Intel DC S3500 Wearout of 001

[herby]

Just got two 480 GB DC S3500 SSDs off eBay with the following SMART data:

Disk 1

Vendor Specific SMART Attributes with Thresholds:
ID# ATTRIBUTE_NAME          FLAG     VALUE WORST THRESH TYPE      UPDATED  WHEN_FAILED RAW_VALUE
  5 Reallocated_Sector_Ct   0x0032   100   100   000    Old_age   Always       -       0
  9 Power_On_Hours          0x0032   100   100   000    Old_age   Always       -       8
 12 Power_Cycle_Count       0x0032   100   100   000    Old_age   Always       -       10
170 Available_Reservd_Space 0x0033   100   100   010    Pre-fail  Always       -       0
171 Program_Fail_Count      0x0032   100   100   000    Old_age   Always       -       0
172 Erase_Fail_Count        0x0032   100   100   000    Old_age   Always       -       0
174 Unsafe_Shutdown_Count   0x0032   100   100   000    Old_age   Always       -       8
175 Power_Loss_Cap_Test     0x0033   100   100   010    Pre-fail  Always       -       642 (1 397)
183 SATA_Downshift_Count    0x0032   100   100   000    Old_age   Always       -       0
184 End-to-End_Error        0x0033   100   100   090    Pre-fail  Always       -       0
187 Reported_Uncorrect      0x0032   100   100   000    Old_age   Always       -       0
190 Temperature_Case        0x0022   075   066   000    Old_age   Always       -       25 (Min/Max 23/37)
192 Unsafe_Shutdown_Count   0x0032   100   100   000    Old_age   Always       -       8
194 Temperature_Internal    0x0022   100   100   000    Old_age   Always       -       36
197 Current_Pending_Sector  0x0032   100   100   000    Old_age   Always       -       0
199 CRC_Error_Count         0x003e   100   100   000    Old_age   Always       -       0
225 Host_Writes_32MiB       0x0032   100   100   000    Old_age   Always       -       83824
226 Workld_Media_Wear_Indic 0x0032   100   100   000    Old_age   Always       -       133
227 Workld_Host_Reads_Perc  0x0032   100   100   000    Old_age   Always       -       54
228 Workload_Minutes        0x0032   100   100   000    Old_age   Always       -       488
232 Available_Reservd_Space 0x0033   100   100   010    Pre-fail  Always       -       0
233 Media_Wearout_Indicator 0x0032   100   100   000    Old_age   Always       -       0
234 Thermal_Throttle        0x0032   100   100   000    Old_age   Always       -       0/0
241 Host_Writes_32MiB       0x0032   100   100   000    Old_age   Always       -       83824
242 Host_Reads_32MiB        0x0032   100   100   000    Old_age   Always       -       100631

Disk 2

Vendor Specific SMART Attributes with Thresholds:                                                          [18/188]
ID# ATTRIBUTE_NAME          FLAG     VALUE WORST THRESH TYPE      UPDATED  WHEN_FAILED RAW_VALUE
  5 Reallocated_Sector_Ct   0x0032   100   100   000    Old_age   Always       -       0
  9 Power_On_Hours          0x0032   100   100   000    Old_age   Always       -       16236
 12 Power_Cycle_Count       0x0032   100   100   000    Old_age   Always       -       65
170 Available_Reservd_Space 0x0033   100   100   010    Pre-fail  Always       -       0
171 Program_Fail_Count      0x0032   100   100   000    Old_age   Always       -       0
172 Erase_Fail_Count        0x0032   100   100   000    Old_age   Always       -       0
174 Unsafe_Shutdown_Count   0x0032   100   100   000    Old_age   Always       -       61
175 Power_Loss_Cap_Test     0x0033   100   100   010    Pre-fail  Always       -       648 (88 397)
183 SATA_Downshift_Count    0x0032   100   100   000    Old_age   Always       -       0
184 End-to-End_Error        0x0033   100   100   090    Pre-fail  Always       -       0
187 Reported_Uncorrect      0x0032   100   100   000    Old_age   Always       -       0
190 Temperature_Case        0x0022   074   068   000    Old_age   Always       -       26 (Min/Max 22/43)
192 Unsafe_Shutdown_Count   0x0032   100   100   000    Old_age   Always       -       61
194 Temperature_Internal    0x0022   100   100   000    Old_age   Always       -       37
197 Current_Pending_Sector  0x0032   100   100   000    Old_age   Always       -       0
199 CRC_Error_Count         0x003e   100   100   000    Old_age   Always       -       0
225 Host_Writes_32MiB       0x0032   100   100   000    Old_age   Always       -       44311014
226 Workld_Media_Wear_Indic 0x0032   100   100   000    Old_age   Always       -       102400
227 Workld_Host_Reads_Perc  0x0032   100   100   000    Old_age   Always       -       4
228 Workload_Minutes        0x0032   100   100   000    Old_age   Always       -       973747
232 Available_Reservd_Space 0x0033   100   100   010    Pre-fail  Always       -       0
233 Media_Wearout_Indicator 0x0032   001   001   000    Old_age   Always       -       0
234 Thermal_Throttle        0x0032   100   100   000    Old_age   Always       -       0/0
241 Host_Writes_32MiB       0x0032   100   100   000    Old_age   Always       -       44311014
242 Host_Reads_32MiB        0x0032   100   100   000    Old_age   Always       -       2323260

As far as I can tell disk 1 was barely used, but disk 2 was completely dogged. 1.4 PB written if my math (and understanding of Intel's attributes) is right!

