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

[chrcoluk]

Just FYI, we've tried RMAing multiple drives here in Europe, it's a joke. The disks started failing, we pulled them out of the server, stuck them on a desktop machine, ran Magician *before* and *after* the firmware upgrade. SMART does say there are errors, but the full test from Samsung M. just shows all green. Which means they won't take disks for RMA...

Here's the list of things they ask you to provide *for each disk*:


• A brief description of the problem you have with your drive.
• Please provide us with screenshots of any error messages or error codes.
• Please provide us with a short video clip showing the problem you are
experiencing with your drive and any troubleshooting steps you may have
performed, if possible.
• A screenshot of the Samsung Magician Drive Details page
• A screenshot of the results of the Diagnostic Scan (Please note that the
Diagnostic Scan is only possible on the following models: 860 PRO, 860/870
EVO, 860/870 QVO, 970 EVO PLUS, 980 PRO, 980).
• The SMART test result from the Samsung Magician.
Samsung Magician can be downloaded via the following link: https://
www.samsung.com/semiconductor/minisite/ssd/download/tools/
Please note that the software is only compatible with Windows and the SSD
must be connected directly to the motherboard.
To obtain the SMART result, simply click on the 'SMART' button located at
the top-right corner of the Samsung Magician Drive Dashboard page.
• A copy of your proof of purchase*
• What is your country of residence?*
• Where was the SSD being used? Server, Nas, Desktop PC, etc?
• Are you an end-user or re-seller of this drive?*
• Photos of the front and backside of the entire SSD clearly indicating the
serial number on the label on the drive.**
• A close-up photo of the connector of the SSD.**

We have 20+ drives in production, but you can clearly tell they're not going to help out.

Your drives are passing the full LBA test? Everyone else (with the problems) seems to fail on that.

 

[pcmikeymike]

found this thread while searching for samsung 870 failures. purchased 17+ of them in 1st half of 2021 and installed for clients. last month i had to reinstall a laptop because of bad lba on drive in a critical windows file(laptop wouldn't boot anymore). last week i was working for another client where i installed 10 of these drives, clearing off an old user's profile from a pc and got source read errors. ran magician and found a gang of bad lba's. then i got curious and ran magician on all of the 870's i installed there. 7 of 10 have bad lba's. none of them have been able to be cloned so i'm reinstalling each pc on a new drive, about 3 to 4 hours each to reinstall, load software, get everything signed in, get it to where the user is happy with it. called samsung and they emailed a form to fill out with all of the info. going to try to get them to replace all 17 drives since i don't trust them anymore but they said that their techs will test drives they receive and send back the good units and repair the rest. grateful that prices shot up on samsung drives after a couple months of buying these and i started to get sk hynix drives. will start tracking down the other 6 drives this week to notify the owners that they'll need to be replaced asap. i've always paid the extra money for samsung reliability and have 3 of them in my 2 pc's at home right now and have purchased at least a couple hundred 850/860/970/980's over the past 4 years-very disappointed at the moment.

 

[2BitNandBand]

found this thread while searching for samsung 870 failures. purchased 17+ of them in 1st half of 2021 and installed for clients. last month i had to reinstall a laptop because of bad lba on drive in a critical windows file(laptop wouldn't boot anymore). last week i was working for another client where i installed 10 of these drives, clearing off an old user's profile from a pc and got source read errors. ran magician and found a gang of bad lba's. then i got curious and ran magician on all of the 870's i installed there. 7 of 10 have bad lba's. none of them have been able to be cloned so i'm reinstalling each pc on a new drive, about 3 to 4 hours each to reinstall, load software, get everything signed in, get it to where the user is happy with it. called samsung and they emailed a form to fill out with all of the info. going to try to get them to replace all 17 drives since i don't trust them anymore but they said that their techs will test drives they receive and send back the good units and repair the rest. grateful that prices shot up on samsung drives after a couple months of buying these and i started to get sk hynix drives. will start tracking down the other 6 drives this week to notify the owners that they'll need to be replaced asap. i've always paid the extra money for samsung reliability and have 3 of them in my 2 pc's at home right now and have purchased at least a couple hundred 850/860/970/980's over the past 4 years-very disappointed at the moment.

That's pretty rough. I wonder how wide that window of producing defective units really was.

