Samsung 870 EVO - Beware, certain batches prone to failure!

[P4-630]

My 870 Evo 500GB, used as external backup.

 

[CiTay]

Like i wrote, under "Function" -> Advanced Feature -> Raw Values, select "10 [DEC]" to have human-readable values. You can then upload screenshots of the data, press CTRL-S to screenshot in there.

But i already see even in hexadecimal that the important values for this issue are all 0, so you'e not affected. Plus the SSD seems almost brand new. What's the date printed on the SSD or the packaging?

 

[P4-630]

What's the date printed on the SSD or the packaging?

2021-11-18 made in Thailand.

 

[CiTay]

Yeah that's way newer than the failing ones are usually from. I don't think you will have problems with this one. But definitely keep an eye on the SMART values, i'd check every couple weeks or so.
There are four values which are important, i named them in the first post.

 

[RJARRRPCGP]

randomly corrupt some of the files

That is what's most likely to occur with faulty SSDs. The bad-sectors report by Magician on post #12, is what usually happens with platter drives, albeit unless super bad, usually will be some sectors beyond 50 percent of the drive or closer to the half-point, versus the beginning.

Mea culpa=I can't tell if the bad sectors are at the beginning or not. (with the Magician block display)

 

[chrcoluk]

mine is manufactured june 2021

 

[Tarte]

I use it as an external backup, formally internal for data.

 

[CiTay]

I use it as an external backup, formally internal for data.

That one is for sure affected, RMA it. Meets both criteria:

Elevated "Reallocated Sector Count", "Used Reserve Block" and "Runtime Bad Block" count
Non-zero "Uncorrectable Error Count" and "ECC Error Rate", and especially if those two keep rising when you read/write files

What date was it made?

 

[Tarte]

Made date is 02.2021.
I will RMA it.
Thank you!

 

[ThrashZone]

Hi,
Thankfully I haven't gone past 860 versions yet but at 99% but only at 500gb data
Even my old 850's look good at 99% 256gb os

 

[CiTay]

Again, under "Function" -> Advanced Feature -> Raw Values, select "10 [DEC]" to have human-readable values. I don't know why it doesn't default to that, the hexadecimal values are harder to read (except 0).

Health status means absolutely nothing for this particular problem. Health status percentage relies mostly on a conservative estimate from the manufacurer about how many times the cells can be overwritten (i.e. the TBW value), and the real endurance is often 10x higher. It makes sense that the TBW is a conservative value, this way the manufacturer can decline RMA if you have used the SSD excessively and are above the TBW value. This way, when you use an SSD professionally in a server environment, you would get a server-grade SSD instead. And that's not just pure marketing, the best NAND flash is always used in server-grade SSDs.


Anyway, for this problem, the only thing you have to look for in the SMART values is:

"Reallocated Sector Count" & "Used Reserve Block" & "Runtime Bad Block" going above 0
and
"Uncorrectable Error Count" & "ECC Error Rate" high/increasing

The "and" is also important. A couple reallocated sectors are not the end of the world when no data is affected. Like the name says, the affected blocks get reallocated/remapped with reserve blocks, and will be avoided for saving any data in the future. But when you have uncorrectable errors / ECC errors, it means that data is affected. Some blocks that already had data in them have gone bad, and thus some file cannot be read anymore, because once the SSD tries to read those blocks where part of the file is saved, it aborts, and the Error count is increased. That's when you know you need to RMA, this should not happen with an SSD.

 

[ThrashZone]

Hi,
Missed where you used samsung magician and what it said about the ssd's

 

[PhilR]

Have already 3 failed 2 TB EVOs with serials S62* - didn't check production date yet. All failed at around 13-14 TB written, pops up with read error and smart reallocation warning.
They were in production for less than a year. Samsung's tool says they are GOOD.

 

[P4-630]

Have already 3 failed 2 TB EVOs with serials S62* - didn't check production date yet. All failed at around 13-14 TB written, pops up with read error and smart reallocation warning.
They were in production for less than a year. Samsung's tool says they are GOOD.

They were still under warranty, so I assume you got new ones from Samsung and they are OK?

 

[Powernoob]

Today I realized that my Samsung 870 EVO is super slow and the task manager said it was used 100 % all the time. At first I thought it's a Windows search thing but then I saw in the windows logs that it complained about bad sectors on the drive - MANY bad sectors.
Crystal DiskInfo provided this and I wanted to share it with you guys since your replies in this forum where very helpful. Thanks by the way!

Its almost unusable now. I try to rescue some more important files, but the speed (to another SSD) fluctuates between 9 MByte/s and 0....

 

[Tetras]

Today I realized that my Samsung 870 EVO is super slow and the task manager said it was used 100 % all the time. At first I thought it's a Windows search thing but then I saw in the windows logs that it complained about bad sectors on the drive - MANY bad sectors.
Crystal DiskInfo provided this and I wanted to share it with you guys since your replies in this forum where very helpful. Thanks by the way!

Its almost unusable now. I try to rescue some more important files, but the speed (to another SSD) fluctuates between 9 MByte/s and 0....

Wow, only 3000GB written, they usually seem to die 10000+

 

[Icon Charlie]

When the 870 came out I looked at their construction. I went and bought the 6, 860 EVO instead. IMHO Samsung went cheap on the 870. I did not like how the 870 was put together. It takes a little bit of research now these days because you can not just buy on name brand only.

