SanDisk SDSSDH2256G drops out

[maximus]

I have a SanDisk SDSSDH2256G as a primary drive in my computer for close to 3 years now, and it has dropped out at least 3 times. By dropped out I mean it either locked up Windows or blue screened, and afterwards it was required to change the drive order in bios so that the computer would boot again. So when it dropped out the bios must have lost it. I maintain a full backup so I am not in danger of losing everything, but this is very annoying. Are there any known issues with this drive? It seems more likely to happen under heavy activity, but overheating should not be an issue as it is in a very well ventilated computer.

[maximus]

I would like to update this post. I am still using this drive, although I think it is time to replace it. The drive works perfectly right up until it has to deal with very large reads and writes (such as cleaning up a virtual machine, which I don't do very often due to this issue). If I can catch it when it locks up, I can power cycle the machine and save myself a bigger headache (if the computer blue screens and reboots the drive is not detected, the bios picks a different drive to try to boot from, and my raid 1 feels compelled to perform a full check which takes 8 hours). So it would seem that the drive goes into device fault. The highest temp I have seen on the drive is 42C during such high activity, which is well within specs. I even moved it in the drive bay to be in front of a fan just to see if that would help, but no luck. But the kicker is that once the drive messes up and I power cycle, I can immediately hammer the drive with reads and writes (perform the same task that killed it and then some) and it will not fault out again. I dare someone to explain why the drive acts as it does, seemingly defying normal logic.

[labtech]

SMART attributes?

[fzabkar]

Have you checked for a firmware update?

Have you tried to limit the SATA speed? For example, if the drive is attached to a 6Gbps port, try moving it to a 3Gbps port.

If the drive supports NCQ, try disabling this feature.

[maximus]

SMART attributes?

Hmm, check the smart data, why didn't I think of that? So I checked and I don't see anything that indicates an issue. The unexpected power loss count had me wondering for a minute, but then I remembered that I have another drive in this system that doesn't always like to come up at startup so I have to do a power cycle before the OS starts loading which would be the likely cause of that count.

----------------------------------------------------------------------------
(4) SanDisk SDSSDH2256G
----------------------------------------------------------------------------
Model : SanDisk SDSSDH2256G
Firmware : X211200
Serial Number : 130504401051
Disk Size : 256.0 GB (8.4/137.4/256.0/256.0)
Buffer Size : Unknown
Queue Depth : 32
# of Sectors : 500118192
Rotation Rate : ---- (SSD)
Interface : Serial ATA
Major Version : ATA8-ACS
Minor Version : ATA8-ACS version 6
Transfer Mode : SATA/600 | SATA/600
Power On Hours : 8988 hours
Power On Count : 1484 count
Host Reads : 1067480426 GB
Host Writes : 355836969 GB
NAND Writes : 83 GB
Temperature : 28 C (82 F)
Health Status : Good (100 %)
Features : S.M.A.R.T., APM, 48bit LBA, NCQ, TRIM
APM Level : 00FEh [ON]
AAM Level : ----

-- S.M.A.R.T. --------------------------------------------------------------
ID Cur Wor Thr RawValues(6) Attribute Name
01 100 100 __4 000000000000 Vendor Specific
05 100 100 __0 000000000000 Reallocated Block Count
09 253 100 __0 00000000231C Power On Hours
0C 100 100 __0 0000000005CC Power Cycle Count
A7 100 100 __0 00000000002C Max Bad Block/Die
A8 100 100 __0 00000000003B Maximum Erase Cycle
A9 100 100 __0 000000000192 Total Bad Block
AB 100 100 __0 000000000000 Program Fail Count
AC 100 100 __0 000000000000 Erase Fail Count
AD 100 100 __0 000000000019 Average Erase Cycle
AE 100 100 __0 0000000001A5 Unexpected Power Loss Count
B8 100 100 __0 000000000000 Vendor Specific
C2 _72 _44 __0 002C0002001C Temperature
D4 100 100 __0 000000000000 SATA PHY Error
E6 100 100 __0 000000000053 Percentage Total P/E Count
E8 100 100 __0 000000000064 Spare Block Remaining
E9 100 100 __0 000000000053 Total GB Written To NAND
F1 253 253 __0 00021535A429 Total GB Written
F2 253 253 __0 00083FA0756A Total GB Read
F3 100 100 __0 000000000000 Vendor Specific

[maximus]

Have you checked for a firmware update?

