Jump to content











Photo
- - - - -

How to refresh the contents of a flash drive?


  • Please log in to reply
45 replies to this topic

#1 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 06 February 2011 - 12:49 PM

As you may know, if you put a flash drive for something like 10 years into a drawer and then try to access it, all data will most likely be corrupted, if not the drive itself is unformated.

To stop this from happening the contents need to be refreshed from time to time and here comes my question.

It is clear that moving the content off the drive and back on, will result in a refresh.
However, will simply connecting the drive periodicly to power do the same?


:ph34r:

#2 Wonko the Sane

Wonko the Sane

    The Finder

  • Advanced user
  • 14919 posts
  • Location:The Outside of the Asylum (gate is closed)
  •  
    Italy

Posted 06 February 2011 - 08:19 PM

However, will simply connecting the drive periodicly to power do the same?

Hmmm. :dubbio:

Nice consideration/thought. :thumbsup:

More questions:

Can you define "periodically"?

By "connecting to power" you do mean "connecting to power only" (example: powered USB HUB not connected to anything but mains through adapter or USB charger) or to a "powered computer with an OS running"?

Is the original effect and proposed timespan reliable/verified?

Does it apply to a particular type of flash memory?

Are we talking of NAND or NOR (or SLC or MLC, or whatever)?

A couple of docs to "warm" the thread ;):
http://www.kingston....h_memory_guide/
http://klabs.org/ric...c4_miyahira.doc

And a presumably dedicated blog :unsure::
http://blog.dataligh...fits/flash_life


:cheers:
Wonko

#3 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 06 February 2011 - 10:21 PM

More questions:

Can you define "periodically"?

By "connecting to power" you do mean "connecting to power only" (example: powered USB HUB not connected to anything but mains through adapter or USB charger) or to a "powered computer with an OS running"?

Is the original effect and proposed timespan reliable/verified?

Does it apply to a particular type of flash memory?

Are we talking of NAND or NOR (or SLC or MLC, or whatever)?

Havn't read the topics linked to. Just a few quick answers.

1. periodicly - duing something repeatedly in intervalls (in this case - way shorter than 10 years :dubbio: )
2. connecting to power - just power (Would it make a difference, if it were a "powered computer with an OS running"?)
3. The effect is verified, the lifespan, i guess, is projected.
4. No it applies to all EPROM. But the time varies for different versions.
10years is the value i've read mentioned for cheap NAND like in USB Sticks, NOR is supposedly better, though i can't see why.
From the elecrical point of view, i would think, that circuits that allow faster writing as well as circuits that have been deleted and rewritten more often, will hold their charge less perfectly.


:thumbsup:

#4 Wonko the Sane

Wonko the Sane

    The Finder

  • Advanced user
  • 14919 posts
  • Location:The Outside of the Asylum (gate is closed)
  •  
    Italy

Posted 07 February 2011 - 08:39 AM

2. connecting to power - just power (Would it make a difference, if it were a "powered computer with an OS running"?)

Well, I think we have to draw a line somewhere between "intelligent" devices (like USB flash sticks) i.e. those that do have a controller inside and "dumb ones" (like CF or SD cards).
For the latter ones powering with or without an OS should not :cheers: make a difference, for the former ones, it is possible that to actually "get the power to the chip", the controller needs to be initialized (let's say by the OS or BIOS probing the USB bus) :dubbio:
Probably I am overcautious, but what if when in 2021, after having duly powered your USB stick at say 6 to 12 months intervals, you find that your stick has not anymore your data because of this? ;)

Idea :thumbsup:
Create a "USB stick sheath" or "USB stick scabbard".
Materials needed:
  • A BIG battery, say this type:
    Posted Image
  • A resistor, current limiting, whatever to draw the 5 V from it with a max of 200 or 500 mA
  • A timer, switching on every six months for, say, 5 minutes.
  • One or more Female USB type A connectors.
  • ...or maybe a Li-Ion or NiCD battery and a connection to mains?

OT, always remember that the hand is faster than the eye :unsure::


:doh7:
Wonko

#5 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 07 February 2011 - 01:50 PM

Read the 3 documents you've linked to.
The blog is useless, the one from NASA deals with a completely different point and the Kingston one, just states what i've already wrote in the first post. Data Retention up to 10 years.