Every corporation is cutting corners these days. Samsung is no exception.

 

[CiTay]

Crystal DiskInfo provided this and I wanted to share it with you guys since your replies in this forum where very helpful. Thanks by the way!

Yeah, there's an ungodly amount of ECC errors, while there aren't even that many bad blocks (at least in relation). What this tells me is that the bad sectors affect some important or commonly read files, perhaps even a folder structure instead of a file, or that you have repeatedly attempted to read out certain damaged files.

Let me tell you from my own attempts: There is no way you can successfully read out files anymore that contain bad blocks. The controller or Windows will only stall when you try to do that, and the ECC errors will increase. Those files are basically lost. They might still contain 99% valid data, but the damaged 1% will cause insurmountable trouble for the SSD and Windows.

Can you tell me the serial number of your drive (minus the last four digits) and/or what production date it is (should be on the SSD and the packaging)? I want to know if this is another of the early production drives.

 

[Powernoob]

Sure - the serial number is S6BCNG0R21xxxxx and its production date 2/21.

 

[R-T-B]

When the 870 came out I looked at their construction. I went and bought the 6, 860 EVO instead. IMHO Samsung went cheap on the 870. I did not like how the 870 was put together. It takes a little bit of research now these days because you can not just buy on name brand only.

Every corporation is cutting corners these days. Samsung is no exception.

Not sure what you can construe from "construction" when everything is down to NAND chip and controller these days... and those all look the same construction wise.

 

[CiTay]

Sure - the serial number is S6BCNG0R21xxxxx and its production date 2/21.

Very similar to my affected one, which is a S6BCNG0R207xxxx from February '21. So it's another one from those early production batches.

There must be thousands of those early 870 EVOs around, silently developing bad sectors, which will only be discovered when some file can't be read out anymore someday...

 

[Powernoob]

The odd thing: I'm not sure if I actually lost a file (my last full backup was 5 months ago, so there was not too much to sync). Maybe I was lucky(?) enough to realize from the extreme speed decrease that something is wrong before losing data. At least no important data seem to be lost.

 

[CiTay]

Good. It's good to always make a backup of truly important data anyway. As we see, things can go wrong even with a top-shelf SSD. If you RMA it, it would be nice if you could write how that goes.

 

[Powernoob]

I already contacted Samsung. They wanted some information and screenshots of Magician's diagnose, smart and drive details page. Also they required me to send photos of the drive (top, bottom, interface) and a copy of the receipt as proof of purchase. After that I got a link where I had to fill in some of the infos again. Now I wait for the RMA-labes to be sent to me and the parcel service to contact me.
We'll see...

 

[thatoneguy111]

I'll add my points.

Two 870 EVO 1TB, one with significant Uncorrectables and the other some. Just over 1 year old, SN in the S5Y2NJ0NC0xxxxx range

Model Name

Samsung SSD 870 EVO 1TB

ID	Description	Threshold	Current Value	Worst Value	Raw Data
5​	Reallocated Sector Count	10​	85​	85​	164​	OK
9​	Power-on Hours	0​	98​	98​	9104​	OK
12​	Power-on Count	0​	99​	99​	203​	OK
177​	Wear Leveling Count	0​	98​	98​	41​	OK
179​	Used Reserved Block Count (total)	10​	85​	85​	164​	OK
181​	Program Fail Count (total)	10​	100​	100​	0​	OK
182​	Erase Fail Count (total)	10​	100​	100​	0​	OK
183​	Runtime Bad Count (total)	10​	85​	85​	164​	OK
187​	Uncorrectable Error Count	0​	99​	99​	5439​	CRITICAL
190​	Airflow Temperature	0​	70​	54​	30​	OK
195​	ECC Error Rate	0​	199​	199​	5439​	CRITICAL
199​	CRC Error Count	0​	100​	100​	0​	OK
235​	POR Recovery Count	0​	99​	99​	189​	OK
241​	Total LBAs Written	0​	99​	99​	76905203513​	OK

Model Name

Samsung SSD 870 EVO 1TB

ID	Description	Threshold	Current Value	Worst Value	Raw Data
5​	Reallocated Sector Count	10​	99​	99​	9​	OK
9​	Power-on Hours	0​	98​	98​	9118​	OK
12​	Power-on Count	0​	99​	99​	205​	OK
177​	Wear Leveling Count	0​	98​	98​	40​	OK
179​	Used Reserved Block Count (total)	10​	99​	99​	9​	OK
181​	Program Fail Count (total)	10​	100​	100​	0​	OK
182​	Erase Fail Count (total)	10​	100​	100​	0​	OK
183​	Runtime Bad Count (total)	10​	99​	99​	9​	OK
187​	Uncorrectable Error Count	0​	99​	99​	180​	CRITICAL
190​	Airflow Temperature	0​	70​	48​	30​	OK
195​	ECC Error Rate	0​	199​	199​	180​	CRITICAL
199​	CRC Error Count	0​	100​	100​	0​	OK
235​	POR Recovery Count	0​	99​	99​	126​	OK
241​	Total LBAs Written	0​	99​	99​	7.6E+10​	OK