SSD failure rates

StoreLab of Russia has different numbers and says just the opposite about Hitachi, that they're the most reliable:


http://www.tomshardware.com/reviews/hdd-reliability-storelab,2681-2.html

worldwide market shares:


StoreLab's distribution of failed drives:


However market shares in Russia can vary greatly from world market shares, and StoreLab said Seagate has a 40% share in Russia vs. 31% worldwide.

Is anybody a Google insider who's seen their failure rates by brand?

King107s said:

That does sound a little high.... all these numbers sound high to me.

Click to expand...

I believe those drive were the worst offenders. Add to that the fact that failure rate seems to scale up with size, some smaller drivers would have lower rates. And the that n for individual drives can be as low as 100so the error bars would be big. Given this the overall failure rates are lower.

EDIT: With n=100 the confidence interval for the 2TB Caviar Black would be (3.91%, 15.5%). The range is quite huge.

larrymoencurly said:

StoreLab of Russia has different numbers and says just the opposite about Hitachi, that they're the most reliable:

Click to expand...

What's their source of information? Looking through google it seems they do data recovery while this French company was in retail. What's the difference?

The retail store can tell you the percentage of sold drives that faild, telling you the reliability (in theory). The data recovery company can tell you what drives where brought for recovery... Interesting, but what does it mean?

It's not surprising the numbers are different when you are not measuring the same thing.

Seems with retail, you'd get bias towards drives that fail OOTB since most have limited time return policies, while the recovery guys can pretty much only tell you what percentages of failed drives they received are what brand... which would potentially have selection bias towards the most popular (most sold) drive.

What would be nice is if the manufacturers had to disclose the numbers for warrantied replacements of their products... hmmm...

The repair guys would have strong bias to whatever drives corporations / wealthier citizens are using... Seriously, most countries have the stores handling RMA's, for example in France the store will have warranty on any drive for two years (whatever the manufacturer says), and people will be used to contacting the store regarding any problems. It should give a more objective picture.

However looking at the results you can see WD having the most reliable and least reliable 2TB drive, and it should remind you that these things can be very dependent on the specific product, not just manufacturer.

aphellyon said:

Yea, I think one of my raptors is going now after a year and a half... getting occasional bsods and when i reboot i get intel matrix raid manager warnings on one of the drive. So far, I've been able to mark the drive as normal and keep my array going... but i doubt that will last.

Click to expand...

I get error messages on my raid manager too but that typically is dealing with a bad write when it BSOD's which was caused by something I did. All data is intact and works perfectly fine, just a bad write that couldn't finish getting all its data to the drive is all. Thats on my SSD's in Raid0 too.

dropadrop said:

What's their source of information? Looking through google it seems they do data recovery while this French company was in retail. What's the difference?

The retail store can tell you the percentage of sold drives that faild, telling you the reliability (in theory). The data recovery company can tell you what drives where brought for recovery... Interesting, but what does it mean?

It's not surprising the numbers are different when you are not measuring the same thing.

Click to expand...

The StoreLab article indicates they're going by number of drives that came in for repair, both for manufacturing defects and damage caused by dropping, and they said all 200 of the Hitachis they received failed from dropping.

I didn't mean that one of those companies was right and the other was wrong but that there doesn't seem to be nearly enough information available to the public to let us reach any valid conclusions.

aphellyon said:

Seems with retail, you'd get bias towards drives that fail OOTB since most have limited time return policies, while the recovery guys can pretty much only tell you what percentages of failed drives they received are what brand... which would potentially have selection bias towards the most popular (most sold) drive.

Click to expand...

So why did data recovery company StoreLab favor Hitachi, one of the least popular brands, rather than Seagate or WD?

dropadrop said:

T... you can see WD having the most reliable and least reliable 2TB drive, ...

Click to expand...

That's why when I build a RAID I match drives from different manufacturers. Only time will tell if a particular drive turns out to be a turkey.

Speaking of which, I recently retired an IBM Deathstar - yeah, the particular model that made them infamous. It had over 6 years of operating time according to the SMART data. I retired it because it started throwing errors.

My backup strategy includes RAID 1 storage on site along with another off site. The off site gets woken up via WOL when the backup starts and then shuts down when the backup is complete. It takes less than an hour except when I rip some DVDs. Those take a while to backup over the Internet.

larrymoencurly said:

So why did data recovery company StoreLab favor Hitachi, one of the least popular brands, rather than Seagate or WD?

Click to expand...

sorry, I should have been more specific... selection bias towards getting more failed drives from the most popular brands... that would favor them choosing Hitachi (one of the least popular). I didn't mean it to read "them receiving more failed drives from the more popular brands would lead to a selection bias that would have them favoring those drives.". That wouldn't make any sense.

At any rate, they chose Hitachi because their numbers favored them and because none of the ones they received died due to manufacturing defects or faulty firmware. Which is nice, but with a sample size of 200, and comparing drives of different sizes and such... meh. Imo, the most interesting part of their little analysis is comparing WD to Seagate. Same market share but disproportionate failure rates. But hey, without the actual data all we can really infer is there may or may not be some trend that may or may not be statistically significant from a dataset that may or may not be tightly constrained. So, while I find it interesting, I take it with a grain of salt.

deathman20 said:

I get error messages on my raid manager too but that typically is dealing with a bad write when it BSOD's which was caused by something I did. All data is intact and works perfectly fine, just a bad write that couldn't finish getting all its data to the drive is all. Thats on my SSD's in Raid0 too.

Click to expand...

What were you doing to cause the bad writes?

aphellyon said:

What were you doing to cause the bad writes?

Click to expand...

BSOD from too high OC. Sometimes I have gotten it during a reboot or a shutdown before (normal not forced).

Just little hick ups here or there but it never warranted a system restore from backup or anything, just resetting the controller (in windows) to say everything is OK.

deathman20 said:

BSOD from too high OC. Sometimes I have gotten it during a reboot or a shutdown before (normal not forced).

Just little hick ups here or there but it never warranted a system restore from backup or anything, just resetting the controller (in windows) to say everything is OK.

Click to expand...

Ahhh, I got you. This happened to me once about 6 months ago, and just started back a few weeks ago... the bsods are getting more and more frequent now. I'm running stock clocks, so I figured one of my raptors is going bad. I, um, don't have enough empty spare drives to back-up all my data *facepalm*, so I haven't torn down my array to test them. At any rate, I already have 2 ssds and 2 1TB backup drives on the way and hope to avoid any future raid failure issues.

Just an update on the numbers. I saw that in the latest Intel SSD review Anandtech had a slide from Intel that confirms at least the Intel failure numbers.

According to Intel the Annual Return Rate of the X25-M is <0.7%, and the Annual Failure Rate is ~0.4%. Pretty darn good.

Mr Alpha said:

Just an update on the numbers. I saw that in the latest A<strong></strong>MD SSD review Anandtech had a slide from A<strong></strong>MD that confirms at least the A<strong></strong>MD failure numbers.

According to A<strong></strong>MD the Annual Return Rate of the X25-M is <0.7%, and the Annual Failure Rate is ~0.4%. Pretty darn good.

Click to expand...

Yeah saw that AMD posted those numbers the other day. Pretty good considering and they are suppose to make the 320's even better than that. Thats 4-5x better then the competition.