Well, I think we have to draw a line somewhere between "intelligent" devices (like USB flash sticks) i.e. those that do have a controller inside and "dumb ones" (like CF or SD cards).

I don't know if CF and SD Cards are as dumb as you think. Doesn't all flashmemory use an internal controller theses days, to distribute the wear and tear more evenly?

I originally thought only about USB-Sticks, but with your point above, i remembered that something else will have this problem too, SSD-Drives!


:cheers:

#6 Wonko the Sane

Wonko the Sane

    The Finder

  • Advanced user
  • 14919 posts
  • Location:The Outside of the Asylum (gate is closed)
  •  
    Italy

Posted 07 February 2011 - 04:03 PM

Read the 3 documents you've linked to.
The blog is useless, the one from NASA deals with a completely different point and the Kingston one, just states what i've already wrote in the first post. Data Retention up to 10 years.

Obviously I did not read "them", I read between the lines of them, and curiously enough came up to a number of interesting IMHO docs from datalight, which I won't share, but that you can - optionally - read by subscribing to their site.

This one I can share:
http://download.micr...inHEC_Cooke.pdf
(please don't read it, as it is - still obviously - completely unrelated, though maybe page 50/61 may be of interest to you).

Something else that you should NOT read, as it is so totally unrelated :) comprehends:
http://www.cl.cam.ac...em_CHES2005.pdf
http://www.eetasia.c...OURCES=DOWNLOAD

But what if the "for something like" was nearer to 100 years then to 10 years? :cheers:

:whistling:
Wonko

#7 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 07 February 2011 - 05:41 PM

Obviously I did not read "them", I read between the lines of them, and curiously enough came up to a number of interesting IMHO docs from datalight, which I won't share, but that you can - optionally - read by subscribing to their site.

Well, if you think you've read something useful, regarding this topic, in those documents, why don't you tell us?

But what if the "for something like" was nearer to 100 years then to 10 years? :dubbio:

Then data retention would be less of a problem, but it would still not answer the question if simply applying power would refresh the data or not.

:whistling:

#8 Wonko the Sane

Wonko the Sane

    The Finder

  • Advanced user
  • 14919 posts
  • Location:The Outside of the Asylum (gate is closed)
  •  
    Italy

Posted 07 February 2011 - 06:03 PM

Well, if you think you've read something useful, regarding this topic, in those documents, why don't you tell us?

Because if I would do so that would invalidate my new signature (someone could mistakenly take what I post for valid/meaningful)? ;)

Then data retention would be less of a problem, but it would still not answer the question if simply applying power would refresh the data or not.

Sure, the question remains the same :whistling: , but the level of urgency to find an answer to it may be slightly lessened and in the meantime some more info may be available.
As you can understand - even by NOT reading the documents, any info is coming either from the manufacturers or from academic/theoretical work, and it is VERY scarce (and - as I see it - largely unproved/untested/unverified).
I think that there is NOT a valid answer available right now, but as always I may well be wrong :dubbio:
The amount of different technologies for the Flash, the possible different "quality" of flash chips from different manufacturers - even with exactly the same technologies - the presumed effect (the one that you should NOT read on page 49 and 50/61) of wear leveling and of "read disturb", the difference behaviour of controllers and what not create so many possible variables, that it makes it impossible to "know for sure".

The only thing we can rely upon right now is "common sense" and "better be safe than sorry" approaches, i.e. do a "full filesystem refresh" periodically AND have a backup on other media, the nice :cheers: alternative of just powering on the thingy from time to time - which can possibly be enough - not enough documented to rely upon :unsure:.



:)
Wonko

#9 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 07 February 2011 - 07:48 PM

What i've gathered, from the various documents, you've linked to, is basicly the same, i already assumed above.

- The more often a cell is erased/written the lower the data retention time.
- The faster a cell can be erased/written the lower the data retention time.
- NOR cells take significantly longer to erase than NAND ones and have a longer data retention, which would suggest that there is also another difference, than just the implied difference in the used logic circuit.
- SLC flash has a higher data retention time, due to it storing only 2 instead of 4 states, which makes it four times as easy to tell them apart.

New to me was that the reading of a cell does also cause the data retention time to drop.
This would be most important for storing an OS and less for data.

Data retention is given in one of the documents for a single write, never read to be greater than 1000 years under ideal temperature and still more than 200 years at 50°C.