Didn't think of that. On the SanDisk site it shows that I am a couple firmware versions behind. But when I installed the SanDisk Dashboard and check the firmware, it says it is up to date. WTF?!

Have you tried to limit the SATA speed? For example, if the drive is attached to a 6Gbps port, try moving it to a 3Gbps port.

Not sure if I wan to play around with moving it to a different port. Because I know I will have to go into the bios and adjust which drive it should boot from, and I would bet that my raid 1 will decide it needs to check itself.

If the drive supports NCQ, try disabling this feature.

How can I tell if the drive supports it, and how can it be disabled?

[maximus]

After checking and double checking that I have the correct model (only difference is SanDisk website has a dash in the number and the toolkit doesn't), I have found a manual firmware upgrade option. So if there are more posts on this and I don't respond for a few days, that will likely mean that I nuked the drive and crippled my system

[maximus]

(sigh) No compatible device found for firmware update. I think it is time to shop for a new SSD, and while I have had good luck with their flash drives, it won't be a SanDisk.

[fzabkar]

Queue Depth : 32
Features : S.M.A.R.T., APM, 48bit LBA, NCQ, TRIM

You can disable NCQ by changing the queue depth to 1.

One way to do this is with hdparm:

hdparm -Q1 /dev/ice

http://man.he.net/man8/hdparm

Google also turns up the following alternative method:

echo 1 > /sys/block/sda/device/queue_depth

Smartmontools (smartctl) may provide a better rundown of the SMART attribute names. CrystalDiskInfo doesn't appear to understand them properly. The data are correct, though.

For example, the following attributes are incorrectly interpreted:

E9 100 100 __0 000000000053 Total GB Written To NAND
F1 253 253 __0 00021535A429 Total GB Written
F2 253 253 __0 00083FA0756A Total GB Read

F1/F2 should be Total LBAs Written/Read by Host.

0x00021535A429 LBAs x 512 bytes per LBA = 4.58 TB

The Average Erase Cycle is 0x19, so this would suggest that the total data written to the NAND array over the lifetime of the drive is ...

0x19 x 256 GB = 6.4 TB

This in turn suggests a write amplification factor of ...

6.4 TB / 4.58 TB = 1.40

The temperature attribute is reporting that the max/min/current temperatures for the current power cycle are 44C/2C/28C, respectively.

C2 _72 _44 __0 002C0002001C Temperature

The highest temperature recorded by the drive over its lifetime is 56C (= 100 - 44).

The drive has recorded no SATA PHY Errors, so this would suggest that there is SATA communications problem.

[fzabkar]

The Average Erase Cycle is 0x19, so this would suggest that the total data written to the NAND array over the lifetime of the drive is ...

0x19 x 256 GB = 6.4 TB

This in turn suggests a write amplification factor of ...

6.4 TB / 4.58 TB = 1.40

Sorry, I should have included the overprovisioned capacity in the calculation (ie the entire NAND capacity), not just the user space.

0x19 x 256 GiB = 6.87 TB

6.87 TB / 4.58 TB = 1.5

[maximus]

I broke down and ordered a 500GB Samsung SSD. I needed more space anyway. When I swap it out, I see if I can do some tests on this one to see if I can reproduce the results of getting it to crash. I also ordered a new 4TB WD to replace the 2TB Hitachi that is giving me the other issue of not coming up all the time when first booted (also need more space here too).

F1/F2 should be Total LBAs Written/Read by Host.

Believe it or not, the actual SanDisk Dashboard lists the names exactly like CrystalDiskInfo. But the values look like they are limited to 32 bit so they are off and basically useless to even count LBA.

FYI this was the output of HDDSuperTool smart data script. I used CrystalDiskInfo because it named the three value 167, 168, and 169.

