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

[80251]

Could samsung be engaging in a little planned obsolescence in their SSD designs? So these failures aren't so much a design flaw as a design feature?

 

[tpudine]

Have there been any cases of defective 870 EVOs manufactured in 2022 or later? (i.e. those that shipped with the newer SVT02B6Q firmware)
I read this thread and looked through the Github repo, and it seems every bad drive was made in 2021, came with SVT01B6Q firmware.

I know it takes a couple of months for the errors to manifest, but at this point can we safely assume the issue only effects drives manufactured in 2021?

 

[Nospheratu]

Not sure if anyone has posted this as I haven't gone through all pages of this thread. You can find it on the below Samsung page under Firmware
DOWNLOAD – Tools & Software

Interesting to see no other model in their range has a note under the model name of a revised manufacturing process and they have SSD's listed from model 470 to the 980...

For those of you that might not be able to see the screen-print,
"*The 870 EVO model will be manufactured with a revised V6 process starting November 2022."

 

[A&P211]

Hi,
Yeah sammy what is happening to you dips

I stopped buying samsung SSDs when they stopped at 2tb, I need more space. Today, 2tb is becoming more mainstream.

 

[keithtpuc]

I don't know if there's much value in another me too post, but yeah, I've been bitten by the 870 EVO bug as well.

Bought in November 2021 from Amazon. Model MZ-77E1T0 with a date code of 2021.09 on it.

First noticed an I/O error when reading a file in Ubuntu EXT4 Partition. Used dd_rescue to image the disk --- there were over 5000 areas throwing read errors. Managed to trace the block numbers back to inodes, giving me filenames of corrupted files. Ended up with about (10) corrupted files I actually cared about in a group of 1 million original files. I had a few backups, both local, and in the cloud. The real pain in the butt was the fact that I needed to a) determine if the saved files were corrupted or not, b) determine whether the corrupted files were backed up, c) determine which copies are the good copies, and d) merge all these different sources together. Useful utilities were GNU's ddrescue, beyond compare 4, hashdeep, and debugfs.

I was throwing around multi-hundred-GB zip files and directories, and I will say that despite reasonable hardware, including 10 gbps connections and 260MB/s spinning rust, it was a pretty hefty job. Part of the problem is that even with fast drives, SSDs, etc, when you're dealing with a million of anything --- IOPS is more important than throughput. Small overheads per file really add up.

This was also "silent corruption" in that given it was a boot drive, and powered on 24/7, there was no real notice that there were issues. This is pretty scary, and I'm starting to take some steps to alert me sooner to this happening. I manage a couple hundred TB at home, but the bulk of it is surfaced scanned and SMART monitored regularly. I've owned, let's call it, about (2) dozen SSDs --- and never saw a single failure before this.

I will say that Samsung has fallen off my go-to list because of this. Their email support went from being very responsive -- to ignoring me completely. I called 800-726-7864 and was able to get a UPS label emailed. I've had both Western Digital (although not many!) fail, and Seagates (a bit more often) fail but both do over overnight advanced replacement. A quick painless resolution to a warranty claim really helps.

Hope this post helps someone.... your posts here certainly helped me!

Thanks

P.S. Quick edit: My firmware was not what you see listed initially......I only upgraded after I moved the disk from Ubuntu to a Windows machine for inspection via CrystalDiskInfo and Samsung Magician. If the theory was an outdated corrupted firmware caused the issue, then it's likely I NEVER upgraded the firmware from the factory, until a problem was discovered.

 

[80251]

I stopped buying samsung SSDs when they stopped at 2tb, I need more space. Today, 2tb is becoming more mainstream.

My Samsung 870 QVO is a 4 TiB model, they even have a 8 TiB model. It's a SATA QLC though so it's much slower than any other Samsung design.

 

[Mussels]

"Colour me with sadness when within just a couple of days of buying the 990 Pro 2TB, I noticed that the drive health according to SMART data from both Samsung Magician and third party tools had dropped to 99%. For the record I have other Samsung SSDs with over 40TB written and still at 99% health 1.5 years later, so I knew this was not normal.
Within another day or so it had dropped to 98%, by this point I'd not even written 2TB to the drive. Fast forward a couple more days and the drive health was sitting at 95%."

