SSD Drive health at 3% :o

[wonko]

Hi all, I've got a Crucial CT64M225 SSD that's been a Windows 7 boot drive since (I think) mid 2009.

I have just updated it to firmware 2030. Part of the update apparently reduces the predicted number of erase cycles that cells can do (from 10000 to 5000 cycles), with the result that the SMART D1 parameter drive health/remaining drive life has dropped to 3% . This, apparently, is just an estimate, and the drive won't suddenly stop working when it hits 0%.

So, has anyone got any experience of this with the M225 drives? Any ideas on how long can I expect it to work for would be appreciated, as would any thoughts on warranty if the expected life of the drive is less than that expected at time of purchase.

Thanks

 

[Undesirable]

How many writes do you think you've made to the drive over the time you've used it? Is the swap file on the SSD? Have you opened a lot of archive files (RAR, ZIP, etc.) with the temp directory set to the SSD? Do you use the hibernate feature a lot?

 

[g0th2000]

It's also possible that the firmware update has change the way remaining lifetime is reported, i.e. they changed it from % to 100-% and it's really 97%, that's the problem with SMART - the standard is there is no standard

Given that it's an M225 I'd be more worried about it dropping dead suddenly than you wearing it out

 

[Dirk]

I assume some of you haven't been following the endurance testing here:

http://www.xtremesystems.org/forums/showthread.php?271063-SSD-Write-Endurance-25nm-Vs-34nm

The Samsung 470 was good for ~550TiB before dying a death, while the Vertex Turbo is still going strong on 601TiB, and the Crucial M4 is cruising having had 768TiB of writes so far.

That's about 900 years of use for my daily writes, before nand death occurs. Should be enough to last most people until the next upgrade cycle at least.

 

[wonko]

How many writes do you think you've made to the drive over the time you've used it? Is the swap file on the SSD? Have you opened a lot of archive files (RAR, ZIP, etc.) with the temp directory set to the SSD? Do you use the hibernate feature a lot?

I've no idea about the number of writes TBH. The temp folder is on the SSD, and I do open a lot of rars so it has had quite a bit of use.

I suppose what I'm trying to understand is how realistic is the 'remaining life' calculation - if it's pessimistic and the drive will likely carry on working for another couple of years at current usage then I'll be happy. If it's on the edge of dying right now then I'd like to know!

It's also possible that the firmware update has change the way remaining lifetime is reported, i.e. they changed it from % to 100-% and it's really 97%, that's the problem with SMART - the standard is there is no standard

Given that it's an M225 I'd be more worried about it dropping dead suddenly than you wearing it out

I understand your point about there being no standards on the SMART data, but the D0 parameter (agerage erase count) is very close to the CD (Maximum PE Count Specification) so don't think I've misinterpreted it.

I hadn't heard that the M225 had a reliability problem. Mine has been rock solid so far, and is well backed up on my WHS, so if anything I'd rather it died under warranty than wore out out of warranty

 

[wonko]

I assume some of you haven't been following the endurance testing here:

http://www.xtremesystems.org/forums/showthread.php?271063-SSD-Write-Endurance-25nm-Vs-34nm

The Samsung 470 was good for ~550TiB before dying a death, while the Vertex Turbo is still going strong on 601TiB, and the Crucial M4 is cruising having had 768TiB of writes so far.

That's about 900 years of use for my daily writes, before nand death occurs. Should be enough to last most people until the next upgrade cycle at least.

Thanks, that looks like an interesting link, and at 99 pages could take a bit of time to read .

SSDLife reckons that a total of 11283GB of data has been written, which makes it a youngster compared to the ones you quote. Even if the M225 is an older generation which wears out quicker I'd hope it's not that much different.

Looking a bit more at the SMART data, the average erase cycle count is nearly at the (revised) specified CD parameter value, and yet the C4 number of erase failures is zero. If the CD parameter was meant to be an average expected lifetime, then it would seem to be very conservative. Makes you wonder why they revised it in the firmware

Ah well. Maybe my excuse for a shiny new SSD isn't as imminent as I thought it might be.

 

[Dirk]

Bottom line, SMART attributes, especially MWI data, is next to useless as a real world guide on the life of your SSD.

Here's an old graph from that thread:

MWI for the M4 was at 170TiB, so it's worn out it's 'expected' life 4.5 times already.

 

[wonko]

OK, thanks, so looking at the purple "M225->Vertex Turbo 64MB' bar, I'm reading this as meaning that the 'predicted' lifetime based upon the the drive's own Media Wearout Indicator is circa 190 TiB (for the usage pattern in the test setup), and that so far it has lasted nearly twice this long without dying yet.

There's quite a difference between the 11T(i)B or so that mine has written vs. this 190 TiB, but I'm presuming that this will be down to the difference between my real-world usage to that of the test setup (?)

So I'm still puzzled why Crucial changed the predicted lifetime to something so pessimistic, and it also begs the question of how to tell when your SSD is nearing end of life. Presumably once you start seeing C3 and C4 going up (Program and Erase Failure Block Counts) it's a bad sign.

 

[Dirk]

That chart is quite old now. The Vertex Turbo is over 600TiB now so about three times it's predicted life span. All the manufacturers have erred on the side of caution though as you can see above. Their estimates of nand lifecycle have been very conservative.

The test itself is designed to mimic real world writes, the only difference being the intensity of the test which is far higher than real world usage. In fact, given that nand is believed to have even greater endurance given longer rest periods, the test is itself gives very conservative end results.

For example the now dead Samsung 470 which took in excess of 500TiB of constant writes, may have lasted ten times that if the data had been written under more usual loads, with the nand being given more time to recover between writes. Of course that's just speculation at this point, but there was a very thorough white-paper written about this. It's linked in that thread somewhere.

Edit: Found it:

'How I Learned to Stop Worrying and Love Flash Endurance':

"...recovery periods of such durations can significantly boost endurance, allowing the blocks to undergo several millions of P/E cycles before reaching the endurance limit.
The amount of time required for reaching the endurance limit is much longer than the NAND flash retention period. Therefore, endurance is not a major flash reliability concern under realistic data center usage scenarios and a much wider array of I/O intensive applications can leverage the performance and power benefits of flash-based SSDs than previously assumed."

As far as how you would tell if your SSD is nearing the end, you should start seeing the number of re-allocated sectors rising sharply, with a corresponding slow-down in Write speed.

 

[wonko]

This is obviously something you've taken quite an interest in - cheers again for the answers and pointers to more than enough bedtime reading.

I shall stop taking any notice of the remaining life attribute, just like I stopped being bothered by my WD Green Load/Unload cycle count .