eeprom read write limits

hi guys

just a quick question really

i know the atmega328 has about 100,000 read write routines before it fails

but what about the 168? i am aware that the smaller the memory the more read write routines before epic fail but is it true in this case?

cheers

Per the data sheet from ATMEL, the 168 and the 328 EEPROM have the same write endurance... 100,000 writes.

The FLASH memory ( into which your sketch is loaded ) is rated for 10,000 writes.

i know the atmega328 has about 100,000 read write routines before it fails

That 100K minimum value is for erase/write cycles. There is no limit on read cycles.

Lefty

The 100,000 read/write routines are for the chip's EEPROM

The read/write cycles for running sketches (using RAM) are incredibly high (something to the 14?)

Mowcius

That 100K minimum value is for erase/write cycles. There is no limit on read cycles.

Lefty

if that is the case then brilliant ;D

By the way "failure" means the guaranteed data rentention time is not guaranteed anymore. It does not mean that it stops working immediately. It means that it may lose memory sooner. That is: if you do not need a high retention time, you can overwrite it much more often degrading the retention time further and furter.

Is it each bit that is written count as one (so a byte would = 8) or is it each time the function is called?
If I code something to write and I make a mistake so that it loops will It quickly use up the 100,000 limit?

If I code something to read or write and I make a mistake so that it loops will It quickly use up the 100,000 limit?

100,000 isn't a limit it's the guarantee from Atmel. They guarantee at least 100,000 write cycles. You could easily get more writes or a bit fewer.

Is it each bit that is read or written count as one (so a byte would = 8) or is it each time the function is called?

There isn't a count. Each time an EEPROM cell is erased or written it undergoes physical wear and tear. At a certain point, the cell just wears out.

Curious to see if anyone had actually experienced this kind of write/read failures. It seems that 100,000 cycle is not a whole lot...

A few people on AVR Freaks went to the effort to deliberately fail an EEPROM...
http://www.avrfreaks.net/index.php?name=PNphpBB2&file=viewtopic&t=85029
http://www.avrfreaks.net/index.php?name=PNphpBB2&file=viewtopic&p=609550#609550

So basically fails generally happen some time after 3M cycles. That's not bad going considering they are often guaranteed for 100,000 or so.

Mowcius

So basically fails generally happen some time after 3M cycles

Not necessarily. The point of EEPROM is to preserve the value without power. I don't think either of those tests include a power cycle.

I've been told that EEPROM can work fine, through many write cycles, until the power is turned off.

I've been told that EEPROM can work fine, through many write cycles, until the power is turned off.

Yes this is true. Back in 2001 ATMEL were selling a flash part that they claimed to have a endurance of 1000000 write cycles. I was involved in testing them and found the real life was only 80 to 120 cycles if you powered then down between writes. That was flash not eeprom but I imaging the same applies.