I've got this 4TB 870 here that was produced in Nov 2021, and it's still in the store return window. Not sure if I'd be overly paranoid in exchanging it for an MX500.

 

[Tetras]

That's pretty rough. I wonder how wide that window of producing defective units really was.

I've got this 4TB 870 here that was produced in Nov 2021, and it's still in the store return window. Not sure if I'd be overly paranoid in exchanging it for an MX500.

If it isn't showing any symptoms (reserved blocks used or ECC errors) and survives a scan in magician (assuming you haven't formatted it since installation) then I'd feel fairly safe, for now.

 

[floogulinc]

I have three 870 EVO drives all made in January or February 2021 (one of them has a SN just +4 of the other), they were all purchased 2021-04-23. All three have been throwing bad block errors in Windows event viewer as data is read and show high uncorrectable error count in SMART data. I am unable to run Samsung Magician because the drives are in a Windows Storage Spaces mirrored array and it won't recognize them.

S620NG0R108XXX (2021-01-24)	S620NG0R201XXX (2021-02-08)	S620NG0R201XXX (2021-02-08)

According to Event Viewer they started producing bad blocks in mid December. Thankfully Storage Spaces has been able to silently fix these bad blocks using the mirrored data so I haven't actually lost anything, my main application for these drives stores files with SHA256 hashes so I was able to verify all files were still correct.

I am of course still trying to take these out of service as soon as I can and hopefully I will be able to RMA them. I specifically spent extra on these drives for Samsung's reputation for reliability. I see this as an extreme failure on Samsung's part and I doubt I will buy their storage products again.

 

[Pwnstix]

Samsung got back with me and are supposed to be shipping a replacement 870 to me soon. I don't think they took very much time inspecting my previous drive.

 

[aaronnoraa]

Have two 870 EVO 1TB drives mfg dates 2021.6 and 2021.4, both purchased from different local big box stores (Office Depot, Best Buy).

Both have the same issues here, hundreds of thousands of ECC and Uncorrectable counts. Spent 10 hours of work/waiting in reallocating sectors with HDAT2 to minimize lost data and force cloning in MediaTools Pro (nothing else would clone even after full disk scans which should reset unreadable sectors, new ones appeared so quickly Macrium, Acronis and old Ghost Corporate all failed even with verification OFF), Media Tools ended up taking 7.75 and 9.5 hrs to clone the data off those drives after lowering retrys to 5 with forced resume.

Used in an office environment meaning the lightest load possible both drives less than 2 TB written, after looking at Windows logs the bad blocks/read errors were going back to 2 months after purchase for one of them, both skyrocketed over the last week for unreadable sectors/bad blocks.

Not looking forward to getting any mfg date 870 EVO drive as replacement from the RMA since I absolutely do not trust these drives anymore (company just let me purchase new SSDs immediately for replacements to get the workstations back up, RMA doesn't mean jack in the real world for companies which cannot wait two weeks nor even a day of a downed workstation waiting for RMAs to come back), really feel like either a refund or a different model is appropriate for these RMAs.

This is clearly a big problem but no way in hell is Samsung going to admit to it.

 

[Tetras]

Have two 870 EVO 1TB drives mfg dates 2021.6 and 2021.4, both purchased from different local big box stores (Office Depot, Best Buy).

Wow, much newer drives than most of the ones with issues Starting to wonder if stores should even still be selling these.

 

[BSim500]

I have three 870 EVO drives all made in January or February 2021 (one of them has a SN just +4 of the other), they were all purchased 2021-04-23. All three have been throwing bad block errors in Windows event viewer as data is read and show high uncorrectable error count in SMART data.

Have two 870 EVO 1TB drives mfg dates 2021.6 and 2021.4, both purchased from different local big box stores (Office Depot, Best Buy). Both have the same issues here, hundreds of thousands of ECC and Uncorrectable counts.

Wait a minute, so whatever manufacturing defect there was went on undetected for 6 months?!? That's one hell of a lot of drives affected...

