What's new

Careful: Samsung EVO 4TB SSD (high failure rates)

MegaPixel

Developer by day, DAW tinkerer by night.
Hi,

Just letting people know, be careful when your buying a large size SSD, seems Samsung (who also sell their chips to other manufacturers to use in their products) are doing sneaky things like using the good (nand?) chips for the early part of the storage, but the cheaper or even B grade stuff for the last part of it.

I bought a 4TB EVO in August 2021, it failed after 1 month. Amazon sent a replacement and now this one is failing also.

I've not had any issues with any of my 1TB or 2TB EVO SSDs, rock solid and some 5+ years old and been driven every day 24x7x365 but just an FYI on the 4TB and above SSDs, be careful, there's a lot of crap out there at the moment, especially with the chip / rare earth material shortage.

FYI
Samsung QVO Drives = 3 year warranty (budget buyers, cheapest parts etc)
Samsung EVO Drives = 3 to 5 year warranty (middle ground)
Samsung PRO Drives = 5 year warranty (supposedly the best you can get from samsung, bar going NAS SSD RED PRO)
 

Attachments

  • ssd-log-1.png
    ssd-log-1.png
    65.5 KB · Views: 73
  • ssd-event-log.png
    ssd-event-log.png
    73.9 KB · Views: 70
Did SSD fail completely or just part of it?

Could you post S.M.A.R.T. values of the drive? I have Samsung 860 EVO 4TB and I'm wondering if it is possible to predict SSD failure using SMART.
 
That's wonderful! I have a 4TB QVO for part of my sample libraries. And since all Samsung devices I've owned (two DVD players, a cell phone and a vacuum cleaner) have broken…

Paolo
 
Did SSD fail completely or just part of it?

Could you post S.M.A.R.T. values of the drive? I have Samsung 860 EVO 4TB and I'm wondering if it is possible to predict SSD failure using SMART.

Unable to post SMART results at the moment doing a full scan... It says it will take 6 hours, but for the last drive it took nearly 12...

However, SMART did show 2 failures on the previous drive, and there are 2 failure on the quick SMART scan on this one.

Luckily after last failure, I bought 2 x WD Black 8TB drives to ensure I didn't loose anything again. 1 internal frequent backup and for external infrequent backup.
 

Attachments

  • ssd-scan.PNG
    ssd-scan.PNG
    77.6 KB · Views: 40
My 870 EVO 2 TB SSD failed after 6 months with many bad blocks. SMART-Values 05, B3, B7, BB, C3 increased daily. Everything looked similar to the log posted by @MegaPixel. The drive was replaced by Samsung. There seems to be a problem with the 870 series, I can see similar feedback of other users in other forums. Hopefully only a specific batch is affected. I bought my drive mid 2021.

I have been using Samsung drives for many years and also in the same computer in which the failing one was operated. I have never had a single S.M.A.R.T. error before.
 
Last edited:
My 870 EVO 2 TB SSD failed after 6 months with many bad blocks. SMART-Values 05, B3, B7, BB, C3 increased daily. Everything looked similar to the log posted by @MegaPixel. The drive was replaced by Samsung.

There seems to be a serious problem with the 870 series. Hopefully only a specific batch is affected. Bought my drive mid 2021.

I bought the drive in August 21, it failed after a month. Amazon sent out a replacement but now this one is going also.
 
RE: Samsung Magician
The latest version of Samsung Magician I've not been able to get to work for over a year now, I have to use an older version of the software. Many have this problem, Samsung know about it but don't seem to care.

RE: RE: Samsung Magician
With my first 4TB Evo, the full system scan didn't show any bad sectors at all which is why I got my hands on HardDrive Sentinel. But there were 1000s of CRC errors reading and writing files. Samsung Magician only seems to be able to show SMART errors, I think the full scan is just a re-allocation (repair scan) so quite possibly they made it so that it wont show any errors anyway due to re-allocation of bad sectors. But as SMART failures are occurring on reads and writes it will just keep re-allocating and flagging those sectors as non usable. But it does have a CSV export which gives a bit more info, so that boxes thing is not really helpful, just a big progress bar lol
 
Just for you information, in my full diagnostic scan of the drive with the Samsung Magician Tool, there were many red blocks visible in the graph (bad blocks). And they never went away, not even after repeated runs.
 
Is it possible there were some defective NAND flash chips from the start and errors started showing up when drive got filled up to a point where defective cells were located?
 
seems Samsung are doing sneaky things like using the good (nand?) chips for the early part of the storage, but the cheaper or even B grade stuff for the last part of it.
Following that logic, wouldn't the QVO drives be showing more frequent failure rates? Or is the problem only the controller hardware? I bought my 4TB QVO in mid July; it's about half full and I have not had any problems with it yet.
 
Just for you information, in my full diagnostic scan of the drive with the Samsung Magician Tool, there were many red blocks visible in the graph (bad blocks). And they never went away, not even after repeated runs.