The same document later gives the data retention time as only 16 years at 50°C, when data refresh is discussed.
The given methode to do it, is the simple copy and delete methode. An in place refresh (just power approach) is not mentioned.


:dubbio:

#10 sbaeder

sbaeder

    Gold Member

  • .script developer
  • 1338 posts
  • Location:usa - massachusettes
  •  
    United States

Posted 08 February 2011 - 04:43 AM

- The more often a cell is erased/written the lower the data retention time.

Not sure that there is a big difference her at the physics level on the transistor...Yes, MAYBE you could make a claim that there is some "wear and tear" on the device by changing it's state, but that shouldn't be a major issue here...(definately not a first order effect).

- The faster a cell can be erased/written the lower the data retention time.

And again, not sure even with those posts above that this is a first order effect. at the end of the day, it MAY be related just because a faster transistor means it is smaller, and might have more "leakage" which at the end of the day is what is going to make a difference here...How fast does the charge "leak"...

- NOR cells take significantly longer to erase than NAND ones and have a longer data retention, which would suggest that there is also another difference, than just the implied difference in the used logic circuit.

Now you are starting to get "warmer"...i.e. it is the way the transistors are actually created in the memory "cell" and how they are trying to encode the data...

- SLC flash has a higher data retention time, due to it storing only 2 instead of 4 states, which makes it four times as easy to tell them apart.

Also a bit over simplified...but close enough...

New to me was that the reading of a cell does also cause the data retention time to drop.
This would be most important for storing an OS and less for data.

Data retention is given in one of the documents for a single write, never read to be greater than 1000 years under ideal temperature and still more than 200 years at 50°C.

The same document later gives the data retention time as only 16 years at 50°C, when data refresh is discussed.
The given methode to do it, is the simple copy and delete methode. An in place refresh (just power approach) is not mentioned.

I guess the bottom line is that all data - even the stuff carved in rock tablets - has a finite life. Paper also will deteriorate over time, even in ideal conditions, and magnetic media and other "modern" forms of data storage also have their issues...I guess the issue is that if you want something to be "permanent", you're out of luck...Next best thing is to just make sure to double check things often and update them to newer technology if available...

But an interesting discussion :buehehe:

#11 Wonko the Sane

Wonko the Sane

    The Finder

  • Advanced user
  • 14919 posts
  • Location:The Outside of the Asylum (gate is closed)
  •  
    Italy

Posted 08 February 2011 - 09:18 AM

So, now that we have supposedly explored the theories, we need to get practical. :w00t:

Setting appointments every (say) 3 1/2 years to refresh data (3 1/2years being 1/3 of the guaranteed data retention from Kingston seems logical) is not very handy, humans are "programmed" for - at the most - yearly recurrences.

:ranting2:
  • Let's institute "Flash Refreshing Day", we could make it a day like the 26th of December, or some other common holiday so that people is at home and can refresh their flash.
  • What about an automatical refresher? A small HD enclosure with a USB female connector to which you plug in a stick and with three leds, Yellow, Red, Green:

  • Yellow: reading flash
  • Red: writing flash
  • Green: competed and verified
It could be a way to "recycle" old, low capacity, laptop 2.5" hard disks :buehehe: and at the same time allowing grandfathers and grandmothers to be able to refresh their flashes effortlessly. :)

:cheers:
Wonko

#12 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 08 February 2011 - 12:21 PM

Not sure that there is a big difference her at the physics level on the transistor...Yes, MAYBE you could make a claim that there is some "wear and tear" on the device by changing it's state, but that shouldn't be a major issue here...(definately not a first order effect).

It's exactly this effect, that limits the number of possible write cycles to a cell. Which at the moment is the effect that causes the most concern and is hardest worked on to overcome.
So i would definitely call it a first order effect.

However, if you don't fully agree with what i said, why don't you write a post about, how it works in your opinion?
I like to learn new things.


:smiling9:

#13 paraglider

paraglider

    Gold Member

  • .script developer
  • 1729 posts
  • Location:NC,USA
  •  
    United States

Posted 08 February 2011 - 01:13 PM

The key information from the Kingston doc is this statement:

Kingston Flash Data Retention: Kingston Flash storage devices are rated for up to 10 years under normal use. Important information should also be backed up on other media for long-term safekeeping.


I would read, erase and rewrite the drive every couple of years maybe upgrading to more recent technology at the same time.

