experience with max block erase

experience with max block erase

[dennis noermann]

Hi
i am intrested in testing some Intel Flash
how much erase cycles it realy can do

does someone has done this before ?

i have some intel strata flash 28F640J3A

"Each block can be independently erased 100,000 times"

is this like the "food expiry date" ore "best before date" ?
(sorry i dont know the right englisch word , but i think you undertstand it :) )


ore have i to write some eraseprogramm and wait a long time :)

dennis

Re: experience with max block erase

[Vipin Malik]

I would take their word for it as there is no way you can infer anything from
one (or even a couple of) tests.

I take it to mean that they "gurantee" that each sector can _at least_ handle 100K
erases without any chip failures or erase failures.
That's just my take.

AMD claims (gurantees?) 1million erase cycles/sector!

Vipin


dennis noermann wrote:

> Hi
> i am intrested in testing some Intel Flash
> how much erase cycles it realy can do
>
> does someone has done this before ?
>
> i have some intel strata flash 28F640J3A
>
> "Each block can be independently erased 100,000 times"
>
> is this like the "food expiry date" ore "best before date" ?
> (sorry i dont know the right englisch word , but i think you undertstand it :) )
>
> ore have i to write some eraseprogramm and wait a long time :)
>
> dennis
>
> ______________________________________________________
> Linux MTD discussion mailing list
> http://lists.infradead.org/mailman/listinfo/linux-mtd/

Re: experience with max block erase

[Cam Mayor]

On Wednesday 16 January 2002 18:50, you wrote:
>
> "Each block can be independently erased 100,000 times"
>
> is this like the "food expiry date" ore "best before date" ?
> (sorry i dont know the right englisch word , but i think you undertstand it
> :) )

yes.  _statistically_ it should work at least the advertised number of times. 
 after that, it may work, but they won't guarantee it.  there is also a 
nonzero chance that they'll stop working or degrade before that time, much 
like that jar of mayonnaise i had in the refrigerator that my wife forbade me 
to eat anymore after she decided it was expired.

cheers,
cam

Re: experience with max block erase

[Herman Oosthuysen]

In practice, your product will probably fail due to some other reason,
before the flash memory will fail, so don't worry too much about it.  For
instance, if the product has a keyboard of sorts, that will almost certainly
fail first and if it has batteries, it will go to the big bit bucket in the
sky even sooner.

100 000 cycles is lot.  You have to erase the chip about ten times per hour
continuously for 100 years to reach that number...
--
Herman Oosthuysen
Herman@WirelessNetworksInc.com
Suite 300, #3016, 5th Ave NE,
Calgary, Alberta, T2A 6K4, Canada
Phone: (403) 569-5688, Fax: (403) 235-3965
----- Original Message -----
From: dennis noermann <dennis.noermann@noernet.de>
To: <linux-mtd@lists.infradead.org>
Sent: Wednesday, January 16, 2002 5:50 PM
Subject: experience with max block erase


> Hi
> i am intrested in testing some Intel Flash
> how much erase cycles it realy can do
>
> does someone has done this before ?
>
> i have some intel strata flash 28F640J3A
>
> "Each block can be independently erased 100,000 times"
>
> is this like the "food expiry date" ore "best before date" ?
> (sorry i dont know the right englisch word , but i think you undertstand
it :) )
>
>
> ore have i to write some eraseprogramm and wait a long time :)
>
> dennis
>
>
>
>
>
>
> ______________________________________________________
> Linux MTD discussion mailing list
> http://lists.infradead.org/mailman/listinfo/linux-mtd/

Re: experience with max block erase

[Vipin Malik]

Herman Oosthuysen wrote:

> 100 000 cycles is lot.  You have to erase the chip about ten times per hour
> continuously for 100 years to reach that number...

Let's state 100K cycles another way:

for a 8mbit chip (1024KB)

It's only 100K * 1024KB of data life (assuming a circular use- which is the
best case and what you will get with wear leveling).

If you want your product to last 8 years (pick a number), then the max data
rate that you can send to the flash chip is "only":
100000 * 1024/(8*365*24*60*60) = 0.4KB/sec!

That's not a lot. This does not even take into account the extra data generated
due to the overhead of any filesystem one may be using. Tack in another 20% for
that (not unusual for JFFS2 system- it depends on write sizes- this assumes a
small write size/write) then the data rate falls to 0.32KB/sec

OR let's say that the system is generating data (that needs to be saved to
flash) at the rate of 10KB/sec (not very high number too), then the life of the
system is only 0.32years! (Yikes!)

Vipin

Re: experience with max block erase

[Herman Oosthuysen]

Well, if you have to write data to flash continuously, then you are using
the wrong media.  Flash is designed for stuff that doesn't have to change
often.

Anyhoo, getting a million or more cycles from flash is not unrealistic, but
then you probably would get some failures in some products and the erase
cycle time would get unacceptably slow eventually.

The important thing to understand is that a flash disk drive isn't one...

BTW, I like your writeups Vipin - it is well presented.
--
Herman Oosthuysen
Herman@WirelessNetworksInc.com
Suite 300, #3016, 5th Ave NE,
Calgary, Alberta, T2A 6K4, Canada
Phone: (403) 569-5688, Fax: (403) 235-3965
----- Original Message -----
From: Vipin Malik <vipin.malik@daniel.com>
To: Herman Oosthuysen <Herman@WirelessNetworksInc.com>
Cc: dennis noermann <dennis.noermann@noernet.de>;
<linux-mtd@lists.infradead.org>
Sent: Thursday, January 17, 2002 2:41 PM
Subject: Re: experience with max block erase


> Herman Oosthuysen wrote:
>
> > 100 000 cycles is lot.  You have to erase the chip about ten times per
hour
> > continuously for 100 years to reach that number...
>
> Let's state 100K cycles another way:
>
> for a 8mbit chip (1024KB)
>
> It's only 100K * 1024KB of data life (assuming a circular use- which is
the
> best case and what you will get with wear leveling).
>
> If you want your product to last 8 years (pick a number), then the max
data
> rate that you can send to the flash chip is "only":
> 100000 * 1024/(8*365*24*60*60) = 0.4KB/sec!
>
> That's not a lot. This does not even take into account the extra data
generated
> due to the overhead of any filesystem one may be using. Tack in another
20% for
> that (not unusual for JFFS2 system- it depends on write sizes- this
assumes a
> small write size/write) then the data rate falls to 0.32KB/sec
>
> OR let's say that the system is generating data (that needs to be saved to
> flash) at the rate of 10KB/sec (not very high number too), then the life
of the
> system is only 0.32years! (Yikes!)
>
> Vipin
>
>
> ______________________________________________________
> Linux MTD discussion mailing list
> http://lists.infradead.org/mailman/listinfo/linux-mtd/