Also, the wear leveling will attempt to spread writes across the blocks so as to prevent any one block from having a lot of writes.
Unless you're doing a truly stupendous amount of writing, it's unlikely to damage the eMMC in terms of expending the write capacity. As mentioned above, you'd need to write ~160TB to the eMMC, which is only ~0.002 TB, so we're talking orders of magnitude of difference here. I forget how fast writes are to the eMMC, but I think you'd need to be writing to it, continuously, for several years to even make a dent in your write capacity. On Tuesday, February 25, 2014 9:34:50 AM UTC-5, Timbo wrote: > > > On Tuesday, February 25, 2014 3:20:37 AM UTC+1, Carl-Fredrik Sundström > wrote: >> >> Has anyone actually experienced an emmc go bad due to it getting worn out >> ? Does the controller just refuse any more writes when it reaches the 3000 >> writes or does it attempt writing anyways ? >> > > A certain fraction of the memory is kept back and used to replace bad > blocks as they are detected. From the spec sheet: > > "If a defective block is identified, JEDEC eMMC completely replaces the > defective block with one of the spare blocks. This process is invisible to > the host and generally does not affect data space allocated for the user." > > After a block replacement there is a risk that the filesystem will be left > corrupt, and a possibility that fsck will not be able to repair the data. > But re-flashing from backups should restore function. Of course the supply > of spare blocks will eventually run out, but I suspect that this will take > a long time. > > -- For more options, visit http://beagleboard.org/discuss --- You received this message because you are subscribed to the Google Groups "BeagleBoard" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/groups/opt_out.