#14 steve6375

steve6375

    Platinum Member

  • Developer
  • 7033 posts
  • Location:UK
  • Interests:computers, programming (masm,vb6,C,vbs), photography,TV,films,guitars, www.easy2boot.com
  •  
    United Kingdom

Posted 08 February 2011 - 01:19 PM

I have asked my Kingston contact for a definitive answer and emailed him the Qs.

What I am unclear about is, if reads can reduce the charge in a cell, the why don't they specify a 'retention time at xx reads per second' figure? If you have a flash SSD drive and use it for 5 years in 'read-only' mode (e.g. large fast database) then are you in danger of getting corrupt data?

Maybe some devices have ECC correction and if a block starts getting ECC errors then it re-writes (read/erase/write) or re-allocates?

#15 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 08 February 2011 - 01:42 PM

What I am unclear about is, if reads can reduce the charge in a cell, the why don't they specify a 'retention time at xx reads per second' figure? If you have a flash SSD drive and use it for 5 years in 'read-only' mode (e.g. large fast database) then are you in danger of getting corrupt data?

YEs, the stuff about the danger of excessive reading, as they call it, was news to me too. Until now i had only ever heared about the danger of "excessive" writing.

Maybe some devices have ECC correction and if a block starts getting ECC errors then it re-writes (read/erase/write) or re-allocates?

I would guess that SSD drives have the same error prediction and protection like a normal HDD, in this regard.


:smiling9:

#16 Wonko the Sane

Wonko the Sane

    The Finder

  • Advanced user
  • 14919 posts
  • Location:The Outside of the Asylum (gate is closed)
  •  
    Italy

Posted 08 February 2011 - 01:42 PM

The point I was previously trying to make is that 10 years is "reasonable" AND that it is a "rated" value.

Other sources are more like 50 year or 100 years.

It is clear enough from the gathered documentation that data retention time is NOT carved in stone AND it depends both on the technology used AND on the actual use of the device (number of writes and - surprisingly enough - number of reads too).

If I were from Kingston, or from any other manufacturer, I would try to "rate" my product to anything that is:
  • "reasonable"
  • "one-size-fits-all"
  • "cautelative enough"

I have the impression that the "10 years" represents the above and not actual results of tests/experimentation.

That would be more like conditioned to a number of factors, of which we seem like being able to consider AT LEAST:
  • number of writes
  • number of reads
  • temperature of the storage

From the things that I have not read ;), it seems to me like a flash chip that:
  • has been extensively written
  • has been extensively read
  • stored in a metal box in the open in (say :cheers:) Nigeria

will have quite a different expected retention time from the SAME exact model that has been:
  • written ONCE
  • read ONCE (only to verify the successful write above)
  • stored in a drawer in a (say :)) Pharmacy/Chemist (as an example of a climatized environment with constant temperature all the year round)

So, do the "10 years" represent the former or the latter hypothetical scenario?

I don't think it can't represent both. :smiling9:

:ranting2:
Wonko

#17 steve6375

steve6375

    Platinum Member

  • Developer
  • 7033 posts
  • Location:UK
  • Interests:computers, programming (masm,vb6,C,vbs), photography,TV,films,guitars, www.easy2boot.com
  •  
    United Kingdom

Posted 08 February 2011 - 01:49 PM

Sure - 10 years is just a finger in air stab (and probably conservative to be on the safe side). If data does need to be refreshed by read/write cycling the whole flash drive, then how come we don't see 'refresh' programs that do this? Maybe this is a new feature for inclusion in Windows 8 :smiling9:

#18 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 08 February 2011 - 05:40 PM

So, do the "10 years" represent the former or the latter hypothetical scenario?

I don't think it can't represent both. :dubbio:

There is something called the guaranteed data retention time, a time the data will be at least save under worst circumstances, not destroying the hardware. Cells of current SSD from SanDisk, for instance, are rated to have a guaranteed data retention time of the very last write, before they die, of at least 1 year.

Unfortunately the Kingston document doesn't state things as clearly, by saying 'up to' and not naming clearly, what kind of value they are giving. (maybe steves Kingston source will give better answers)


:cheers:

#19 Wonko the Sane

Wonko the Sane

    The Finder

  • Advanced user
  • 14919 posts
  • Location:The Outside of the Asylum (gate is closed)
  •  
    Italy

Posted 08 February 2011 - 06:18 PM

I would like to have the same kind of info as this (which is for FRAM) document:
http://www.fujitsu.c...fram/index.html

for the Flash chips. :dubbio:

This is another hint:
http://cache.freesca...letin/EB618.pdf
that the 10 years is a "commercial/safe/one-size-fits-all" value.

