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

[CiTay]

Yeah, fair enough, i'm convinced. I'm easy to convince with facts, don't worry

 

[E.S]

Zeromis said:

What is FC even meant to represent though.

SMART ID "FC" (252) is described on wikipedia as follows:

252 0xFC

Newly Added Bad Flash Block

The Newly Added Bad Flash Block attribute indicates the total number of bad flash blocks the drive detected since it was first initialized in manufacturing.[81]

Given this explanation, increasing this value seems like a problem.
I've been using the 870evo for over a year now and the "FC" value has been increasing recently.

 

[JagCube]

Since 0xFC / 252 is officially a "vendor specific" attribute, are there really any guarantees it means exactly the same thing on these current Samsung drives with the new firmware, as somebody somewhere on the Internet years ago determined it means on their possibly not even Samsung drive?

Not to defend Samsung or anything. I bought a 1TB 870 EVO myself a little less than a year ago which it seems I now need to return and hope for a non-defective replacement, and an eventual refund if the replacement turns out to have the same problem.

 

[E.S]

Since 0xFC / 252 is officially a "vendor specific" attribute, are there really any guarantees it means exactly the same thing on these current Samsung drives with the new firmware, as somebody somewhere on the Internet years ago determined it means on their possibly not even Samsung drive?

Not to defend Samsung or anything. I bought a 1TB 870 EVO myself a little less than a year ago which it seems I now need to return and hope for a non-defective replacement, and an eventual refund if the replacement turns out to have the same problem.

Samsung has not explained anything about this big problem.
that's abnormal.
We should suspect Samsung as much as possible.

 

[JagCube]

Samsung has not explained anything about this big problem.

Yes, that would certainly be appreciated. Also them telling what exactly 0xFC means and providing a changelog for the firmware would be very, very nice.

We should suspect Samsung as much as possible.

However, if you check the SMART attribute descriptions used by CrystalDiskInfo you can already find an alternate description for attribute 0xFC (see under the various SmartMicron headings):

"FC=Total NAND Read Plane Count (High 4Bytes)"

It seems my old 512 GB Intel-branded SSD in one of my PCs also reports a 0xFC SMART attribute which is currently at 0x22 (34 dec) which I have no idea what it might be, and whether it was lower when the disk was new. It could even plausibly be a temperature value (34°C). I can at least say I've never had any problems whatsoever with that disk, and all the error counters and reallocated sector count are still at zero after many years.

 

[E.S]

Presumably　The update to firmware "SVT02B6Q" is intended to fix our problem.
The addition of "FC" is of course also relevant.
"Total NAND Read Plane Count" is irrelevant to this problem.
It would be reasonable to interpret it as "Newly Added Bad Flash Block"
I suspect that they are trying to avoid RMT by replacing fatal values SMART ID "05" (Reallocated Sector Count) and "183" (Runtime Bad Block) with "FC".

 

[JagCube]

Presumably　The update to firmware "SVT02B6Q" is intended to fix our problem.
The addition of "FC" is of course also relevant.
"Total NAND Read Plane Count" is irrelevant to this problem.
It would be reasonable to interpret it as "Newly Added Bad Flash Block"
I suspect that they are trying to avoid RMT by replacing fatal values SMART ID "05" (Reallocated Sector Count) and "183" (Runtime Bad Block) with "FC".

If the intention is to hide the existence of broken sectors from users, why have any externally visible counter for them at all? The drive could still internally keep track of them for remapping without ever telling the user anything about them. And why pick 0xFC for it when there's "prior art" of using this attribute number for a bad sector count?

They cannot hide the disks losing sectors forever anyway because eventually the user is going to notice that some files they stored earlier cannot be read anymore. Unless the new firmware just ignores uncorrectable read errors and returns garbage data in such a case? (I wouldn't know, I've not used it, my EVO came with the old firmware from factory and I never updated it)

I don't know how things would work out outside in countries with less legally mandated consumer protection, but I'm pretty sure that here I could get the drive exchanged or refunded regardless of any SMART counters if some tens to hundreds of sectors just become unreadable. Especially if the new firmware still stores uncorrectable read errors in the SMART log and fails the SMART self test on such unreadable sectors, but possibly even just an OS always reporting I/O errors when reading certain sectors could be enough.

edit: I am of course lucky enough to be in a situation where I can deal with the retailer (less than 2 years from purchase date). Probably things wouldn't go so well if I had to deal directly with Samsung.

