Samsung SSDs Dying! 980 Pro, 990, 970 EVO PLUS, PM9A1 And More

I had a number of SSD's fail, and I think they were all 870's. What was very suspicious about the failures was the way they failed and the fact that even when utterly broken, Samsung Magician still said they were OK.
To this day, I don't know what was wrong with them, other than they really slowed down. Eventually to a point where certain operations would never complete. Yet they checked out just fine and had a small amount of use.
This is not the first batch of Samsung drives that have done this. I, certainly, am not finding them as reliable as they are supposed to be.
interesting anecdote.

i can surmise. only. that there are occasional bad batches off the assembly lines (or even just singular bad samples). whether it is the memory cells or the controllers or any of the other fiddly bits on the pcb's.

some engineers would know better.
as customers can do the warranty thing if it still applies within the warranty period.

bad samples can pass tests earlier on as they are tested prior to delivery. but end up being marginal cases i suppose, in that they pass basic testing then can fail soon after. i'm not saying anything new that i haven't heard before.

my own experiences of the least reliable parts have been motherboards. most other items I have / had typically lasted around 10 yrs or longer. ive heard anecdotes of some people still using older motherboards, and yet others that have replaced a few components and they function yet again.

I got an early crucial ssd circa 2008. 128GB it showed errors after 10 years. thing is ssd's have only been around since about then in the consumer market. and newer ones are stacking the cells increasing the density, and also running hotter. `used to be generally said that SLC was the most reliable, then MLC TLC QLC - in decreasing order of Endurance. now there is V-nand (stacked vertically)

 
Last edited:
bad samples can pass tests earlier on as they are tested prior to delivery. but end up being marginal cases i suppose, in that they pass basic testing then can fail soon after. i'm not saying anything new that i haven't heard before.
In the case of the 870 Evo, what appears to happen is that the drive slowly corrupts itself and either damages the flash, or it exposes existing bad flash, since the data in those sectors goes from mostly readable to nearly unreadable and produces loads of ECC uncorrectable errors.

For some reason, software would tend to ignore the bad blocks and uncorrectable errors in the overall health report of the drive, unless they reached abominable proportions and unless someone did a full scan of the drive it didn't pick up how much of the data on it was corrupted.

Samsung have never made a formal statement on the 870 Evo issues (that I have seen), but I believe they have suggested to some customers that the problem was resolved with a firmware update, but this does not fix existing damage (e.g. reserved blocks remain used and corrupted data will always be unreadable/nearly unreadable).

The firmware update has certainly not stopped all drives from failing, but it might have been that they were too badly damaged at that point to be recoverable. Samsung don't manufacture these drives with the same process anymore (at least, it says so on their website), so presumably newer 870 Evo drives are unaffected, but earlier drives still continue to fail.
 
i got an 870 evo fairly recently 2TB. just a backup. so really it isnt being used. write once. sits on shelf.
recently bought (yet another) wd black 1tb and a 870 qvo 1 tb. same thing just back up drives.

i recall doing a firmware thing to the crucial ssd i mentioned. its all history. drive is obsolete / old. showing errors. gonski.

there are larger hdd's available. ...but i tend not to go to higher capacity ones because i dont need that much storage., plus drives have x amount of life then theyre gone. the wdblack 1tb hdd makes my fourth. still have others that are less than 10 yrs old. they are all good. (well apparently they are)

in retrospect i probably shouldnt have gone with the 870 evo. but i dont know. the qvo should theoretically have higher endurance. also a reason i still use hdd's they have proven to be reliable for me. despite being mechanical in nature. the weakest link in the chain principle applies to devices.

in terms of gen4. hmm. while there are gen 3's still available i think ill stick with them for now. I use these things until they show problems. as ive gotten older. speed is less important to me than reliability.
 
