What happen if make write data to "ST EEPROM IC M24C64-F" the same address more than 4 million write cycles (spec)?

[User16715311169089326328]

Hello: 

I have a ST EEPROM IC type is "M24C64-F". If I try make write data to the same address over than spec write cycles(4million) , so, does it affect other address's retention data? Does it only affect the data of the address that is written repeatedly?

[Muhammed Güler]

If you write data in a cell too many times, that cell will be corrupted. If million writes aren't enough, I suggest you look into FRAMs that support trillion writes.

[Peter BENSCH · Accepted solution]

To be a little more precise: the guaranteed Write Cycle Endurance of 4 million write cycles applies at the specified ambient temperature of ≤25°C and suitable supply voltage.

As soon as this temperature range is exceeded, the write cycle endurance decreases. For simplicity, a value of 1.2 million at 85°C is given here.

All values of the Write Cycle Endurance do not mean that the 4'000'001st or 1'200'001st write access necessarily destroys the respective byte, the success of the write is just no longer guaranteed.

Regarding the second question: the number of write cycles applies to each individual byte, regardless of the number of cycles of the neighbouring byte, for example.

Does it answer your question?

Regards

/Peter

[Tesla DeLorean]

You get bit rot

[User16715311169089326328]

Thank you Mr. Peter for your reply!

I have another question to ask, that is, I recently had a customer's device with a ST EEPROM IC model M24C64-F, and the customer reported that the IC had an unexpected value in a cell at a certain address (0x57E) (EX: 0x84-->0xA4), causing the device fail to start.

It has been confirmed many times that the MCU program does not write the value 0xA4 in this address, so what will cause the value retention in EEPROM to be changed? Could it be a defect of the IC itself?

[Peter BENSCH]

Well, there can be several reasons for a bit of an EEPROM cell to inadvertently flip to a different value. One would be a not quite complete write of the original value, another reason could be an external influence like radiation.

If data security requirements are high, it therefore makes a lot of sense to take additional security measures, such as EEPROMs with ECC (e.g. M24C64-A125), see also AN2440.

Regards

/Peter

[Tesla DeLorean]

If not at an MCU level, perhaps the integrity of the I2C communication, ie speed, pull-up, etc. That would be hard to log and validate. I2C can also support multiple byte writes, and might need to manage page/sector boundaries. ​

[User16715311169089326328]

Do you mean that even if there is no real write command from the MCU, there will be a chance to trigger the possibility of the EEPROM being written?

[Peter BENSCH]

By this I mean that an EEPROM is not written accidentally, but its contents can be influenced by external influences or unfavourable programming.

Regards

/Peter

[S.Ma]

Assuming 3 msec per write, about 3 hours to reach 4M cycles. So when debugging Eeprom api, avoid keeping running the test write function in the while loop when releasing the code after a debug session....