in the research environments I have worked in which needed 100S of TBs of storage,since they needed to backup huge amounts of data especially bio imaging and bio informatics data stacks which can be massive(for at least 5 to 7 years IIRC),due to legal requirements,SSDs won't be replacing current methods anytime soon,which will be HDDs and tape.

For anyone who is into stuff like photography or video stuff,it looks like it is better to stick to HDDs.

Yes,I get it for people with casual backup requirements it isn't an issue,but when a single day of shooting pics can easily generate 32GB of files,and that can still mean 10GB to 15GB of backups just for one day after processing images and keeping RAWs,you will soon realise SSDs are not that great,especially when keeping local backups for your entire lifetime.

This gets compounded if you do video.

[DanceswithUnix]

So basically you are saying for storage purposes,SSDs are useless as it says one week to 52 weeks. It also means in the research environments I have worked in which needed TBs of storage,since they needed to backup huge amounts of data,due to legal requirements,SSDs won't be replacing current methods anytime soon.

For anyone who is into stuff like photography or video stuff,it looks like it is better to stick to HDDs.

Actually that chart says 1 week to 404 weeks, but most importantly if you control the temperature you get to choose the number.

For now, the cost advantage of backup to hard drive is massive so I think it is moot. You still have to watch temperatures though, hard drives like to be a bit warm but the 55C operating temperatures in that chart would be enough for me to expect some hard drives to see damage to their platters.

I wouldn't expect either storage type to survive a year in my garage.

[peterb]

I don't think you will, it comes down to extended life tests and trusting the manufacturer data as by the time an independent reviewer has spent 3 years testing a product it will be obsolete an no-one will care about the results.

https://www.anandtech.com/show/9248/...data-retention




See the above, it is all about temperature. Storing in a greenhouse is bad, but if you were after an excuse for that wine cellar you always wanted to get built...
Edit: Though hard drives are cheaper, and you can spend the difference on wine

True, but I can't gurantee the storage conditions - and as this would be a stand-by solution, it could be 4 or 5 years before it is needed. And for the cost of an SSD I could do two or three cloned hard drives, and while one might fail after 5 years of non use, the chances of two or three failing is small. But the chance of an SSD losing its data after 5 years unpowered (unless its kept in a freezer!) quite high.

Hmm - keeping an SSD in a freezer - adds a new meaning to frozen chips

[CAT-THE-FIFTH]

Actually that chart says 1 week to 404 weeks, but most importantly if you control the temperature you get to choose the number.

For now, the cost advantage of backup to hard drive is massive so I think it is moot. You still have to watch temperatures though, hard drives like to be a bit warm but the 55C operating temperatures in that chart would be enough for me to expect some hard drives to see damage to their platters.

I wouldn't expect either storage type to survive a year in my garage.

They say one to 52 weeks,unless conditions are perfect and its marketing bumpf from Intel for their MLC drives(not QLC), considering I have a 14 year old WD 80GB IDE drive,which ran as hot as the sun,had 6 years of usage,and still has data readable off it after dumped in a box in storage,and the same goes with my own tests of older drives.

Plus as others say,if the drive has issues,you can still pull data off it. Every SSD failure I have seen,has meant if you didn't back up you were screwed,and after the POS Sandisk died SUDDENLY during boot,I am certainly not going to trust SSDs for any degree of longterm storage when powered off.

Until proper longterm tests are done,and the companies push out proper independent testing for EVERY new drive,they can claim anything since they don't need to back it up.

Especially when SSDs still cost more per GB.

Edit!!

Plus I don't care if its slower - its like all the people on forums moaning over SATA3 SSDs being "slow" and PCI-E being the future!!

I would have endurance,longterm data retention and capacity anytime over "speed" of SSDs.

[DanceswithUnix]

And for the cost of an SSD I could do two or three cloned hard drives,

That is the killer. HDD are the new tape, albeit you have to be careful not to drop them. I tend to use USB laptop drives, the case offers limited knock/drop protection for little cost over a bare drive and if it is sat in the back of a drawer/cupboard idle the fact it won't last long in use isn't relevant, the MTBF just has to be much longer than the time to backup & restore.

