On Fri, 09 Apr 2021 07:00:14 -0400, Paul
wrote:

Writing at low temperature, increases damage to flash cells
but the data lasts longer.

Writing at high temperature, comes closer to annealing,
reduces wear, but may also have the data lasting less long.

Powerful ECC is used to hide these factors and make the
device function for 3000 writes on TLC (less for QLC).

My best SSDs (a medium rating) run as much as 20 degrees (F) cooler
than alongside a couple of (lesser rated and most inexpensive) QLC
class and decidedly hotter operating SSDs. Were incidence leveled,
however, into some hypothetical independence, for neither having
occurred significance for their main matrix of total terabytes
written, whereupon removal of the SSD is effected in its entirety, of
one residual to a storage consideration, temperature, independent of
electrical operation, most will then factor for SSD data retention
cohesion. Anywhere from 1 to 404 weeks seems a peculiar way to toss
the dice. (Anandtech's, indirectly, unfortunately for their link not
to have survived my software buffer, as I'd formerly construed the
below. . .)

-

This would vary by NAND type, manufacturer, etc., and personally, I am
not aware of any concrete data on the subject. I have not seen anyone
do an experiment like many people did to test endurance. With that
said, there was a controversial article last year which was based on
some JEDEC/Intel data which showed that under some conditions SSDs
could lose data in as little as a week. Here's Anandtech's take on it:
The Truth About SSD Data Retention -- The basic takeaway from all of
that is that data retention periods will depend on SSD's operating
temperature and its storage temperature. If you manage to write your
data to an SSD that is running at 55C and then store it at 25C, you'd
be looking at 404 weeks of data retention for a client SSD.