 

[aaronnoraa]

Was over checking out a Reddit thread about this here and came across this comment and thought it should be quoted here to help illustrate the frequency of these failures-

3RAD1CAT0R

I bought 14x 1TB 870 EVOs mid November 2021. 7 of them failed on the same day about a week ago with varying amounts of uncorrectable sectors. All about 850GB written, 90 days of power on time at time of failure. 3 in one 4 drive raid 0, and all 4 in another 4 drive raid 0. The remaining 1 survived a 2tbw stress test, so I'm gonna assume it's fine. Moved all content off the other 6 I have just in case, still need to stress those a bit.

Who went on to say-

Judging by the manufacture date code in the serials, all were manufactured in September of 2021.

Wait a minute, so whatever manufacturing defect there was went on undetected for 6 months?!? That's one hell of a lot of drives affected...

Wow, much newer drives than most of the ones with issues Starting to wonder if stores should even still be selling these.

Looking like at least 9 months now.

 

[Jiginata]

Purchased September 26th 2021. 870 Evo1TB Serial: S5Y2NJOR101****


Check out the legendary "Samsung reliability" 870 Evo heap of junk in all it's glory. Can't even handle 6 TB's of data writing before it falls on it's arse. All marketing, zero substance.


Can anyone recommend me a a battle tested reliable 1TB SSD from a non-scam company? Preferably one that doesn't die after 6 months from general home use and please don't recommend me another Samsung SSD because I will never be buying an SSD from this company again.


THIS DRIVE SHOULD BE TAKEN OFF THE MARKET.

 

[Captain Danger]

Purchased September 26th 2021.

Manufacturing date?

Can anyone recommend me a a battle tested reliable 1TB SSD from a non-scam company?

Crucial MX500.

 

[ThrashZone]

Hi,
Think crucial has it's own issues

Sammy is fine for other series
870 evo I'm surprised it hasn't been recalled
I have several 860 evo/ 850 pro working just fine.

Look at western digital

 

[Pwnstix]

My 1TB 850 Pro was the first SSD I'd ever used. I'm still using it as my OS drive on a recent reinstallation. It's been a real workhorse since I first installed it in January 2018: 27 TBW, 97% wear leveling/lifetime remaining.

I've also had a 1TB 970 EVO NVMe in use since 2020 and it's been rock solid.

The 4TB 870 EVO was my first SSD over 1TB and I'd just assumed that Samsung's reliability and my own luck with them so far would hold over into the 870. Hopefully the replacement I've got coming in will hold up...but I'm definitely not putting any critically important/personal data on it this time.

 

[chrcoluk]

Samsung have handled this wrong in my opinion, I would have increased faith in the product if they announced a public recall, at no cost to the consumer and stated they fixed the issue. Instead silence and a firmware which has no changelog.

Curious if 870 EVO will get replaced with a plus model.

MX500 for anyone buying one I suggest you check if you have the revised models, as the earlier one's have their own issues with massively accelerated erase cycles and potential corruption, both my MX500's now dont work properly.

My 870 EVO still clean on SMART but its only got 688 power on hours, still 0 erase cycles on 356GB host writes so at least doesnt have crazy erase cycles like the MX500.

 

[ThrashZone]

Hi,
MX100 was the last good version still have some in operation and even it had bad firmware where linux will kill them in short time because trim never runs.

 

[Yraggul666]

Mine came in last week, box says fab date november 17 2021.
Haven't connected it yet, it's still in the box, i'll put it on when i upgrade the PSU, CPU, cooler, etcaetera.
With my luck it will be one from the bad batches, i wish i was wrong about that tho...

 

[BSim500]

MX500 for anyone buying one I suggest you check if you have the revised models, as the earlier one's have their own issues with massively accelerated erase cycles and potential corruption, both my MX500's now dont work properly.

Just curious but is that stuff related to this?

Crucial MX500 mit Firmware „M3CR033“ (2020) im Test

Seit 2017 wird die Crucial MX500 empfohlen, dabei wurden jüngst Firmware und Hardware angepasst. Überzeugt die Revision „M3CR033“ weiterhin?

www.computerbase.de

From what I understand the original MX500 was a full sized board fully populated with 16x chips, had a SM2258H controller and an old M3CR00x-M3CR02x firmware but at some point they updated the hardware to a much shorter board with 4x higher capacity chips, different SM2259H controller and a new M3CR033 firmware? I have the new 2TB M3CR033 and it's still 100% lifespan after 6 months so it seems unaffected. Given how much crap there is going on at the moment with buggy drives, bait & switches, etc, I'm highly reluctant to "upgrade"...