Interesting, I've never seen any errors on the full scan but the export after it has shown me errors. EG. HardDrive Sentinel screen shot is catching them as Samsung Magician is doing its scan.


Is it possible there were some defective NAND flash chips from the start and errors started showing up when drive got filled up to a point where defective cells were located?

I'm at 70 to 75% full, I try never to fill an SSD past 80% or 85% at most. I had about 5 crucial SSDs fail on me when filled past 90%.

Following that logic, wouldn't the QVO drives be showing more frequent failure rates? I bought my 4TB QVO in mid July, it's about half full and have not had any problems with it yet.
Could be a number of factors, QVO is lower write rating and I think a different kind of chip (stacking), and probably a different kind of controller to handle read/writes. So all could be fine with them but they really screwed something up on the larger 870 EVOs. See screen shot, my other EVOs are rock solid and taken way more of a hammering than the 4TB one has.
 

Attachments

  • ssd-log-2.png
    ssd-log-2.png
    73.1 KB · Views: 14
Thanks for the warning!
A substitute maybe the PNY XLR8 CS3040, it is fast enough, excellent TBW and good price +- $520

4TB Model M280CS3040-4TB-RB
Total Terabytes Written (TBW) 6800TB
Mean Time Between Failures (MTBF, hours) 2,000,000
DWPD 0.93DWPD

 
I'm at 70 to 75% full, I try never to fill an SSD past 80% or 85% at most. I had about 5 crucial SSDs fail on me when filled past 90%
Filling it up once completely after formatting it for the first time, might help detect defective chips. However, this does use up one write of each cell.
 
Thanks for the warning!
A substitute maybe the PNY XLR8 CS3040, it is fast enough, excellent TBW and good price +- $520

4TB Model M280CS3040-4TB-RB
Total Terabytes Written (TBW) 6800TB
Mean Time Between Failures (MTBF, hours) 2,000,000
DWPD 0.93DWPD


I just ordered a WD Black 8TB from Amazon.

I will keep using the smaller SSDs and my NVME for boot but will wait 3 or 4 years till PCIE Gen 4 mobos are more mature (and cheaper). As well as incorporating the newer architecture of direct access (like how I think the ps5 or xbox latest works for cpu to storage). And hopefully there are 4 or maybe 5 NVME PCIE Gen4 slots which can all run at max speed on the motherboard. And And lol DDR5 will have matured by then also. I do a full machine update every 5 years, often excluding the drives but I think this time I will include the drives and just keep my case. It's going to be an expensive one but it's work and it's my entertainment system also.

But if I do buy another SSD it will probably be the PRO series from now on, or maybe it's time I bite the bullet and get a 6 Bay NAS on the go with 10GB ethernet and buy a 10GB ethernet card for my machine also. Problem is samsung typically sell their chips to other manufacturers.


Filling it up once completely after formatting it for the first time, might help detect defective chips. However, this does use up one write of each cell.

Once this full scan is done, I have a full backup so I will 100% max out the drive with some script just duplicating files till its full to see what falls over, then run a CRC check on all files and see how many fail. Either way, it's a phone call to Amazon in a few days to see if they will send out a replacement, if not it's an email to Samsung.
 
870s were mentioned. Is it safe to assume 970s might also be affected?
My two 970s are still running very fine ;)
I don't think 870s are generally bad. I more think there could have been an error in the production process at a specific time, bad wafers, whatever. They give you 5 years warranty, so they seem to be quite confident about the stability of this product.
 
Last edited:
I had a 1TB Samsung 860 QVO fail on me after a couple of months. Got a refund from Amazon. I have a number of T5s and they are all good (so far!).
 
My two 970s still running very fine ;)
I don't think 870s are generally bad. I more think there could have been an error in the production process at a specific time, bad wafers, whatever. They give you 5 years warranty, so they seem to be quite sure about the stability of their products, normally.

My 1TB EVO 850 is running great (very old, was my work drive, now is mirroring my main work drive and is still going strong). And the smaller 970 EVOs are fine too, but not sure I will trust another 870 EVO 4TB again... Not with work data anyway. Maybe sell it...

I have a number of T5s and they are all good (so far!).
I will have to have a look into the T5s and some others, see where the chips are coming from, 3D nand, the controllers etc.
 

Attachments

  • 850.png
    850.png
    4.3 KB · Views: 10
Thanks for the suggestion @Pictus! Any other brands you’d recommend?
You are welcome.
It is more of model than brand, for 4TB with higher TBW: (PNY 3040 is better)

Sabrent Rocket 4TB (GEN3)
Model = SB-ROCKET-4TB
MTBF = 1,800,000
TBW = 6070

Seagate FireCuda 530 4TB (GEN4)
Model = ZP4000GM30013
MTBF = 1,800,000
TBW = 5100

There are SSDs with very big TBW
https://basic-tutorials.com/news/adata-prospector-ssd-endurance-m-2-ssd-for-chia-mining-unveiled/
But if it is a drive you fill with libraries and only keep reading, there is no need to worry about TBW.
 
Top Bottom