Last edited:
interesting anecdote.

i can surmise. only. that there are occasional bad batches off the assembly lines (or even just singular bad samples). whether it is the memory cells or the controllers or any of the other fiddly bits on the pcb's.

some engineers would know better.
as customers can do the warranty thing if it still applies within the warranty period.

bad samples can pass tests earlier on as they are tested prior to delivery. but end up being marginal cases i suppose, in that they pass basic testing then can fail soon after. i'm not saying anything new that i haven't heard before.

my own experiences of the least reliable parts have been motherboards. most other items I have / had typically lasted around 10 yrs or longer. ive heard anecdotes of some people still using older motherboards, and yet others that have replaced a few components and they function yet again.

I got an early crucial ssd circa 2008. 128GB it showed errors after 10 years. thing is ssd's have only been around since about then in the consumer market. and newer ones are stacking the cells increasing the density, and also running hotter. `used to be generally said that SLC was the most reliable, then MLC TLC QLC - in decreasing order of Endurance. now there is V-nand (stacked vertically)


Fortunately, most of mine that failed were in warranty so I had them refunded.

I know that EPROM's slow down the same way when you approach their end-of-life. BUt what was really strange was the way that nothing detected the fault. Even Samsung Magician would take forever to test the SSD, yet proudly stated nothing was wrong. It seemed to think that taking two hours to do a ten second test was fine. That made me think that it wasn't an SSD failure as such, rather the was the controller chip that was failing and the test simply didn't take account of that. It kinda makes sense because the controller chip is the one that gets hot. Anyway, I refunded them so I will never know.
 
Fortunately, most of mine that failed were in warranty so I had them refunded.

I know that EPROM's slow down the same way when you approach their end-of-life. BUt what was really strange was the way that nothing detected the fault. Even Samsung Magician would take forever to test the SSD, yet proudly stated nothing was wrong. It seemed to think that taking two hours to do a ten second test was fine. That made me think that it wasn't an SSD failure as such, rather the was the controller chip that was failing and the test simply didn't take account of that. It kinda makes sense because the controller chip is the one that gets hot. Anyway, I refunded them so I will never know.
makes sense. the controller is a processor. my guess is they are speeding up the processor in later iterations to increase data/bits access speed. and i've noticed how much processors like excessive heat. :-/ whether heatsinks are adequate for gen4/5 nvme's is questionable. but your example is an ssd. which dont get that hot typically. spose it depends where the temp. sensors are placed.

im actually using a crucial mx 2tb for the main drive in one of 2 pc's i use. it is relatively slow. it also is showing at 98% "good" in crystaldiskinfo.
10.4TB writes, 26C, 3402 power on hours. and my daughter managed to short out and kill another 2tb evo i only recently got. (no liquids around the pc is the rule now XD) not to be anti samsung specifically. theyve had some really good products...but i've seen that skhynix have an nvme gen 3 gold p31 model that supposedly is very power efficient. might give it a try to clone off this crucial thing. if samsung still made an 850 pro equivalent. id probably get it.
 
Last edited:
From my testing on Samsung drives iv noticed if you store data in them after few months then read speed of that data tanks hard and you have to clone the drives data back to itself for speed to pick back up for only the cycle to repeat itself.

So why buy Samsung when there are plenty of other brands that do t have this issue? Simps gonna simp I guess
 
From my testing on Samsung drives iv noticed if you store data in them after few months then read speed of that data tanks hard and you have to clone the drives data back to itself for speed to pick back up for only the cycle to repeat itself.

So why buy Samsung when there are plenty of other brands that do t have this issue? Simps gonna simp I guess
Which drive would you recommend?
 
Lots of options, micron, WD, SanDisk, lexar etc.
I just personally avoid Samsung for next few generations till they are tried and tested as being properly fixed
I'm not saying Samsung are perfect, because they aren't - but they're what you might call the best of a bad bunch.

My own experience with Samsung drives has been good, bar one drive - an 840 EVO that dropped dead without warning during the night.
 
I'm not saying Samsung are perfect, because they aren't - but they're what you might call the best of a bad bunch.

My own experience with Samsung drives has been good

Likewise. As you say no manufacturer is perfect, but Samsung (for SATA and NVME SSDs) and WD (for HDDs) have been pretty good for me.
 
Last edited:
Back
Top Bottom