Heike C. Zimmerer wrote:
You're sure? Maybe I'm not up to date, but I remember figures in the
order of typically around 100,000 or even as low as 10,000, depending
of the kind of write cycle (like MLC) for today's standard devices. A
pointer to the data of the device actually used by the 770 would be
much appreciated.
Well, I don't know what the Nokia uses, but the Flash chips I've worked
with (AMD and others) quoted 100K as a bare minimum with 1M as typical.
In the "bad old days" when the programming cycle was controlled by the
computer, and the timing was at best approximate, the chips tended to
get burned out. Now, that the write cycle is controlled by the chip
itself, and thus is tuned to the process, the chip does not overstress
the oxide layer.
Again, maybe I'm not up to date: isn't wear leveling only effective if
there's enough room left on your flash which gets freed from time to
time? IOW, the more static data there is on a flash (like programs
and fixed data), the less is gained from any leveling algorithm?
You always do wear leveling, because of the block erase nature of flash
- you have to keep a certain amount of the flash in reserve so that when
a write occurs you have a place to put the new data until you have a
full block.
I've done both low-level flash programming (including the bad-ole-days
timing with NOPs and JMP $+2's) and flash file systems before.
_______________________________________________
maemo-developers mailing list
[email protected]
https://maemo.org/mailman/listinfo/maemo-developers
