Persistence of Flash Memory

[Witling]

Are there any figures on how long flash memory can be expected to remain reliable, i.e., when it starts to degrade?

 

[KalTorak]

Sure - in the flash components' datasheets. Now, everything in those sheets is done through extrapolation; a lot of those estimates are on the order of a hundred years, and since there wasn't any flash memory around in 1904, that's kinda hard to empirically test...

 

[Eskimo]

Originally posted by: KalTorak
Sure - in the flash components' datasheets. Now, everything in those sheets is done through extrapolation; a lot of those estimates are on the order of a hundred years, and since there wasn't any flash memory around in 1904, that's kinda hard to empirically test...

Actually flash is usualy spec'd in terms of program/erase cycles, not so much in terms of time. In my experience the first samples out of the fab are usually recipients of quite intensive non stop testing under extreme conditions (heat/voltage). After a certain number of cycles at those conditions without fail the part is qualified for the fab to begin production starts. Before the first production part is ever shipped to a customer though those initial sample parts have received 45-60 days of 24/7 erase/program cycles. On the order of 300K cycles. Every company is different so obviously I can't speak for them all.

 

[Witling]

Thank you for the answers. It lets me know what I wanted to know.

 

Reading takes little toll on the memory, however flash memory can be rewritten a very limited number of times, some are rated at only 10000 writes, but the question is, how often are you really erasing and rewriting your data?

Flash memory basically works by shooting electrons onto a layer of glass so hard that they stick there, you can only do this so many times before the glass starts to break.