High failure rate of OCZ SSDs yields answers that raise troubling questions

[gongdonger]

Heres my tuppence :

every ocz drive that i have been using for heavy database testing and heavy photoshop work has failed quickly. heavy mysql / xp2sp usage killed a ocz 30gb solid in < 2days. frequent photoshop image manipulation killed a ocz vertex 2 60gb < 1 month xp2sp. for drives that dont really do anything, that is ocz 30gb solid in networked epos terminals, over 12 months, out of 40 drives installed wepos cloned using norton ghost, 3 were DOA, 4 more failed < 12 months.

i would never trust important information to a SSD because when they fail, in general its instant with no warning, and complete and total loss of everything - use it just for installing OSs and apps - but put your data elsewhere.

as for aligning partitions, yes i agree that should increase lifespan of flash - but the failures we are seeing are not reduced diskspace / bad blocks as you would expect for accelerated wear out, we are almost always seeing 'not detected by bios' failures which indicates the controller is being fried - nothing to do with wearing out. so i dont buy the 'you need to align your partitions for xp' being bandied about as a reason for failure. nb. when i first started in the epos industry, there was a notorious brand of motherboard that would have its video interface fried just by plugging or unplugging the vga cable while it was powered up. i wonder if the ocz controllers are just as easily susceptible to that level of static damage ?

OCZ replied today and it's interesting what they wrote.

Firmware updates should only be applied if the drive is malfunctioning similar to your case. We do not recommend users to update the firmware if the drive is functioning properly. You can apply the firmware update without the secure erase and see if the issues have been resolved. If they haven't we then recommend that you perform the secure erase and a fresh install.

As for image cloning, SSD Controllers have to translate logical blocks to physical nand locations. This complicates mapping snapshots when imaging 'incrementally'. If you decide to clone your drives, please use whole disk images as they include all tables in the disk signature and are preferable with your OCZ SSD. Please also perform the clones from external bootable media rather than internal to the OS.

Two questions immediately arose, which I submitted.

Thank you for your informative reply. I wonder if and how cloning could cause the kind of failure 3 of the 4 SSDs had. I mean, the drives worked fine for two or more months before suddenly becoming unrecognizable to any BIOS as drives. Can imaging actually lead to this kind of failure months later?

Secondly, you write that "Firmware updates should only be applied if the drive is malfunctioning similar to your case." In the case of this latest drive, which still shows up in BIOS, I can see how your firmware update software might still communicate with the drive (and I'm having a tech try that as I write this), but can your firmware software update an SSD that the BIOS cannot see?

I plan on also asking what exactly they mean by "external bootable media. Can it not be Windows 7 on another HDD? Their reply also raises a question that I think many individual tech-savvy SSD users might have: can programs like Tereabyte's Image for Windows (or DOS), Acronis Trueimage, or one of Paragon's programs be use to perform the kind of disk copying or imaging that OCZ recommends? For that matter, there's "imaging" and there's simply copying an entire drive. "Imaging," at least to me, means to create a .IMG file of a partition or drive using a program like Terabyte's Image for DOS. It's akin to an intermediate file that must be restored to a drive to end up with what you originally imaged. "Copying a drive," as one can do with Paragon Partition Manager, is quite different. I need to ask OCZ if simply copying an SSD to another drive, like an HDD or another SSD, is possible. And I will.

Any constructive comments will be appreciated.

[dameslap]

Fascinating.

I'm working on a very sizeable project that integrates STEC Mach4 Compact Flash disks into bespoke low-power, XP-based systems.

The in-house process has been to restore at a Partition level using Ghost and we are seeing many (Many) failures as above. Most often we see general performance degradation after a period of months, closely followed by boot failure which can be rectified by re-imaging, but generally by this stage the card has become unreliable and quickly fails again in the same way.

We have also (less often, but in maybe a dozen examples) seen sequential write operations (Ghost imaging and also the use of Paragon Alignment Tool) 'trash' a disk and render it invisible to the OS. Earlier this afternoon I used PAT to re-align our single partition in the hope that it might alleviate some of these issues. It appeared to execute correctly, but on rebooting though the disk can still be identified by the BIOS, nothing else (Ghost, for example) appears able to recognise it as a disk any more.

