[NTLK] flash longevity question

[Victor Rehorst]

Modern flash memory chips use multi-level NAND gates packed hundreds (if 
not thousands?) of times denser than the single-level cell, NOR-based 
chips used for any Newton, or compatible linear flash cards. They're 
different beasts designed to perform different jobs.  So I don't think 
that the durability (ability to retain data when unpowered) of 
high-density NAND flash can be compared to NOR flash from many 
generations ago.

(6) is a good summary of the differences between NAND and NOR in terms 
of reliability.  Particularly interesting tidbit from that article is 
the note that reading from modern NAND devices can eventually disturb 
adjacent cells and cause data loss; NOR does not have this problem.

And then I did a research deep-dive ...

NOR flash uses the same technologies as EEPROMs, except the latter can 
be erased and reprogrammed at the byte level, whereas NOR flash can only 
be erased at the device (chip) or block level. The durability of EEPROM 
and NOR flash should therefore be comparable, as long as the same 
silicon processes are used.

This PDF from Intel(1) in September 1994 talks about the differences 
between Intel Series 1, Series 2, and Series 2+ flash cards.  All of 
these use chips with "ETOX" or EEPROM Tunnel Oxide technology.

I found this paper(2) from Proceedings of IEEE in 1993 that examined the 
reliability of different flash memory technologies, specifically ETOX 
type as well as NAND type.  The paper cites another (3) which mentions 
that ETOX-type flash was tested at 250°C for 168 hours and 10,000 
erase/program cycles, and the stored voltage was still well within 
readable parameters.

This "bake" test sounds like a less formal version of the standardized 
reliability tests now performed on some flash chips. As far as I can 
tell, no such standard existed in the 1990s, unfortunately, but with 
today's standardized tests and some mathematical models, we can 
extrapolate some expected durability performance.

For example, some test reports from STMicro state some data retention 
safe values:

* more than 200 years at 55°C or less for high density EEPROMs (4)
* more than 40 years at 55°C or less for EEPROMs embedded within NFC 
chips (5)

All this to say that temperature is probably the most important factor 
when it comes to storage durability of flash.  Also, I'm not worried 
about the data on any of my Newtons; I'm pretty confident I'll boot one 
up in another ten years and everything will still be there.

-Victor

1) 
http://intel-vintage-developer.eu5.org/DESIGN/FLCARD/APPLNOTS/29215801.PDF

2) https://www.slac.stanford.edu/exp/npa/misc/00220908.pdf

3) https://ieeexplore.ieee.org/document/572591

4) 
https://www.st.com/resource/en/application_note/an4653-cycling-endurance-and-data-retention-of-high-density-eeprom-based-on-cmos-f8h-process-stmicroelectronics.pdf

5) 
https://www.st.com/resource/en/application_note/an5085-cycling-endurance-and-data-retention-of-eeproms-in-products-of-the-st25dvi2c-series-based-on-the-cmos-f8h-process-stmicroelectronics.pdf

6) 
https://www.curtisswrightds.com/sites/default/files/2021-10/Flash-Memory-Lifespan-and-Reliability-white-paper.pdf


On 2022-09-12 22:15, Dan wrote:
> Hi all,
> I always thought that while flash had a limited number of write cycles, the actual data storage was pretty reliable.  But recently I have run across several papers that say long term storage isn't a good idea and will eventually fade away (giving a general time frame of 10 years).    Never heard of that before, and we have often talked here how a "newton never forgets" after being in a drawer for over a decade and everything is still intact as the day we put it there.
>
> Is this situation with flash something that has recently been found out?  Or is it the newer flash chips have such extreme density that they deteriorate in 10 years?  Such as MLC or even TLC vs SLC.
>
> So far all my cards still have all their data and are functioning fine.
>
> -Dan
> ----------------------------------------------------------------------
>
> http://newtontalk.net
> http://twitter.com/newtontalk