Samsung refusing to acknowledge and replace 990 Pro SSD with rapid health drops [Update]

Multiple users around the web, including myself, have found that the 990 Pro, the latest and fastest consumer SSD currently, might actually have an issue with the reported drive health remaining.

www.neowin.net

that article seems really weird and written by someone killing their own drive

THEY TRIED TO WARRANTY BECAUSE IT DROPPED TO 99%


They also think that the drive would reset to 100% if it was "reset" which... what?

The author of this has no idea what they're talking about. For all we know they're runinng XP without TRIM and torrenting to the drive - we have no idea.

 

[A&P211]

My Samsung 870 QVO is a 4 TiB model, they even have a 8 TiB model. It's a SATA QLC though so it's much slower than any other Samsung design.

oh sorry, I meant to say only m.2 versions instead of sata drives. I do have a 8tb qvo version in my laptop, its a wonderful mass storage for movies.

that article seems really weird and written by someone killing their own drive

View attachment 280756


THEY TRIED TO WARRANTY BECAUSE IT DROPPED TO 99%


They also think that the drive would reset to 100% if it was "reset" which... what?
View attachment 280757


The author of this has no idea what they're talking about. For all we know they're runinng XP without TRIM and torrenting to the drive - we have no idea.

If they are using the drive to torrent, most of the time its only reading from the drive instead of writing to it. I dont think reading causes that much wear and tear to a SSD.

 

[80251]

@A&P211
How is your 8 TiB QVO holding up? Any problems (CRC errors, reallocated sectors etc.)?

I would've considered an 8 TiB QVO as backup storage if they weren't so expensive.

 

[kevin335200]

"Colour me with sadness when within just a couple of days of buying the 990 Pro 2TB, I noticed that the drive health according to SMART data from both Samsung Magician and third party tools had dropped to 99%. For the record I have other Samsung SSDs with over 40TB written and still at 99% health 1.5 years later, so I knew this was not normal.
Within another day or so it had dropped to 98%, by this point I'd not even written 2TB to the drive. Fast forward a couple more days and the drive health was sitting at 95%."

Samsung refusing to acknowledge and replace 990 Pro SSD with rapid health drops [Update]

Multiple users around the web, including myself, have found that the 990 Pro, the latest and fastest consumer SSD currently, might actually have an issue with the reported drive health remaining.

www.neowin.net

There's already one 0E issue case reported from a 990 PRO user:

Source: https://tieba.baidu.com/p/8165358667

Recently, a user tested several popular PCIe 4.0 SSDs, and the Samsung 980 PRO experienced speed drops during the game testing phase, which in turn resulted in 0E errors and partition loss, it is dead before the aging test. The data cannot be recovered using PC-3000.

This 980 PRO is manifactured in 2022.11

Source:

 

[A&P211]

@A&P211
How is your 8 TiB QVO holding up? Any problems (CRC errors, reallocated sectors etc.)?

I would've considered an 8 TiB QVO as backup storage if they weren't so expensive.

Its holding up just fine, I leave it with 200gb remaining. I only use it for storage of videos.

 

[Mixmax]

Created this account just for this discussion. My 870EVO bought in July -22. I noted some odd behavior going on in Crystaldiskinfo when it printed a mangled drive name and readings on the 870 after I had Win11 fresh installed on another SSD, a Vertex4 with 35746 hours, 54425 GB writes and 95% remaining life! Good stuff, I wonder how OCZ is faring now...

Anyways, I don't fully trust neither Crystaldiskinfo nor Samsung on this, but after a reboot everything looks fine on this ssd. My 4tb WD Red had the seek error value rising as I mistakenly had Robocopy retry copying the same one file multiple times- and I ended up formatting (the throughout type) that drive to ReFS. Sucks to hear lotsa people having problems and Samsung unwilling to come forward with their SSD longevity issues. I'm only an intermediate PC user at best but I hope this info helps narrow down the affected product batches.

 

[E.S]

I'm only an intermediate PC user at best but I hope this info helps narrow down the affected product batches.

"*The 870 EVO model will be manufactured with a revised V6 process starting November 2022."
This message indicates that the manufacturing process remained the same until November 2022.

