Bitrot
Beyond variances between NAND type and design, for JEDEC claims interpretively given unrefreshed SSD memory, within an inoperable stasis for an least affected, pure state of hypothetical storage, that data loss is a potential factor within a week;- whereas another and proposition given a web-based hardware site is one of related temperature coefficients, such that the lower temperature measure over storage, than what characteristically a SSD is designed to operate at upon being written to, is a (presumption the) lower subsequent inoperable temperature will relate adversely upon expected longevity of data cohesion (an extensively suppositional if populist allowance at quite some leeway exceeding a former JEDEC's published reception). (And apt dated for an observation posted five years ago in a server-type forum.) |
Bitrot
Flasherly wrote:
Beyond variances between NAND type and design, for JEDEC claims interpretively given unrefreshed SSD memory, within an inoperable stasis for an least affected, pure state of hypothetical storage, that data loss is a potential factor within a week;- whereas another and proposition given a web-based hardware site is one of related temperature coefficients, such that the lower temperature measure over storage, than what characteristically a SSD is designed to operate at upon being written to, is a (presumption the) lower subsequent inoperable temperature will relate adversely upon expected longevity of data cohesion (an extensively suppositional if populist allowance at quite some leeway exceeding a former JEDEC's published reception). (And apt dated for an observation posted five years ago in a server-type forum.) There may be some observations possible at end-of-life, but generally not seen during normal life period. 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). Paul |
Bitrot
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. |
All times are GMT +1. The time now is 01:17 PM. |
Powered by vBulletin® Version 3.6.4
Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
HardwareBanter.com