ssd wear level counts...

[eccerr0r]

Necroing my old post just because.

I recently acquired some used NVMe SSDs and was wondering their longevity.

I had gotten a second hand WD Black and a second hand Samsung (IIRC, not handy right now).

The WD Black is a 250G, had 8.09TB written, and 5% used according to SMART. This calculates out to 31 erases on average and 600 cycle lifetime => It's a TLC or worse...

The (Samsung?) 500G, I had done a similar calculation and estimated the cycle life to be around 2000 cycles. Might be an "enterprise" drive compared to the former which must be a consumer/value model or something...

As for the Intel/Sandforce 180G it's unfortunately sitting idle at the moment. I can't seem to get lvm to start up the cache with it in my initramfs yet, will need to work on it more. Definitely do see a bit of a speedup using it as a cache to my HDD RAID5...

[NeddySeagoon]

eccerr0r,

A 600 erase cycle lifetime is about the warranty life of TLC FLASH.
That means that WD expect it to do a lot more.

QLC FLASH has warning label of a 300 rase cycle lifetime and in usually slower that TLC.

I've never seen anything above QLC advertised but I don't doubt they are in development,

[saellaven]

Just to add some more data. Mostly Samsung EVO drives including 3 NVME, along with a Kingston that is beaten and abused as the ccache and /var/tmp/portage whipping boy.

I also have a bunch of spinning rust that is used for slower access, backup, and diversification should the SSDs fail. That said, my oldest SSD is pushing 10 years of operation and I haven't had an SSD fail yet.


Home Desktop/Server

old / that's mostly retired but kept around as an emergency boot option
4.39 TB

Code: Select all

Device Model:     Samsung SSD 850 EVO 250GB
  9 Power_On_Hours          0x0032   083   083   000    Old_age   Always       -       83420
177 Wear_Leveling_Count     0x0013   099   099   000    Pre-fail  Always       -       12
241 Total_LBAs_Written      0x0032   099   099   000    Old_age   Always       -       9428457096

(steam A) 4.22 TB

Code: Select all

Device Model:     Samsung SSD 850 EVO 500GB
  9 Power_On_Hours          0x0032   086   086   000    Old_age   Always       -       70037
177 Wear_Leveling_Count     0x0013   099   099   000    Pre-fail  Always       -       8
241 Total_LBAs_Written      0x0032   099   099   000    Old_age   Always       -       9066481753

(steam B) 0.64 TB

Code: Select all

 
Device Model:     Samsung SSD 870 EVO 2TB
  9 Power_On_Hours          0x0032   097   097   000    Old_age   Always       -       12000
177 Wear_Leveling_Count     0x0013   100   100   000    Pre-fail  Always       -       0
241 Total_LBAs_Written      0x0032   099   099   000    Old_age   Always       -       1373110256

(steam C) 0.27 TB

Code: Select all

Device Model:     Samsung SSD 870 EVO 2TB
  9 Power_On_Hours          0x0032   098   098   000    Old_age   Always       -       6186
177 Wear_Leveling_Count     0x0013   100   100   000    Pre-fail  Always       -       0
241 Total_LBAs_Written      0x0032   099   099   000    Old_age   Always       -       573130840

NVME: (root)

Code: Select all

Model Number:                       Samsung SSD 970 EVO Plus 500GB
Available Spare:                    100%
Available Spare Threshold:          10%
Percentage Used:                    1%
Data Units Read:                    36,960,766 [18.9 TB]
Data Units Written:                 23,934,524 [12.2 TB]

NVME: (/home - this is a fairly new drive that hasn't seen much action yet)

Code: Select all

Model Number:                       Samsung SSD 990 EVO 2TB
Available Spare:                    100%
Available Spare Threshold:          10%
Percentage Used:                    0%
Data Units Read:                    8,143,142 [4.16 TB]
Data Units Written:                 4,588,232 [2.34 TB]
Power On Hours:                     2,897

Work Desktop/server
/var/tmp/portage and ccache

Code: Select all

Device Model:     KINGSTON SA400S37240G                                                                                 
User Capacity:    240,057,409,536 bytes [240 GB]                                                                        
  9 Power_On_Hours          0x0032   100   100   000    Old_age   Always       -       31689
231 SSD_Life_Left           0x0000   081   081   000    Old_age   Offline      -       81
233 Flash_Writes_GiB        0x0032   100   100   000    Old_age   Always       -       18030
241 Lifetime_Writes_GiB     0x0032   100   100   000    Old_age   Always       -       19877
242 Lifetime_Reads_GiB      0x0032   100   100   000    Old_age   Always       -       6153
244 Average_Erase_Count     0x0000   100   100   000    Old_age   Offline      -       197
245 Max_Erase_Count         0x0000   100   100   000    Old_age   Offline      -       208
246 Total_Erase_Count       0x0000   100   100   000    Old_age   Offline      -       266049

sdd /home 16.40 TB written

Code: Select all

Device Model:     Samsung SSD 860 EVO 1TB 
  9 Power_On_Hours          0x0032   091   091   000    Old_age   Always       -       42340                            
177 Wear_Leveling_Count     0x0013   099   099   000    Pre-fail  Always       -       15                               
241 Total_LBAs_Written      0x0032   099   099   000    Old_age   Always       -       35221617115

NVME: /

Code: Select all

Model Number:                       Samsung SSD 970 EVO Plus 500GB
Available Spare:                    100%
Available Spare Threshold:          10%
Percentage Used:                    1%
Data Units Read:                    37,541,182 [19.2 TB]
Data Units Written:                 36,141,732 [18.5 TB]
Power On Hours:                     4,350

[krumpf]

There was a series of articles about SSD endurance (couldn't fish them all out), the survivors died after writing more than 2 PetaBytes

https://techreport.com/review/the-ssd-e ... ter-1-5pb/

https://techreport.com/review/the-ssd-e ... petabytes/

https://techreport.com/review/the-ssd-e ... -all-dead/

[NeddySeagoon]

saellaven,

I've had three SSDs fail.
Two Samsung 1TB and a Crucial 250GB.
They all failed with bad blocks all over the place.

The Crucial 250GB was old and well out of warranty
The two Samsung were almost new. Not even 1TB written.

After experiencing Samsungs warranty process and duration, I will never buy Samsung again.

[eccerr0r]

My 180G SSD(SATA):

Code: Select all

233 Media_Wearout_Indicator 0x0032   100   100   000    Old_age   Always       -       0
241 Host_Writes_32MiB       0x0032   100   100   000    Old_age   Always       -       618218
242 Host_Reads_32MiB        0x0032   100   100   000    Old_age   Always       -       1778704
249 NAND_Writes_1GiB        0x0013   100   100   000    Pre-fail  Always       -       29499

Seems this means WA is like ... 52% ?

And since it's 180GB this means each block has been erased 163 times on average. As a MLC drive this should be 5% of its life?

The Samsung NVMe appears to be a 512G, 100TB written, 11% life used... this doesn't quite mesh anymore...

[NeddySeagoon]

eccerr0r,

The raw values are often not useful, so I don't know how to interpret that.
The CURRENT, WORST and THREShold values are normalised, so

Code: Select all

233 Media_Wearout_Indicator 0x0032   100   100   000 

says that there has been no media wear out.

Erase blocks and write brocks are not the same size. An erase block is a group of write blocks.
Its needing to move write blocks out of an erase block, because they are still needed, that leads to write amplification.

180G is a funny size. I suspect that there is some manufacturers over provisioning going on.
The entire (over provisioned) space takes part in wear levelling not just the 180G exposed to the user.