I certainly hope if the retailer sends me another 870 EVO that if it's going to develop this problem, it'll develop it before the 2 years are up so I can just ask them instead of Samsung for a refund.

btw. here are the details of my broken 1TB EVO:
- Manufactured 2021.07
- Made in China
- PN MZ7L31T0HBLB
- Model MZ-77E1T0
- Factory firmware SVT01B6Q
- Serial S6PUNF0R[...]

 

[E.S]

If the intention is to hide the existence of broken sectors from users, why have any externally visible counter for them at all? The drive could still internally keep track of them for remapping without ever telling the user anything about them. And why pick 0xFC for it when there's "prior art" of using this attribute number for a bad sector count?

I don't know.
Maybe samsung wants the data.
The 870evo is Samsung's first 1xxL 3D TLC NAND. You may want to collect data using users as test subjects.
Also, most people don't really care about the "FC" value.
In fact, no one on any forum pointed out that "FC" is "Newly Added Bad Flash Block" until I pointed it out here.
There was no one.

 

[CiTay]

Now it gets a bit weird, are you suggesting that Samsung is receiving all SMART data as telemetry over the internet or something?

Yes, you are the only one repeatedly claiming that FC is related to new bad blocks. Sadly, an unproven claim doesn't become more well-founded by repeating it a lot.

I don't want to exclude the possibility that this FC value means something negative going on internally. But i would actually have to see that become reality for someone, it really preceding a failure with actual bad sectors on the drive. That's the point where i would become worried about the FC value. Otherwise, it could mean whatever.

The other day i installed a Verbatim Vi550 S3 SSD in someone's old notebook. I'm not exaggerating, it had like 30+ SMART IDs/values, and only a handful had a description, all the rest were vendor specific!

 

[E.S]

They cannot hide the disks losing sectors forever anyway because eventually the user is going to notice that some files they stored earlier cannot be read anymore.

It may be Reallocated before it becomes unreadable, and then retired as a bad block.

Sadly, an unproven claim doesn't become more well-founded by repeating it a lot.

You said that early batch was the cause without any evidence.
It has led many users in the wrong direction.
Although samsung may be pleased.

 

[CiTay]

You said that early batch was the cause without any evidence.
It has led many users in the wrong direction.

It was the best information i had at the time, according to the cases i saw. I would very much like to edit my first post for various reasons, but the forum doesn't allow it.

 

[Daytrader]

I got this problem on 2 of my Samsung 870 4TB SSD, made date feb 21 and sept 21, made in korea, trying to get rma accepted, nightmare proof they need.

​

 

[Captain Danger]

TLDR: It usually happened when the Total Host Writes ≈ 10TB, most affected SSDs are manifactured in 2021 (so the eighth digit is mainly R), but there still exists a few cases in 2020 (i.e. the eighth digit is N) and January 2022 (i.e. the eighth digit is T, the ninth digit is 1)

Do you have any data as to when it crossed 10 TB and failed, how many of them were from each capacity? (250GB, 500GB, 1TB, 2TB and 4TB.)

 

[Mussels]

It was the best information i had at the time, according to the cases i saw. I would very much like to edit my first post for various reasons, but the forum doesn't allow it.

If you report your own post you can ask for permissions to be altered to allow edits

 

[Tomgang]

Aw, I just buying such a drive today. Now I'm not sure if I should return it.

Any one knows if this issue has been fixed or still effect ssd produced later?

 

[kevin335200]

Do you have any data as to when it crossed 10 TB and failed, how many of them were from each capacity? (250GB, 500GB, 1TB, 2TB and 4TB.)

This speculation comes from the observation of a large number of CDI screenshots, I do not yet have spare time to create a sheet for accurate statistics, but in general I don't feel a strong correlation with the capacity (almost all of the screenshots are somewhere between 7T - 15T). As well, the expression around 10T is not really precise, it is only an approximate confidence interval from my observation. I feel like it could be a little bit higher, but no more than 15T.