[CAT-THE-FIFTH]

True, but I can't gurantee the storage conditions - and as this would be a stand-by solution, it could be 4 or 5 years before it is needed. And for the cost of an SSD I could do two or three cloned hard drives, and while one might fail after 5 years of non use, the chances of two or three failing is small. But the chance of an SSD losing its data after 5 years unpowered (unless its kept in a freezer!) quite high.

Hmm - keeping an SSD in a freezer - adds a new meaning to frozen chips

This basically - I have two local backups on two(maybe three) different types of HDDs,and an online backup. This way I hedge my bets I will have some kind of backup in place if crap happens.

Edit!!

Also maybe I should also consider dual online backups,when I get around to it.

[DanceswithUnix]

I would have endurance,longterm data retention and capacity anytime over "speed" of SSDs.

Then I congratulate you on your choice of HGST enterprise hard drives.

If that isn't what you are running, then I guess you actually want low cost

[CAT-THE-FIFTH]

Then I congratulate you on your choice of HGST enterprise hard drives.

If that isn't what you are running, then I guess you actually want low cost

As opposed to E-PEEN speed as a consideration over capacity,endurance,etc!

Yes,I get it hardware enthusiasts like you only run benchmarks worrying whether their SSD is being limited by the connector on the board.

I think looking at other threads - YOU don't have any need for longterm storage needs,so YOU think nobody else needs or wants it.

Hence,you keep on denigrating HDDs and pushing SSDs for everything.

If YOU are so confident in SSDs retaining data for 5 to 10 years,then I assume we should all switch to SSDs,and you can refund us if there is any data loss in that period!

[peterb]

Then I congratulate you on your choice of HGST enterprise hard drives.

If that isn't what you are running, then I guess you actually want low cost

I'm running HGST Deskstar (so not quite enterprise) - (four in two RAID1 arrays actually) and they are backed up to tape and to portable USB hard drives. (the latter for a quick restore of an accidentally deleted file - the tape for a wholescale data recovery in the event of catastrophic failure of the server. TheOS is on an SSD - and thats the weak link on the system - not because it contains user data, but because of the disruption if (when) it fails. Ideally it wold also be RAID1, but converting it to RAID1 is a bit of a faff - although possible now as the age of the SSDs would be staggered by about 6 months - but the HDD solution is the most cost effective.

Then I congratulate you on your choice of HGST enterprise hard drives.

If that isn't what you are running, then I guess you actually want low cost

As opposed to E-PEEN speed as a consideration over capacity,endurance,etc!

Yes,I get it hardware enthusiasts like you only run benchmarks worrying whether their SSD is being limited by the connector on the board.

I think looking at other threads - YOU don't have any need for longterm storage needs,so YOU think nobody else needs or wants it.

Hence,you keep on denigrating HDDs and pushing SSDs for everything.

If YOU are so confident in SSDs retaining data for 5 to 10 years,then I assume we should all switch to SSDs,and you can refund us if there is any data loss in that period!

I don't think its a case of DanceswithUnix pushing SSDs for everything - its a case of horses for courses. SSDs have many advantages, and the technology is improving over time - but compared with HDDs, the technology is still new - and HDDs are anything but dead - they win on cost/GB but are relatively fragile and slower than SSDs - which is why SSDs are particularly suited as boot devices. In servers where the speed limitation in data transfer is likely to be the network (unless they are part of a SAN), the the slower speed of a HDD is insignificant.

[CAT-THE-FIFTH]

I'm running HGST Deskstar (so not quite enterprise) - (four in two RAID1 arrays actually) and they are backed up to tape and to portable USB hard drives. (the latter for a quick restore of an accidentally deleted file - the tape for a wholescale data recovery in the event of catastrophic failure of the server. TheOS is on an SSD - and thats the weak link on the system - not because it contains user data, but because of the disruption if (when) it fails. Ideally it wold also be RAID1, but converting it to RAID1 is a bit of a faff - although possible now as the age of the SSDs would be staggered by about 6 months - but the HDD solution is the most cost effective.

The thing is I have had a mixture of drives - the record goes to a WD Black drive which still works after 14 years even though its not used anymore.

I have and external 2.5" ruggedised portable drive which still worked after 10 to 11 years and the data is still intact.

