Yes, this bit me just last month. Around October 2022 I purchased 3 Samsung 870 EVO 2TBs for use in a RAID array. By January 2023, all three of them failed within a week of each other! Fortunately they failed one by one, so I was barely was able to recover my RAID array by pulling out one drive at a time, powering the computer off, and waiting for the RMA replacement to arrive. But imagine my shock to see one drive fail... only to replace it with an RMA... and then days later, seeing the next drive fail... and the next!
Doesn’t like it
It's quite possible. Some SSDs are worse offenders for this than others. I have some Samsung 870 EVOs that lost data the way you described. Samsung knew about the issue and quietly swept it under the rug with a firmware update, but once the data was lost, it was gone for good.
Doesn’t like it
This was quite an unpleasant read. Long wall of text on data recovery from HDDs and SSDs and how they changed... just to set the stage for what seems to be the intended message of the article - the assassination of Gibson and his product Spinrite? > And solid state drives... while there is no doubt performance degradation over their lifetimes, failure modes tend to be all-or-nothing. Not all drives are perfect until they suddenly die. In my experience, even modern-ish SSD drives (mainly Samsung 870 EVOs...) do develop problems continually. I've seen them both failing quite quickly, and also RAID controller rejecting several of them after some time of working well, then the drive still lived and could be used elsewhere, but in SMART data had bad numbers, including bad sectors (I know, it's SSD!). It's not all-or-nothing failure with SSDs. > So what about SpinRite? Yeah, what about it? Why spend so much time ( ours time) prepping ground for taking down something you do not really use, or have analyzed in detail? Why do you care? > he doesn't seem to have updated his website in some time... By association, I suspect that GRC's flagship product, SpinRite, doesn't get a lot of active maintenance either. What kind of a novel thought construction is that? These things need not be correlated. And if you checked, you would find he actually does update Spinrite, recently to 6.0 and 6.1. > It seems reasonable on the surface, but it wouldn't make much sense with a drive with inte…
Doesn’t like it
I just had a 2TB 870 EVO fail too! There were SMART errors about uncorrectable errors, and I saw CRC errors in the OS. I lost some data. This issue is all over the Internet, yet Samsung would not acknowledge it was a known issue. They also refused my RMA because the corner of one of the plastic port guides was chipped - we're talking about a miniscule chip, barely visible with the human eye. So despite the fact this drive was obviously defective, Samsung won't replace it. So I'm down £350 (£225 for the original drive, and £125 for the Crucial I had to buy to replace it), through no fault of my own. I'm not buying Samsung SSDs again - problems are one thing, but how you deal with them is paramount.
Doesn’t like it
I can live with my batch of 980pros and 870 EVOs failing. Bad batches happen to every manufacturer. However this seems like a systematic problem at Samsung. What really stinks is that I've been recommending Samsung EVOs and PROs to relatives and friends for some years. If I want to remain honest with myself I have to contact every one of them and have to run the SMART tests. So now there are NO reputable SSD manufacturers left. Only reliable SSDs are MLCs from mid 201Xs especially Intel and Samsung.
Doesn’t like it
In 2021 I bought at least ten 870 EVO 4 TB SATA and six 980 Pro 2 TB NVMe. All devices failed within 6 months on barely used systems. Find some smart data here: https://thomas.glanzmann.de/samsung/ > The pattern is always the same: I have them configured in a raid 1. Once a month debian does a raid check. During the raid check Debian reads all data from both devices. I get uncorrectable read errors. I no longer use Samsung SSDs and replaced them them with SSDSC2KB076T8, Micron SSDs and KC3000 Kingston NVMes. No failures since then. In 2021 I told a friend of mine about the issue. He also had a 870 EVO, issued a dd if=/dev/sdX of=/dev/null bs=8M and guess what, he got uncorrectable read errors. Due to running them in RAID 1 I caught the issue early and I had no data loss or downtime because the Linux software raid compensated for the bad hardware. However I replaced them in a hurry because I no longer trust Samsung SSDs. As you can see from the smart log they're barely used. Less than 4 months in service and 10 TB written. I also got uncorrectable read errors when evacuating data from the devices.
Doesn’t like it
Ars had a piece covering this as well, and I do wonder if there is something going on somewhere else in the Samsung stack, not just the NVMe 900 series line. Pure anecdote, but two years ago I did a NAS for a client using 24x 2TB Samsung 870 Evo drives (they'd gotten some incentive deal for it). While it was all one type vs mixed, there was the "luxury" of time because at that point getting the system they wanted together had a significant lead time. So I did ensure that the drives were purchased over the course of around 7 months, from multiple different reputable sellers (B&H, CDW, Provantage etc) in separate batches. System was solid, an Epyc 2 based SuperMicro server, running TrueNAS. And then last year with around 5500-7500 quite light hours of runtime (primarily reads, ~0.08 DWPD, well under official rating of 0.3 DWPD) drives started failing. These were definitely real failures, first indication came from regular automated ZFS scrubs and reporting increasing checksum errors and ATA errors. It was for so many drives and I'd always considered Samsung SSDs relatively reliable (even for consumer ones) that at first I thought it was a SATA controller failure, and our rep agreed and warranties back the server. They were great, gold plated support contracts pay off once in a while, and motherboard replacement and thorough testing later back in service. More drive problems. SMART short tests said everything was healthy, first longs did too. But then drives exceeded error limit…
Doesn’t like it
It appears Samsung isn't honoring their warranties. Is this happening to anyone else? I bought a Samsung 870 EVO with a supposed 5 year warranty but received an email that stated: "After carefully reviewing the Bill of Sale of your item, this device only has a year warranty starting from the date of purchase: 03/03/2021."
Doesn’t like itDoesn’t say they own it
> IMHO, generally speaking, Samsung SSD products are the best on the market. There seems to be 870 EVO reliability issue going on for years now [1]. Shit happens, what matters is how you handle it. Samsung has yet to issue any public statement on the matter. [1] https://www.techpowerup.com/forums/threads/samsung-870-evo-b...