Samsung 870 EVO 2TB/4TB, 980 Pro 1TB/2TB QA issues manufactured in 2021? Other models too?

RanceJustice

Supreme [H]ardness
Joined
Jun 9, 2003
Messages
6,629
I've recently been informed that there are some potential major issues with certain Samsung SSDs, perhaps determined by manufacturer date! Where users have been looking at one-off issues and Samsung seems to be good about RMAing, a pattern seems to be forming that is of concern. Here is a thread detailing the overall progression of the issue and discussion

https://www.techpowerup.com/forums/...eware-certain-batches-prone-to-failure.291504

There are similar discussions on other sites (many of them referring to this one which seems to be one of the most extensive), and questions regarding the viability of the 980 Pro and if it has similar issues despite being a PRO line. Given that the 850/860/870 EVO line have more or less been THE SATA drive to pick up and that the 980 PRO was a top of the line model , this is certainly concerning. Anyone who has these drives already in their systems, especially those manufactured in 2021 should definitely run CrystalDiskInfo (open source!) and perhaps Samsung Magician (Linux users have similar diagnostics) to check , specifically. The TPU user said

"Check your 870 EVO SSDs for these things:
Elevated "Reallocated Sector Count", "Used Reserve Block" and "Runtime Bad Block" count - first warning sign (my two other 870 EVOs have none).
Non-zero "Uncorrectable Error Count" and "ECC Error Rate", and especially if those two keep rising when you read/write files. Definitely affected then!"


I am unsure about possible 980 PRO issues if the fields in issue are the same, but look for any unusual degradation! Those buying these drives new or have ones yet to install, check for manufacture dates in 2021 (some say 1st half 2021 is worse, others say all of 2021 is a concern) and consider a RMA. It is not clear if simply updating firmware can fix potential issues if your data has not yet begun to degrade, as Samsung hasn't come forth with a distinct recall or given a lot of specific info unfortunately. Those manufactured in 2022, especially anytime after Q1 2022, seem to be safe and without the issue, as well as on an updated firmware. Just wanted to let everyone know, check your drive health!
 
It's just no the 1TB and 2TB drives. I have a batch of 250GB and 500GB 870 EVOs and one of the 500GB drives had this issues as well. MFG 2021.05 on it so the dates are all over the place too.

5.27 TB Writes and it ended up with 66 Bad Sectors and over 126k CRC Errors lol. Flashed all the others right away before use and haven't seen anything yet, but also haven't hit that "certain" TBW threshold that some people seem to think starts the issues.
 
The point of a backup drive is to not have to worry about losing files. But it seems that isn't being accomplished here.

I recently bought a 4TB 870 Evo that was manufactured in February 2022, to use as a backup / storage drive. It has just 877 hours power-on time.

Does an Uncorrectable Error mean that there are already files which are damaged and cannot be recovered?

I have 4 Samsung SSDs in my PC: an 850, two 860s, and an 870. They each have exactly 100 uncorrectable error counts, and all the same other stats:

Reallocated Sector Count - 100
Used Reserve Block - 100
Runtime Bad Block - 100
Uncorrectable Error Count - 100
ECC Error Rate - 200
 
Last edited:
The point of a backup drive is to not have to worry about losing files. But it seems that isn't being accomplished here.

I recently bought a 4TB 870 Evo that was manufactured in February 2022, to use as a backup / storage drive. It has just 877 hours power-on time.

Does an Uncorrectable Error mean that there are already files which are damaged and cannot be recovered?

I have 4 Samsung SSDs in my PC: an 850, two 860s, and an 870. They each have exactly 100 uncorrectable error counts, and all the same other stats:

Reallocated Sector Count - 100
Used Reserve Block - 100
Runtime Bad Block - 100
Uncorrectable Error Count - 100
ECC Error Rate - 200
I'm not entirely sure but, assuming you're seeing those on CrystalDiskInfo I think you're okay. I see those 100s too but if you look at the raw value its probably all 00000000's right? And the "threshold" is 10? I could be wrong but this seems a lot like a "you're currently at 100%, everything is fine" and there either needs to be 10 instances of shit going wrong or you go from 100% > 10% before it flags that field as busted. Not sure which is wihich off the top of my head, but Looking at my 840 Pro and 860 EVO all which have similar fields yet are rated as very healthy, I'd have to check out SMART ratings and see what it means properly in this context. However when I saw people reporting problem drives it wasn't like an even 100 across many fields. I'll have to double check though
 
Use a program like HDSentinel instead of CD Info as it is much better. Has a free version as well.
 
I'm not entirely sure but, assuming you're seeing those on CrystalDiskInfo I think you're okay. I see those 100s too but if you look at the raw value its probably all 00000000's right? And the "threshold" is 10? I could be wrong but this seems a lot like a "you're currently at 100%, everything is fine" and there either needs to be 10 instances of shit going wrong or you go from 100% > 10% before it flags that field as busted. Not sure which is wihich off the top of my head, but Looking at my 840 Pro and 860 EVO all which have similar fields yet are rated as very healthy, I'd have to check out SMART ratings and see what it means properly in this context. However when I saw people reporting problem drives it wasn't like an even 100 across many fields. I'll have to double check though
I think you're right. The raw values for each of those stats are all 000s.
 
Just wanted to update this - do you think its worth it seeing if I can get Samsung to RMA these unopened 2021 made drives rather than roll the dice and try to monitor, then if it goes poorly have to dump all the data off before sennding them in? I'm planning to put these into new systems likely to be built as soon as the 7950X3D releases, so I have a little time and they've been sitting on my shelf. Anyone know if Samsung is willing to RMA based upon this issue ?
 
Just wanted to update this - do you think its worth it seeing if I can get Samsung to RMA these unopened 2021 made drives rather than roll the dice and try to monitor, then if it goes poorly have to dump all the data off before sennding them in? I'm planning to put these into new systems likely to be built as soon as the 7950X3D releases, so I have a little time and they've been sitting on my shelf. Anyone know if Samsung is willing to RMA based upon this issue ?
It's probably worth a try, but I'm going to guess that Samsung doesn't want to bear all this expense. In effect, you are tilting at windmills.
 
Checked my 980 pro last week and was in the clear. Now this..

I have three 870 evos to check.
 
Digging into this a little further, I've read that even Samsung gave a halfass-but-not-really confirmation of the issue by mentioning that updating to the latest possible firmware for 980/990 Pros will apparently stop this degradation from happening. However, it will not reverse any pre-existing degradation from use on an affected drive on older firmware! The same is apparently true for 870 EVO / Pro drives. All of this I'll need to see if there are more substantive confirmations, but it does seem to be promising if it turns out to be accurate. Check your drive health and update to the latest firmware!
 
Via the hotdeals thread on 4tb drives, a nice compilation of affected models and serial numbers for potential bad drives.

https://nascompares.com/2023/02/02/...reporting-failures-everything-we-know-so-far/

Ars Tech article:

https://arstechnica.com/gadgets/202...sds-after-users-report-abnormal-health-drops/

I hadn't heard it was also the 970 drives too. I just checked serial number of the 970 evo plus I have and it misses the april manufacture date by a month. Still, probably worth it to check any drive one might have with the magician software even if not supposedly affected.
 
Back
Top