Though of course we also have NO way to verify that the acceleration given by baking the chips is actually correspondent to an actual timescale. :unsure:

Definition
The intrinsic data retention is by definition inherent to all bits manufactured with the same process
technology. It is an estimate, based on accelerated stress data and the Arrhenius model, of the data
retention life that most bits are expected to achieve.
The Arrheniusmodel is an industry standard for estimating data retention life of floating gate technologies.
It is used to find the acceleration factor between a stress temperature and a use condition, which in turn
can be used to de-rate results from an accelerated stress test.


During the technology certification process, Freescale Semiconductor determines the intrinsic activation
energy Ea for a technology with empirical data. This is done by evaluating time-to-failure under high
temperature stress.


In a nutshell, a mathematical model BTW calculated taking as base a non-better defined "empirical value" is applied, and SINCE the result of this (supposedly accurate :cheers:) model results in a "Typical value" ALWAYS > 100 years (which is a non-value, 101, 1000, 10000 and 100000, as well 1 zillion years :w00t: do fall in the categorization) THEN a guaranteed 10 years is given.

:cheers:
Wonko

#20 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 08 February 2011 - 06:54 PM

This is another hint:
http://cache.freesca...letin/EB618.pdf
that the 10 years is a "commercial/safe/one-size-fits-all" value.

The bigger problem, i see with that document, is that the tests are done at a once and only write situation. So meassured data retention is bound to be the highest possible.

I would love to see a test about data retention, where a USB-Stick had a good workout prior to testing.
Something like 10000 write-delete cycles for MLC NAND, so we can be sure we get the lower end of what's possible and not only the max.

:dubbio:

#21 Wonko the Sane

Wonko the Sane

    The Finder

  • Advanced user
  • 14919 posts
  • Location:The Outside of the Asylum (gate is closed)
  •  
    Italy

Posted 08 February 2011 - 06:59 PM

The bigger problem, i see with that document, is that the tests are done at a once and only write situation. So meassured data retention is bound to be the highest possible.


Sure :cheers: it was just cited as a reference to the methodology presumably used, and of the kind of accuracy (or lack of it :dubbio:) of the results published.

:unsure:
Wonko

#22 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 08 February 2011 - 07:33 PM

btw. Got just an answer from SanDisk. Telling me about their warranty. Guess the qualified technician housewife answering, must have been confused by the word guaranteed. :dubbio:

:cheers:

#23 Wonko the Sane

Wonko the Sane

    The Finder

  • Advanced user
  • 14919 posts
  • Location:The Outside of the Asylum (gate is closed)
  •  
    Italy

Posted 08 February 2011 - 07:38 PM

btw. Got just an answer from SanDisk. Telling me about their warranty. Guess the qualified technician housewife answering, must have been confused by the word guaranteed. :dubbio:

Yep :cheers:, the future is now:
Posted Image

:w00t:

:unsure:
Wonko

#24 steve6375

steve6375

    Platinum Member

  • Developer
  • 7033 posts
  • Location:UK
  • Interests:computers, programming (masm,vb6,C,vbs), photography,TV,films,guitars, www.easy2boot.com
  •  
    United Kingdom

Posted 10 February 2011 - 12:16 PM

I have had a few emails from Kingston trying to get a clear answer and almost succeeded (!), here are the Q's I asked and Kingston's response:

Which of these would refresh the memory for another 10 years (say)?
1. Just plug into a 5V USB port and leave it for a while (I assume no refresh would occur)
2. Plug into a computer and read all sectors on the drive (I assume no refresh would occur)
3. Plug into a computer and read and then write back all sectors on the drive (I assume this would refresh)
A side question is,
4. If a flash memory was filled with important data and then from that date on, it was only read/accessed by a computer but never re-written, would the 10 years (or whatever the data retention spec. says) for the drive still apply. i.e. does the data retention period apply for unpowered flash, powered flash and read-only flash memory (or even, is the data retention period actually shorter if the cells are read a lot during the 10 years, and if so, by how much is the data retention period shortened)?
(in particular I would like to know if constant read access to cells could lead to their loss of charge and if so by how much – i.e. negligible or shorten data renetion by 50%)

i.e. if I constantly read Kingston flash memory as fast as I could for 3 years (with no writes) would you still ‘guarantee’ the data would be intact after 3 years?