I don't think its a case of pushing SSDs for everything - its a case of horses for courses. SSDs have many advantages, and the technology is improving over time - but compared with HDDs, the technology is still new - and HDDs are anything but dead - they win on cost/GB but are relatively fragile and slower than SSDs - which is why SSDs are particularly suited as boot devices. In servers where the speed limitation in data transfer is likely to be the network (unless they are part of a SAN), the the slower speed of a HDD is insignificant.

In my case SSDs have advantages in size and power consumption in addition to speed and noise,for normal stuff,which as a SFF PC fan,so for PC I can see myself not having an HDD in it within a few years.

However - we are talking about a specific niche,ie,longer term storage of largish amounts of data,on drives which won't be always accessed continuously,which is pretty much what a lot of normal people might also being doing when buying an external drive,and dumping family pics,etc on it on and off. The problem is DwL is not a fan of HDDs in general from what I see due to speed,etc,but he does not seem to get HDDs as a storage medium are far better understood and he considers cost the ONLY reason they are an alternative. Its not only cost but issues about how they failed,which you and Watercooled brought up which he seems to think is not an issue,or the real questions about data retention. Remember,Intel only posted data about ideal situations on their MLC drives.

The fact is people should not be lulled in a false sense of security thinking SSDs have no potential issues - they could end up buying one of those external SSDs for backing up family pics and find its lost data within a year.

Basically data retention beyond one year is not a requirement.

The thing is as time progresses we won't have a choice on whether we can get an HDD or SSD,but if the distinct lack of testing or data is worrying.

HDDs are a known quantity,SSDs are not with regards to the area we are talking about.

Edit!!

Don't get me wrong I'm not suggesting they're inherently less reliable than HDDs for storage given they're not vulnerable to a number of HDD failure modes, but it's still not something to simply ignore. Both have inherent advantages and disadvantages. It's interesting from an academic point if nothing else. Some people will outright refuse to accept or discuss any drawbacks of SSDs though, however small.

Pretty much my take on it.

Edit!!

If people have such belief in SSDs,its their money not mine,so let them be the test subjects.

I will stick with stuff which is a better known quantity,so TBH its not really my concern if they want to take a chance. Its not my money or my time TBF!

[DanceswithUnix]

Yes,I get it hardware enthusiasts like you only run benchmarks worrying whether their SSD is being limited by the connector on the board.

lol. That was about as far from the truth as possible

My SSD runs on a SATA 6Gb connector, and was bought as it was decent value but with one rider: I only buy from companies that actually make flash chips on the expectation the engineering will be better integrated and the chips won't be gray market fakes. That limits me to the likes of Crucial (part of Micron), Samsung, Intel, OCZ (part of Toshiba). In truth, recently I have only bought Samsung and Crucial.

From other threads, I run a home network server that handles NAS duties. That makes backup easier. The server runs email backups *hourly* because losing 20 years of email to my personal domain would be irritating. Occasional offsites for photos go on USB portable hard drive to my parent's house, probably should make that automated and cloud storage these days though. My archive data is on actively spinning and monitored disks though.

I used to work in the storage industry, so I am confident that if you lose any data then your backups weren't good enough; most likely backups weren't tested as working, amazing how many people back-up but omit that last bit! I can store data for 10 years on SSD, I'm not guaranteeing you can, and clearly with comments like "HDD is the new tape" that isn't actually what I currently do. I can however see a day when hard drive prices shoot upwards as they become obsoleted by the juggernaut of flash technology, so best to be informed. Or buy lots of actual tapes, or store in the cloud and prey someone else gets it right (I'm pretty sure they don't).

[CAT-THE-FIFTH]

lol. That was about as far from the truth as possible

My SSD runs on a SATA 6Gb connector, and was bought as it was decent value but with one rider: I only buy from companies that actually make flash chips on the expectation the engineering will be better integrated and the chips won't be gray market fakes. That limits me to the likes of Crucial (part of Micron), Samsung, Intel, OCZ (part of Toshiba). In truth, recently I have only bought Samsung and Crucial.

From other threads, I run a home network server that handles NAS duties. That makes backup easier. The server runs email backups *hourly* because losing 20 years of email to my personal domain would be irritating. Occasional offsites for photos go on USB portable hard drive to my parent's house, probably should make that automated and cloud storage these days though. My archive data is on actively spinning and monitored disks though.