They were going to be a mirrored iSCSI target for VMs to replace some Crucial consumer disks but now I'm a little worried.

What are the odds that second disk has some legs left? I'd like a few years of service life.

 

[BackupProphet]

Disk 2 is almost dead! In fact I would consider getting a refund.

I bought a 600GB S3500 Last year with 33% health, been using it on my workstation daily for almost a year now. It has still 33% health left.

 

[keybored]

Still no reallocated sectors or program/erase failures though... That's surprising since these drives are only rated for 275TBW and this particular drive had almost 5x that amount written to it. If refund isn't an option it might still work as a desktop boot drive somewhere...

 

[Evan]

I wonder what they used it for during those 2 years, don’t that often see SSD’s abuses like that.

 

[Stux]

Well... one disk is brand news (8 hours) and the other has been flogged.

Surely this is what you sign up for when you buy second hand enterprise SSDs?

At least the media wearout indicator is still 0.

Might as well use it in mirror until it fails, then replace.

 

[herby]

That none of the attributes that indicate failure to me: Reallocated , Pending, or Uncorrectable have values is some what reassuring so it's not technically a bad drive.

That said the use on this thing is super high. So I'm torn, that kind of wear certainly wasn't what I expected, but it isn't DOA or even not as advertised. Maybe a little in bad faith though.

Oh, the ethical dilemmas of eBay shopping

 

[Jeff Robertson]

Just got two 480 GB DC S3500 SSDs off eBay with the following SMART data:

As far as I can tell disk 1 was barely used, but disk 2 was completely dogged. 1.4 PB written if my math (and understanding of Intel's attributes) is right!

They were going to be a mirrored iSCSI target for VMs to replace some Crucial consumer disks but now I'm a little worried.

What are the odds that second disk has some legs left? I'd like a few years of service life.

It's just getting broken in! In all seriousness even consumer drives will write WAY more than they are rated for, I think they managed multiple petabytes (6 rings a bell?) out of an old Samsung 830 in a test published a few years ago. I would expect you couldn't actually kill the thing if you tried.

 

[Evan]

Oh you can kill drives with out trying too hard ... all depends on what they are used for.
But most in general enterprise environment end up doing not much, I have a heap of SSD that get less drive writes in a month than they are rated for in a day.

 

[funkywizard]

I have plenty of used DC S3700's. Some came nearly untouched, some with tons of writes on them, but, being 3700's, even these still had above 80% health. Given the cost of 3600, 3610, 3700, and 3710's, I wouldn't touch a 3500 -- why bother? The S3500's are easier to come by now that the flood of better models has tapered off, but you can still get 3600's for cheap if you are persistent.

Really happy to get 3600's and higher. When we first started buying SSDs (Intel X-25M and Samsung 830) wear didn't seem like much of an issue. But now that users are accustomed to the performance of SSDs, server workloads have become way more punishing on the NAND. In the past, people would use SSDs as an upgrade from hard drives that were "just barely not fast enough" for their workloads. Now we have people putting workloads on single SSDs that would make an 8 drive raid 10 struggle.

It's not uncommon for us to see customers pile windows VMs on a server, allocating 512mb ram to each -- so low you'll see constant swapfile use at OS *idle*. The drives take a real beating in this environment.

I can't even count the number of failed and low-wear Samsungs (830, 840 Pro, 850 Pro) we've had to stop using -- over 100 easily. I can tell you we don't put low wear drives in customer servers. Even if they haven't failed, they'll be quite slow and may have random problems. The secondhand DC S3700's have been a game changer here -- the rated write lifetime is easily 20 times the Samsung 850 Pro. To put in context, a drive with wear 90 would have burned out an 850 Pro -- TWICE.

Once a drive starts to fail, it fails pretty spectacularly pretty quickly. In many cases you can't even read from the drives, or they won't even detect in the OS. Low wear is often the only warning you get for drive failure. Whereas we've had almost no failures on drives with at least 20 wear remaining, once you get below 10, you're playing russian roulette. I do agree that "Failed" drives (slow and/or flakey and/or low wear) that are still responsive to the OS do show a wide variety of writes when we retire them. For the drives I've processed, the "youngest" wear 001 drives may show 1/4th the TBW as the "oldest" wear 001 drives of that model.

My point is, "They often can be written much much more than rated" is much like saying "this game of russian roulette is ok because there's TONS of empty chambers" -- thanks but no thanks.