Code:

Smart structure version= 4
ID#   FLAG  VALUE WORST THRESH   RAW DATA          ATTRIBUTE NAME
  1  0x000e  100   100     0   0x00000000000000   Read Error Rate
  5  0x0032  100   100     0   0x00000000000000   Reallocated Sectors Count
  9  0x0032  253   100     0   0x0000000000231b   Power-On Hours Count
12  0x0032  100   100     0   0x000000000005cc   Power Cycle Count
167  0x0032  100   100     0   0x0000000000002c   Unknown
168  0x0032  100   100     0   0x0000000000003b   Unknown
169  0x0032  100   100     0   0x00000000000192   Unknown
171  0x0032  100   100     0   0x00000000000000   Program Fail Count
172  0x0032  100   100     0   0x00000000000000   Erase Fail Count
173  0x0032  100   100     0   0x00000000000019   Wear Leveller Worst Case Erase Count
174  0x0032  100   100     0   0x000000000001a5   Unexpected Power Loss
184  0x0032  100   100     0   0x00000000000000   End-to-End error
194  0x0022   70    44     0   0x00002c0002001e   Temperature
212  0x0032  100   100     0   0x00000000000000   Shock During Write
230  0x0032  100   100     0   0x00000000000053   GMR Head Amplitude
232  0x0033  100   100     4   0x00000000000064   Available Reserved Space
233  0x0032  100   100     0   0x00000000000053   Media Wearout Indicator
241  0x0030  253   253     0   0x000002152e09bf   Total LBAs Written
242  0x0030  253   253     0   0x0000083f621853   Total LBAs Read
243  0x0032  100   100     0   0x00000000000000   Unknown

[labtech]

What OS are you using for normal use on the machine? Or is it a dual boot with a Linux based OS?

[maximus]

What OS are you using for normal use on the machine? Or is it a dual boot with a Linux based OS?

It is a dual boot with Win7 and Linux, but spends almost all of its time in Win7 as that is the host that is running VMWare.

[fzabkar]

"GMR Head Amplitude" is a HDD attribute, not SSD. It appears to be related to attribute 233.

Note that "Wear Leveller Worst Case Erase Count" and "Average Erase Cycle" are not the same thing, so attribute 173/0xAD is being interpreted incorrectly by one of the tools.

In short, you can't trust any SMART tool to get it right, not even the manufacturer's.

[maximus]

In short, you can't trust any SMART tool to get it right, not even the manufacturer's.

Yeah, that is for sure. And there is no true accurate full list of smart attributes that can be found anywhere. I did the best I could with the script by comparing as many of the lists that I could find online. And seeing SanDisk's own tool giving inaccurate results is funny in a sad sort of way.

[labtech]

Not sure what the problem may be. Possible a driver issue in Windows. Seen some people complaining about that in some other forums pertaining to Sandisk.

I think upgrading to new disks qnd keeping the old ones for backups sounds like a good plan.

[maximus]

The old ones will be put to use in another machine for testing. But I would like to know if there is any software that can do hard core read and write tests to a drive, not just benchmarking but something that will run as long as you tell it to and just hammer the drive. More of a burn in test. I want to torture this SanDisk drive so I can understand to what level I can trust it in another machine. If there isn't anything like that, I guess I will have to write something of my own for it.

[fzabkar]

Have you checked for known problems with your SATA controller?

[labtech]

The old ones will be put to use in another machine for testing. But I would like to know if there is any software that can do hard core read and write tests to a drive, not just benchmarking but something that will run as long as you tell it to and just hammer the drive. More of a burn in test. I want to torture this SanDisk drive so I can understand to what level I can trust it in another machine. If there isn't anything like that, I guess I will have to write something of my own for it.

Not sure hammering a SSD is a really good idea. The more writing to it, the less life NAND will have.

[jeremyb]

Have you checked for known problems with your SATA controller?

I had a few on-board Intel RST RAID 1's that would BSOD all day if I put any serious IOPS on them. The hard drives would check out fine every time in SMART or using diagnostic software on other computers. The chipset was P67, the only solution was to replace the motherboards.

I wondered if there was an updated Intel RST BIOS module that would solve the problem however I "assume" the ROM only handles INT13 calls since any heavy lifting seems to occur in windows where the software drivers take over.

Intel RAID's really under perform compared to hardware. I'd look into a hardware RAID card from Adaptec or LSI with onboard memory and a dedicated processor. Those are pretty rock solid and start around $300. They're geared towards entry level servers but they can handle IOPS much better than Intel and have optimized read routines to get a little RAID0 performance boost out of RAID1.