Also, this is only the data when the user found the problem and does not represent the amount when the failure started.

 

[Daytrader]

This speculation comes from the observation of a large number of CDI screenshots, I do not yet have spare time to create a sheet for accurate statistics, but in general I don't feel a strong correlation with the capacity (almost all of the screenshots are somewhere between 7T - 15T). As well, the expression around 10T is not really precise, it is only an approximate confidence interval from my observation. I feel like it could be a little bit higher, but no more than 15T.

Also, this is only the data when the user found the problem and does not represent the amount when the failure started.

This is on only 2021 ssd's, or are the 2022 ssd's still have this problem ? thx

 

[kevin335200]

This is on only 2021 ssd's, or are the 2022 ssd's still have this problem ? thx

I have only collected 1 case from January 2022 so far, so it could be outlier. After this time point the problem might have been solved, however, Samsung does not have any official statement yet. Only time will tell.

 

[ThrashZone]

This is on only 2021 ssd's, or are the 2022 ssd's still have this problem ? thx

Hi,
Avoid 870 evo.

 

[P4-630]

This is on only 2021 ssd's, or are the 2022 ssd's still have this problem ? thx

You could get an 860 evo instead.

 

[Daytrader]

I have only collected 1 case from January 2022 so far, so it could be outlier. After this time point the problem might have been solved, however, Samsung does not have any official statement yet. Only time will tell.

Thx, i have one other 870 4TB dated sept 2021, only just started using, like 3TB written to it so far, i reckon this will go bad also, also made in korea like others, i got accepted to rma others, but sent all info via link they give you, and says they contact you within 1 to 2 days, already been 3 days and nothing, is this normal ? thx

Hi,
Avoid 870 evo.

Yeh i might, or go with a different make all together, thx

You could get an 860 evo instead.

Yeh i will look into them, i have a 850 that is 7 years old and my main ssd, and not one error on that, thx

 

[E.S]

This is on only 2021 ssd's, or are the 2022 ssd's still have this problem ? thx

Please read my past posts.
2022 ssd's have new firmware.
There is another issue with this new firmware, which is an increase in the SMART ID "FC" value.
"FC" is a value related to bad blocks according to wikipedia, so increasing this value may be a problem.
Perhaps even the new firmware doesn't fundamentally fix the problem.

I have only collected 1 case from January 2022 so far, so it could be outlier. After this time point the problem might have been solved, however, Samsung does not have any official statement yet. Only time will tell.

why are you ignoring my past posts?
The problem is probably not resolved.

 

[Daytrader]

Please read my past posts.
2022 ssd's have new firmware.
There is another issue with this new firmware, which is an increase in the SMART ID "FC" value.
"FC" is a value related to bad blocks according to wikipedia, so increasing this value may be a problem.
Perhaps even the new firmware doesn't fundamentally fix the problem.


why are you ignoring my past posts?
The problem is probably not resolved.

Oh, i did not realise the problem was the firmware, thought it was the drives themselves, my 09 2021 drive i have only just started using, with latest firmware, so you saying that 09 2021 drive will hopefully be ok ? thx

 

[parityerror]

I bought 4 pcs of 870 EVO 1 TB in June 2021. The first one failed now with about 3 TB written, can no longer read some of the files on the disk, there are bad blocks and lots of read errors in the smart report. Disk manufacture date is 2021.01.

I don't have any windows machines, so I can't use the Samsung tool to update the firmware of my remaining drives. I found some instructions for downloading and extracting the Samsung fumagician software, but it just crashes on my machine. The firmware file (SVT02B6Q.enc) was included with fumagician, but the linux fwupdmgr command doesn't understand the file format.

Is there a way of updating the SSD firmware on linux - one that actually works?

I will post here if I manage to update the fw... Samsung sure isn't making this easy.

 

[E.S]

Oh, i did not realise the problem was the firmware, thought it was the drives themselves, my 09 2021 drive i have only just started using, with latest firmware, so you saying that 09 2021 drive will hopefully be ok ? thx

The root cause of the problem may lie in the drive itself.
The new firmware may appear to fix the problem at first glance, but it may actually be just masking the problem.
We have to note the newly added SMART ID "FC".