Amazon.com, Inc.
Runtime identification of bad memory cells based on difference between data pattern and read data

Abstract:

Technologies are provided for runtime identification of bad memory cells. An uncorrectable error can be detected in data stored in a plurality of memory cells of a memory device. Patterned data can be written to the plurality of memory cells that stored the data in which the uncorrectable error was detected. The data stored in the plurality of memory cells can be read and compared to the patterned data. One or more of the memory cells can be identified as bad memory cells based on differences between the patterned data and the data read from the plurality of memory cells. In at least some embodiments, the one or more identified bad memory cells can be omitted from subsequent data storage operations. Additionally or alternatively, the one or more identified bad memory cells can be repaired, for example, by using a post-package repair operation.