I used to work in the storage industry, so I am confident that if you lose any data then your backups weren't good enough; most likely backups weren't tested as working, amazing how many people back-up but omit that last bit! I can store data for 10 years on SSD, I'm not guaranteeing you can, and clearly with comments like "HDD is the new tape" that isn't actually what I currently do. I can however see a day when hard drive prices shoot upwards as they become obsoleted by the juggernaut of flash technology, so best to be informed. Or buy lots of actual tapes, or store in the cloud and prey someone else gets it right (I'm pretty sure they don't).

That's the point you can't guarantee what you claim which is dangerous thing to do,despite loads of people pointed out problems with SSDs,and trying to bury any mention of potential issues is not good.

HDDs are a known quantity,SSDs are much less of a known one,so don't make a promise you can't keep to. The fact is you say you worked in the industry but seem to be not wanting to admit any issue with SSDs.

I worked in another industry where the data generated was mahoosive in quantity each year,and they had to maintain the data for at least 5 to 7 years for legal reasons.Data retention and data backup was a big deal. They were never going to run over to new tech which is unproven. On a personal level,the stuff I have to store needs to be kept for a lifetime,so even though it means I will have format shift multiple times,it just makes it easier not to use a tech which is somewhat longer lived and cheaper.

Companies will only address concerns if people make sure they are heard and the issue is I get the impression you are more worried about the tech than the utility of it. Yes,SSDs are better in many ways,except not in all ways. Its what I find on forums - the moment anything new comes along,they make anything older look suddenly utterly pointless.

Now,the thing is why should I care?? Its not my money,or my issue if people want to put faith into technologies which are still immature. Its their time and money in the end,not mine.

[watercooled]

And hard drives can't do the same?

Like peterb says I'm talking about the speed of actual data destruction. The only way you can do that on a HDD is if it's a self-encrypting drive where the key can be erased instantly, same for SSDs. However, SSDs have the ability to wipe their entire contents in seconds, it has nothing to do with the sequential write speed - you can issue erase commands to every NAND device and poof, everything is wiped. A similar process is used for things like TRIM and with garbage collection, but on a more granular level rather than the whole device. That ability can be a strength in some instances of course. HDDs just cannot do that though, to actually erase the data you'll be waiting hours to sequentially write to the whole surface of the platters.

Was just going to bring up the bit about 2^4 (16) possible states with QLC. Those differences in voltage are extremely difficult to handle so tbh with QLC I'm as much afraid of the low-ish endurance as I am of data deterioration. It's quite telling that warranty is significantly lower than with TLC drives.

I'd be fibbing if I said I wasn't concerned about the above when it comes to QLC. IMO it's pushing a bit too much into the area of diminishing returns. Interestingly I was reading some of the flash memory summit details at Anandtech and spotted this slide: https://images.anandtech.com/doci/13...2018441828.jpg

Interestingly they put QLC at '>20%' capacity per wafer vs TLC. So not even that close to the theoretical 33%! Is it really worth it for the loss of endurance, retention, speed - basically everything?

[azrael-]

<SNIP>

I'd be fibbing if I said I wasn't concerned about the above when it comes to QLC. IMO it's pushing a bit too much into the area of diminishing returns. Interestingly I was reading some of the flash memory summit details at Anandtech and spotted this slide: https://images.anandtech.com/doci/13...2018441828.jpg

Interestingly they put QLC at '>20%' capacity per wafer vs TLC. So not even that close to the theoretical 33%! Is it really worth it for the loss of endurance, retention, speed - basically everything?

The thing is, to offset the potential data loss you would have to have massive amounts of spare NAND for the inevitable bad blocks. That would cut even more into the capacity surplus of QLC NAND, thus making it even more of a "bad deal". The way I see this is that manufacturers of QLC NAND gladly take the risk of failure on behalf of their customers. After all, even with HDDs manufacturers hold themselves not liable for data loss. Personally, I'll gladly take a 3D NAND TLC drive (and have done, as I'm currently using a 1 TB Samsung 850 EVO), but I probably will not touch a QLC drive unless time shows that data is safe on those drives.