Kingstons response was : The customers response is correct for points 1-4.
Utilizing the drive as a read-only device would reduce any likelyhood of an uncorrectable error occuring, however, a unrecoverable data integrity error would have a near impossible likelyhood to occur even within the products guaranteed life-time configured as a read/write device. SSDs implement multiple fail-safes via detecting and correction methods to ensure that an uncorrectable error does not occur even beyond the products life-time. Hard drives also feature error correcting features since they are suspectible to the same laws of physics that apply to any other storage device.

There is no guarantee on data integrity as there is no standard to measure data integrity on SSDs for archival purposes by which we could test the product by. No storage device manufacturer offers a guarantee that guarantees data integrity on SSDs. Since there is no extensive proof from studies on the degradation of charge in a NAND Flash device there is no way to easily extrapolate the integrity expectancy, however, there are investigations underway to enhance and retrieve further information in this field via the IEEE and other communities.

Our current generation of SSDs are not meant for archival purposes but as performance upgrade solutions.

AND

Flash products built by Kingston are not built for archival purposes but if they are used in this capacity the best way to ensure the data integrity is maintained throughout the life-time of the product is to read all the data off the drive, perform a HDDErase cycle to clear all blocks and lastly, rewrite all the data back to the drive.

Unfortunately there is little information available from studies into this field to decisively state the exact frequency that this should be done since the error correction capability varies between the various NAND flash mediums and their featured controllers/software. There on-going discussions by IEEE and various standards groups to enhance and study this field further.

Hard disk drive manufacturer readily recommend performing a read and re-write on their hard disk drives every 3 years for archival purposes and I would suggest doing the same with NAND flash based products, even when the contents of the stated NAND flash based products are not modified by a read-erase-write cycle.


So, I take it from this response that if reading flash memory causes a CRC error to be detected then the flash controller would probably re-allocate the page if it could (???) or at least fix it with ECC so the data is not corrupt. However, I would question whether ECC correction is relevant for flash memory as ECC works for small runs of bad bits (e.g. as with a damaged hard disk sector), whereas flash memory cells (like CCDs in cameras) would just have bad cells at 'random' places within a page??? So ECC would in practice only be good for single bit errors?

At least this goes some way to answering the OP's question though...

So how long before someone writes a 'Flash drive refresh program' and would it work if the drive had intelligent caching (as it would not bother to actually do a read/erase/write cycle if you wrote the same data back) and how could you check if it actually had done a r/e/w cycle anyway? Hence, I guess, the Kingston guys recommending an HDDErase cycle? :cheers:

Answers on a postcard to ... (here!)

#25 Wonko the Sane

Wonko the Sane

    The Finder

  • Advanced user
  • 14919 posts
  • Location:The Outside of the Asylum (gate is closed)
  •  
    Italy

Posted 10 February 2011 - 12:35 PM

It seems we are back almost, but not quite, at the initial impractical conclusion that you should refresh every 3 1/2 years (derived from "common sense" + "better be safe than sorry"). :cheers:

The only difference is that before we presumed it was "a finger in air stab", now we know it is :ph34r: AND that it is unlikely we'll get a more accurate "refresh interval" backed up by "actual reasons".

Since HDDERASE is part of ATA/ATAPI I doubt :angry7: you can use it on a USB stick. :ph34r:

Maybe a "0 fill" is better suited/adviced/possible on them.

The Kingston guys seem like having done a good job of mishmashing USB sticks and SSD's (in my opinion they are NOT the same thing).
And of course the reference to the hard disk suggested refreshing is completely pointless, chances that "a" magnetical technology (and there are quite a few of them used in the years) has exactly the same "decaying time" then a solid state, flash ( which of the several technologies) one are as I see it VERY little. :ph34r:

As a layman comparison:
Since I change the oil of my car engine every 30,000 Km or three years, then I recommend that you refresh the contents of each and every stored data, no matter the media on which it is, the same day you make the oil change.

Hey, that could be an actual way to remember the "R-day"! :ph34r:

:cheers:
Wonko




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users