My feeling is that there's something about the sequential-write behaviour of disk imaging that can bypass card wear-levelling to cause certain blocks over and over, taking them way beyond block-specific endurance and consequentially depleting the card's reserved blocks.

But the harder I look, the more I find that beyond the marketing gloss, the manufacturers are pretty cagey about precisely how they operate wear-levelling.

[curiousclive]

I'm scared stiff now. For nearly 2 years I have been using Crucial SSD the latest ones are M4's the 2 year old ones are C300 all had the latest firmware installed before use. So far I have had no problems at all with any drive. But it is a bit to early for the M4's as they are only about 2 months old but are still working at the full speed as when new. All drives are derivatives of the first image I created from the normal HDD so none have had a clean install. All machines are using win 7 32bit and 64 bit home.
So far I am very happy with the upgrade to SSD but now after reading this I am having heart attacks. Hope this is not another unfounded scare story.

[dameslap]

Clive, if you're solely using Win7 you can be a lot more optimistic. Unlike XP (that we are stuck with for commercial reasons) W7 aligns its partition properly with the SSD, can issue TRIM instructions to the SSD which should assist the card to keep enough spare capacity to wear-level properly and takes various other file-based measures (prefetching most notably) that should also help.

[curiousclive]

Clive, if you're solely using Win7 you can be a lot more optimistic. Unlike XP (that we are stuck with for commercial reasons) W7 aligns its partition properly with the SSD, can issue TRIM instructions to the SSD which should assist the card to keep enough spare capacity to wear-level properly and takes various other file-based measures (prefetching most notably) that should also help.

Thanks Dameslap
I feel much better now.
I have prefetching disabled to reduce write operations to drive as well as indexing and super-fetch.

I've also used SSD tweaker Pro to supposedly optimize the trim but have no idea what it does if anything.

[gongdonger]

here's an update of my ssd experience - three attempts the ssd industry has had to convince me their stuff works, and three times they have failed. although they were all running xp which by default doesnt align partitions, for first drive i went through the process of using diskpart to align the partition on 1mb boundary, and when it failed, it wasnt such a catastrophic failure but was stuttering badly so i was able to rescue all the data on it, but it still failed after only 47hrs.

this is the lifespan of all the ssd's ive personally installed :
ocz solid - 47hrs
ocz vertex - 3 months
ocz agility - 11 months

compare that to mechanical drives ive installed, about 40 over the last 15 years, ( maxtor and seagate ) and only one developed a corrupt sector that warranted backing all my precious data up then reformat to fix. when ssd's fail, they fail bigtime with no warning and you are left with a brick. files - gone, emails - gone, windows - gone - all in a flash.

[Tinto Tech]

@gongdonger: I wonder if the firmware for the drive is stored in the main flash array?

If it is, it might explain why some drives experience these types of issues: degradation of the flash array due to no TRIM support in XP, misaligned partitions, increasing wear-out, then all of a sudden the firmware is corrupt and the result is "not shown in bios" errors.

[barrydiston]

I Purchased an OCZ Petrol 2 weeks ago. The last few days it's been dissapearing from the BIOS more and more frequently, to the point now, where I can't get it recognised at all.

The whole computer was a completely new system, all just 2 weeks old, latest drivers, windows 7 64 bit, & it's been kept off the net.

this is my second experience with a SSD, both have ended badly. After some research it seems loads of people are having similar problems with these SSD's

Back to drives with moving parts for me now........

[Infinicore]

Just purchased my 7th one, 8th maybe? Mixture of XP and Win7. No problems to date (almost a year for the oldest) except I have had ever so occasional blue screens since installing one (three times over maybe 3 months). I thought it might be happening when the Adobe flash updater popped up with an update, but the last time I could not tie it to that so I'm not sure. The Win7 box is still up 99.9% of the time so I'm not too worried about it.
Mixture of Crucial, OCZ and Samsung drives.