 

[iddqd]

My replacement drive just arrived. It was manufactured on 2022-03-24 in China according to the label on the box. It took exactly one week from pick up to delivery.

I am cautiously hopeful with it being brand new and manufactured in a different factory than the other drives before.

 

[chrcoluk]

Just curious but is that stuff related to this?

Crucial MX500 mit Firmware „M3CR033“ (2020) im Test

Seit 2017 wird die Crucial MX500 empfohlen, dabei wurden jüngst Firmware und Hardware angepasst. Überzeugt die Revision „M3CR033“ weiterhin?

www.computerbase.de

From what I understand the original MX500 was a full sized board fully populated with 16x chips, had a SM2258H controller and an old M3CR00x-M3CR02x firmware but at some point they updated the hardware to a much shorter board with 4x higher capacity chips, different SM2259H controller and a new M3CR033 firmware? I have the new 2TB M3CR033 and it's still 100% lifespan after 6 months so it seems unaffected. Given how much crap there is going on at the moment with buggy drives, bait & switches, etc, I'm highly reluctant to "upgrade"...

To this. Also a thread on toms hardware forums with far more info.

MX500 rapid erase cycles, seems to be poor wear levelling

Hi guys So I decided to open up crystal diskinfo on my laptop, I replaced the SSD with a brand new 500 GB MX500 earlier this year. Usually I use Samsung SSD's. The 94% health took me by surprise, as the SSD is not even half a year old, and the laptop is just running windows, no games, no...

www.techpowerup.com

From what I understand its the newest revisions which have it fixed, not the oldest one. Mine is on the newest firmware for its hardware revision which is M3CR023 https://uk.crucial.com/support/ssd-support/mx500-support. There is I think 3 hardware revisions where crucial decided not to change the SKU model number, like samsung does with say 860 to 870 e.g. Yours is the second revision which I think is ok. (I am curious if the first hw revision works ok on its original very old firmware, mine pre shipped with the newest firmware).

The issues with my MX500 is what led me to buy my 870 EVO only to then discover this thread lol.

 

[floogulinc]

The issues with my MX500 is what led me to buy my 870 EVO only to then discover this thread lol.

Exactly, when I bought my 870s a year ago I had originally purchased 3 MX500s but returned them for the Samsungs after reading about their issues. Obviously that hasn't turned out well lol

At this point I would probably replace the 870s with MX500s if I wasn't moving to enterprise NVMe U.2 drives.

 

[Chrispy_]

Ah shit, just spotted this thread and have a couple of 870EVO 2TB drives caching a QNAP rack NAS :\

 

[2BitNandBand]

Just got my replacement 1TB 870. Took about two weeks from shipping off the bad drive, so not too bad. The replacement is brand new, March 2022 manufacture date.

Definitely gonna keep an eye on my 4TB 870, though. Reading through this thread it seems drives manufactured as late as Sept 2021 are having issues, and my 4TB was manufactured Nov 2021, so not too much later. Also, mine came with earlier firmware (which I since updated). I wonder if it's possible bad firmware could cause the physical failure...probably not, but I don't know enough about how they work.

 

[Chrispy_]

I spec'ed a pair of 1TB 970 Pro for my NAS (MLC, high endurance rating) but the supplier suggested capacity mattered more than endurance so I went with the EVOs.
The NAS write cache is getting hammered right now as about 300TB of data is being seeded to it. If they're going to die, they'll probably do it soon. Once the data's seeded I doubt the SSDs will get much more of a workout than a regular consumer drive.

Manufacturing date was Nov 2021 so I might be okay?

 

[Tetras]

I spec'ed a pair of 1TB 970 Pro for my NAS (MLC, high endurance rating) but the supplier suggested capacity mattered more than endurance so I went with the EVOs.
The NAS write cache is getting hammered right now as about 300TB of data is being seeded to it. If they're going to die, they'll probably do it soon. Once the data's seeded I doubt the SSDs will get much more of a workout than a regular consumer drive.

Manufacturing date was Nov 2021 so I might be okay?

970 or 870? The 970 m.2 drives are unaffected, as far as we know.