SSD infant mortality II

Results of the SSD infant mortality poll were startling. Consumer SSDs may have bigger problems than thought.
Written by Robin Harris, Contributor

Based on what little statistical information released publicly about SST mortality, I expected the actual number of consumers experiencing SSD failure would be around 5%. But the results were much higher: 17% of over 600 respondents reported SST infant mortality.

Even accounting for the extra emotional disappointment that an early failure causes this number seems substantially higher than any previously published information would suggest. To recap, here's the question and the results from the last time I checked:

Assuming you have owned a flash SSD, did it fail in the first 6 months?

No. - 83% (512 votes)

Yes. - 17% (105 votes)

Total Votes: 617

So I'll dig deeper with a more in-depth survey. It will ask for vendor and respondent names to help ensure that the data is coming from real people and not competitors bashing other competitors.

This doesn't fix the problem of audience self-selection but it should at leastgive us some visibility into overall vendor reliability - after adjusting for vendor market share.

I spoke to an Intel SSD engineer at the Flash Memory Summit in Silicon Valley who said that Intel has less than a 1% return rate. While I have no reason to doubt Intel's assertion - didn't ask if the 1% included No Trouble Found drives - this suggests some other vendors have higher return rates.

I'd like to know who they are.

The Storage Bits take

SSD marketing often claims greater reliability because there are no moving parts. But online SSD buzz offers recurring stories of sudden failures that leave users stranded.

With the survey I hope to get some good information of SSD drive reliability - by vendor if enough responses - and to make recommendations. In the meantime though, users should ignore the "reliability" argument and back up their data often. SSDs offer real advances in performance, but their reliability is still unproven.

Comments welcome, of course. Got any questions for the survey?

Editorial standards