Trying to understand low SSD lifetime

stqn · 2013-02-07 23:49:39

Hi. I’ve been using an OCZ Onyx SSD for two years and today had a look at its SMART values. I’m told the disk has already reached half of its lifetime:

205 Max_PE_Count_Spec       0x0000   ---   ---   ---    Old_age   Offline      -       5000
206 Min_Erase_Count         0x0000   ---   ---   ---    Old_age   Offline      -       792
207 Max_Erase_Count         0x0000   ---   ---   ---    Old_age   Offline      -       3838
208 Average_Erase_Count     0x0000   ---   ---   ---    Old_age   Offline      -       2513
209 Remaining_Lifetime_Perc 0x0000   ---   ---   ---    Old_age   Offline      -       50

This seems wrong because:

User Capacity:    64,023,257,088 bytes [64.0 GB]
Sector Size:      512 bytes logical/physical
(…)
199 Write_Sectors_Tot_Ct    0x0000   ---   ---   ---    Old_age   Offline      -       1881433528

According to these values 1881433528*512 = 963293966336 bytes have been written to the disk (963 GB).

According to OCZ (pdf) the formula to measure the amount of data that can be written to an SSD is: DriveSize * PECycles / WriteAmplification, and the worst possible write amplification is 17.

For my disk that makes: 64 GB * 5000 / 17 = 18823 GB.
Which would give a remaining lifetime of: 100 - 100 * 963 / 18823 = 94%!
And we’re talking of a worst case scenario here.

So what is wrong?

The 963 GB of data written to the disk corresponds to about 1.3 GB per day, which seems plausible since I’ve taken care not to write too much on the SSD. I still have pacman’s cache on the SSD, but not Firefox’ cache. I remember using iotop to measure my disk usage and finding that it was low enough (though maybe I didn’t measure disk writes during an update… I don’t remember that.)

I’m using LVM and its snapshot feature (every time I backup&update arch, and the snapshot is removed after the backup is done) but without encryption. Filesystem is ext4 with options discard, noatime, commit=60. There is no swap partition.

My SSD has firmware 1.6 and I see that 1.7 has a fix for a bug that could possibly cause extra wear, but they don’t say how much:

- Enhancement: Change of the behavior of meta data save. With 1.6 FW, there could be such cases that will save the whole meta data even when it is not necessary, for example, during idle time. It depends on usage. Now, the policy of saving meta data is fine-tuned
- Bug fix: Sanitary erase may corrupt write log history. It was fixed.

Can someone see what is going on?

Last edited by stqn (2013-02-08 14:51:43)

debijan · 2013-02-08 15:10:32

Hi stqn,

could you please provide the output of:
$ lsblk -D

I am referring to answer of "mooninite" on https://ask.fedoraproject.org/question/ … lvm-or-not

stqn · 2013-02-08 16:07:06

Hi debijan.

Here it is:

NAME                         DISC-ALN DISC-GRAN DISC-MAX DISC-ZERO
sda                                 0      512B       2G         0
├─sda2                              0      512B       2G         0
│ ├─VGSSD1-ArchRoot (dm-0)          0      512B       2G         0
│ └─VGSSD1-ArchRoot64 (dm-1)        0      512B       2G         0
└─sda3                              0      512B       2G         0

stqn · 2013-02-09 00:02:20

I just found this thread http://forums.anandtech.com/showthread.php?t=2144494 which at least shows that I’m not alone… The Indilinx Barefoot controler must be very very bad…

Edit: according to OCZ the reported remaining lifetime is pretty much useless! Magnificient.

Last edited by stqn (2013-02-09 00:09:34)

blackout23 · 2013-02-09 11:22:05

OCZ has a very bad reputation for SSDs in general. Everything that was Sandforce based like the Vertex 2 and 3 had a abnormally high failure rate. They even manipulated some reviews on price-comparison sites. http://www.theregister.co.uk/2012/02/28 … print.html

I'd only go for Crucial and Samsung SSDs. They are faster more reliable and cheaper, too.

graysky · 2013-02-09 11:33:33

blackout23 wrote:

OCZ has a very bad reputation for SSDs in general
...

Maybe true with older generation models, but the Vertex 4 is a great product.

blackout23 · 2013-02-09 23:02:18

graysky wrote:

blackout23 wrote:
OCZ has a very bad reputation for SSDs in general
...
Maybe true with older generation models, but the Vertex 4 is a great product.

Yes it uses a new controller, too. They learned a bit there. Haven't heard of any common problems with it.
Still as a company they lost a lot of reputation will take them a long time to regain it.

litemotiv · 2013-02-10 10:04:53

blackout23 wrote:

graysky wrote:
blackout23 wrote:
OCZ has a very bad reputation for SSDs in general
...
Maybe true with older generation models, but the Vertex 4 is a great product.
Yes it uses a new controller, too. They learned a bit there. Haven't heard of any common problems with it.
Still as a company they lost a lot of reputation will take them a long time to regain it.

Ah yes, this brings back memories...

Lost all data on home partition, would really appreciate advice

It's a bit irrational, but i am still hesitant to buy an OCZ ssd again, some 5 years later...

WonderWoofy · 2013-02-10 10:30:03

Thanks for that entertaining thread litemotiv Though i am sorry you had to experience that.

I have never owned an OCZ, but currently have two Samsungs and a Mushkin in my laptop, and they have been great to me so far. I have heard that even the older OCZ drives are okay as long as you update to the current firmware. Though after an experience like that, I don't blame you for avoiding them.

Arch Linux

#1 2013-02-07 23:49:39

Trying to understand low SSD lifetime

#2 2013-02-08 15:10:32

Re: Trying to understand low SSD lifetime

#3 2013-02-08 16:07:06

Re: Trying to understand low SSD lifetime

#4 2013-02-09 00:02:20

Re: Trying to understand low SSD lifetime

#5 2013-02-09 11:22:05

Re: Trying to understand low SSD lifetime

#6 2013-02-09 11:33:33

Re: Trying to understand low SSD lifetime

#7 2013-02-09 23:02:18

Re: Trying to understand low SSD lifetime

#8 2013-02-10 10:04:53

Re: Trying to understand low SSD lifetime

#9 2013-02-10 10:30:03

Re: Trying to understand low SSD lifetime

Board footer