Recently, a user tested several popular PCIe 4.0 SSDs, and the Samsung 980 PRO experienced speed drops during the game testing phase, which in turn resulted in 0E errors and partition loss, it is dead before the aging test. The data cannot be recovered using PC-3000.

This 980 PRO is manifactured in 2022.11

I thought that the cause of the problem was read disturb due to cell shrinks, so I was surprised that the error occurred in such a short time.

 

[HLLLLLLL]

Just had my 4TB 870 EVO's "FC" attribute increase by 1, to almost 400. I do not, however, have any other telling attributes go above 0. Do those ever increase anymore after the SVT02B6Q FW update?

Anyway, for somebody following such stuff more closely, how is the "WD Blue 3D NAND" series fairing? I'd like to have a go-to alternative in case I start to see issues.

 

[Ghan]

Samsung 990 Pro SSDs Report Rapid Health Degradation

Some drives lost 10% of health in just a few weeks.

www.tomshardware.com

Seem some Samsung 990 Pro users are experiencing some even worse issues at the moment with their drivers. Samsung responded to that issue quite quickly at least. Hope the issue with the 870 series get brought up too.

 

[80251]

Is Samsung even manufacturing 870 series SSD's anymore? It would seem to me they would want to bury this debacle.

 

[Smuzzle]

I Just got two 4TB 870's and the date is November 26th, so I'm wondering if it was made with that v6 process as they state on the firmware page. Anyone know how to check that and if so could that mean that they fixed the issue?

 

[tpudine]

Is Samsung even manufacturing 870 series SSD's anymore? It would seem to me they would want to bury this debacle.

Read up
"*The 870 EVO model will be manufactured with a revised V6 process starting November 2022."

 

[E.S]

Samsung is currently investigating the 990 issue.
Why do you keep ignoring the 870evo and 980 issue?

 

[Nospheratu]

My guess is that the 990 issue can be reproduced fairly quickly whereas the 870 seems to be a gradual degradation over time. Doesn't hurt that the 990 issue seems to have been picked up by some media outlets fairly quickly as well and bad press about your flagship product will make any company jump.

Perhaps someone with a fairly new defective drive on hand can reach out to tomshardware or gamersnexus, linking to this and the various other threads out there.

edit: In any event they might have the 870 Evo covered already with the new hardware revision in November 2022 as per my earlier post.

 

[E.S]

edit: In any event they might have the 870 Evo covered already with the new hardware revision in November 2022 as per my earlier post.

Why 870evo only new hardware revision?
What about 980?
I think the cause of both problems is probably the same.
Both 870evo and 980 have the same 128 layers.
The 870qvo has 96 layers and there was no problem.

 

[80251]

The 870 QVO really isn't relatable to this issue because it's QLC and vvvvvvvvvvveeeeeeeeeeeeeerrrrrrrrrrrrrrrrrryyyyyyyyyyyyy slow as a result.

 

[kevin335200]

My guess is that the 990 issue can be reproduced fairly quickly whereas the 870 seems to be a gradual degradation over time. Doesn't hurt that the 990 issue seems to have been picked up by some media outlets fairly quickly as well and bad press about your flagship product will make any company jump.

Perhaps someone with a fairly new defective drive on hand can reach out to tomshardware or gamersnexus, linking to this and the various other threads out there.

edit: In any event they might have the 870 Evo covered already with the new hardware revision in November 2022 as per my earlier post.

A heating table can be used to accelerate simulated aging test:

The 0E issue of 980 PRO is also very quickly reproduced in this video, which I don't think is a coincidence.

 

[E.S]

The data retention lifetime of NAND is shortened at high temperature.
If the 870evo were also subjected to the same aging test on a heating table, wouldn't it be possible to quickly reproduce the issue?

 

[AsRock]

that article seems really weird and written by someone killing their own drive

View attachment 280756


THEY TRIED TO WARRANTY BECAUSE IT DROPPED TO 99%


They also think that the drive would reset to 100% if it was "reset" which... what?
View attachment 280757


The author of this has no idea what they're talking about. For all we know they're runinng XP without TRIM and torrenting to the drive - we have no idea.

I personally don't trust smart, like over the last day 3 of my drives have lost 1-2%, i can understand one maybe or even 2 of them but 3.

2 of them being samsung and the other being my m.2 rocket 4 drive.