Discussion Samsung 870 EVO Not To Be Trusted

[fzabkar]

Here is an example where 4 x 500GB Samsung 870 EVO SSDs failed after writing less than 1TB of data.

Samsung 870 EVO Not To Be Trusted!
https://goughlui.com/2022/08/20/notes-ssdraid-recovery-samsung-870-evo-not-to-be-trusted/

There appear to be many similar stories. It seems like they are affected by bad NAND.

 

[USAFRet]

While there may indeed be an issue with the 870 EVO...

Why
Was
There
Not
A
Real
Backup
?

This was a production environment.
Slot in some other drive, recover the data, continue ops, and then diagnose/fix the RAID with the 870's.

 

[fzabkar]

Yes, a backup should have been done. Unforgivable.

Maybe this an example which supports the idea that one should use dissimilar models and vendors in a RAID, to maximise redundancy.

 

[USAFRet]

Yes, a backup should have been done. Unforgivable.

Maybe this an example which supports the idea that one should use dissimilar models and vendors in a RAID, to maximise redundancy.

Yep, that too.

 

[anticeon]

NVME/M2 SSD are mostly already in RAID mode (Multiple NAND combined in one PCB plate)
And since it using PCI-e and high bandwith ... Software level Raid basically a time bomb.
I think NVME need specific Hardware to handle to Very high bandwith of them.

 

[Colif]

There was a bad batch of Evo 870 in January, I can see reports of them from then, but I don't know about more recently.

 

[fishyjack]

Failures have been sprinkled through basically all 2021 batches. January seems to have been the worst one though. I had a 2tb made in October 2021 and it failed in the exact same way as a January 2021 2tb it replaced.

 

[fzabkar]

I'm starting to wonder about Samsung ...

It would appear that the 980 NVMe SSD is affected by a firmware bug which causes it to report bogus 84C temperature spikes on sensor #1. This results in unnecessary throttling. It happens under Linux and Windows.

https://us.community.samsung.com/t5...SD-980-heat-spikes-to-84-C-183-F/td-p/2002779
https://forums.unraid.net/topic/115536-nvme-overheating/page/2/

 

[fzabkar]

https://forum.hddguru.com/viewtopic.php?p=305999#p305999

If anyone is using 870 EVO, 970 EVO Plus, 980, 980 PRO, or PM981A SSDs, please update the firmware as soon as possible. These drives have been associated with premature degradation, and if you notice abnormal values in the Smart (OE) attributes of these drives, please back up your data promptly. Samsung has released corrected firmware updates, but I believe the primary issue stems from premature degradation of flash memory chip cells.

The operation of the corrected firmware may be similar to the failure patterns seen during the time of 840 EVO release. To avoid a large-scale recall, Samsung has added in the firmware logic: after a certain time, the controller rewrites cold data to new locations to mitigate data loss caused by cell degradation. However, in this mode, the actual write rate to the flash memory increases, reducing the flash memory's durability.