Prev: Recovering ownerships and permissions after migration
Next: SATA HDD >1TB
From: Arno Wagner on 18 Apr 2008 00:29
Previously John Turco <jtur(a)concentric.net> wrote:
> Eric Gisin wrote:
>>
>> "John Turco" <jtur(a)concentric.net> wrote in message news:4801CA94.553D1020(a)concentric.net...
>> > JW wrote:
>> >>
>> >> On Fri, 11 Apr 2008 06:10:58 -0500 John Turco <jtur(a)concentric.net> wrote
>> >> in Message id: <47FF4742.D30883(a)concentric.net>:
>> >>
>> >> >I followed the link you've provided, above...still, I'd be concerned about
>> >> >flash memory's relatively limited number of rewrite cycles (approximately
>> >> >100,000, typically), if I were you.
>> >>
>> >> Some industrial flash is good for 2 million writes. Combine that with wear
>> >> leveling this is not as bad as you might think. We've shipped over a
>> >> thousand XP embedded systems on CF in the last few years, and I've yet to
>> >> see a drive wear out.
>> >
>> > Oh, I was already aware of "wear leveling," but, didn't know about that "2
>> > million writes" stuff.
>> >
>> If you knew about wear leveling all along,
>> why do continue to post your paranoid claim about flash memory "wearing out"?


> Hello, Eric:

> No "paranoia" involved, wise guy. <g> Besides, if the Windows "page
> file" is enabled, even 2,000,000 write cycles will be used up,
> pretty quickly -- wear leveling, or not.

No flash can do 2'000'000 writes per cell. That is after wear
leveling and/or ECC. SLC flash does around 100'000 writes reliably
per cell. MLC is still stuck in the 10'000 range. However wear leveling
and ECC are facts and will extend device life significantly.
Some flash drives can even do defect remapping, AFAIK, which will
extend device life even more.

There are also things you can do to reduce wear. One thing is
realizing that pageing actually can be done on ramdisk to a degree.
Surprising, but true. Pageing is not only about having more memory,
but also about being able to use precious direcly mapped memopry in
multiple instances. Some OSes do not need to do that. Linux, for
example, runns pretty well without swap-space. XP is braind-damaged
that way and needs some minimal amount, which can be provided
on ramdisk.

Side note: The german computer magazine c't occasinally tries
to break FLASH devices by overwriting. So far they have not
succeeded with current memory sticks and SSDs.

Arno
From: JW on 18 Apr 2008 06:25
On 18 Apr 2008 04:29:00 GMT Arno Wagner <me(a)privacy.net> wrote in Message
id: <66qmccF2jc12hU1(a)mid.individual.net>:

>No flash can do 2'000'000 writes per cell.


http://ec.transcendusa.com/product/ItemDetail.asp?ItemID=TS8GCF45I-D
"Endurance: 2,000,000 Program/Erase cycles"

From: Arno Wagner on 18 Apr 2008 08:09
Previously JW <none(a)dev.null> wrote:
> On 18 Apr 2008 04:29:00 GMT Arno Wagner <me(a)privacy.net> wrote in Message
> id: <66qmccF2jc12hU1(a)mid.individual.net>:

>>No flash can do 2'000'000 writes per cell.


> http://ec.transcendusa.com/product/ItemDetail.asp?ItemID=TS8GCF45I-D
> "Endurance: 2,000,000 Program/Erase cycles"

I stated "per cell". They state nothing, so likely "per device"
as manufacturers tend to pick the biggest number.

Maybe also the "Built-in dynamic defect management and error
correction ECC technology" is a hint?

Arno
From: Eric Gisin on 18 Apr 2008 09:41
"Rod Speed" <rod.speed.aaa(a)gmail.com> wrote in message news:66qkbpF2lenmkU1(a)mid.individual.net...
> Eric Gisin <gisin(a)uniserve.com> wrote
>> John Turco <jtur(a)concentric.net> wrote
>>> Eric Gisin wrote
>
>
>>> Besides, if the Windows "page file" is enabled, even 2,000,000 write cycles will be used up,
>>> pretty quickly -- wear leveling, or not.
>
>> Raving Lunatic.
>
> Easy to claim, hell of a lot harder to actually substantiate that claim.

Trivial to measure, you stupid pile of dingo dung.
Grab an old HD, put pagefile on it, and compare disk's IO on with perfmon.
>
>> Page file I/O is *much* less than your total I/O.
>
> What matters is the write I/O. Your claim is just plain wrong with many Win systems.
>
Bullshit, Rodbot. There is much more MFT and logfile IO than pagefile IO.

>> Get some RAM if not.
>
> You still get page file write I/O even with the maximum ram the system can handle.
Which does not slow your system down, because it happens when disk are idle.

From: JW on 18 Apr 2008 10:23
On 18 Apr 2008 12:09:39 GMT Arno Wagner <me(a)privacy.net> wrote in Message
id: <66rhc3F2lh7s3U1(a)mid.individual.net>:

>Previously JW <none(a)dev.null> wrote:
>> On 18 Apr 2008 04:29:00 GMT Arno Wagner <me(a)privacy.net> wrote in Message
>> id: <66qmccF2jc12hU1(a)mid.individual.net>:
>
>>>No flash can do 2'000'000 writes per cell.
>
>
>> http://ec.transcendusa.com/product/ItemDetail.asp?ItemID=TS8GCF45I-D
>> "Endurance: 2,000,000 Program/Erase cycles"
>
>I stated "per cell". They state nothing, so likely "per device"
>as manufacturers tend to pick the biggest number.

I dunno, I'm figuring that they mean the entire device. But I could be
wrong.
First  |  Prev  |  Next  |  Last
Pages: 1 2 3 4 5
Prev: Recovering ownerships and permissions after migration
Next: SATA HDD >1TB