An optimal on-demand scrubbing solution for read disturbance errors in phase-change memory

Phase-change memory is a promising technology due to its attractive properties. However, phase-change memory is difficult to commercialize because of its reliability issues. A read disturbance error, which is the main cause of reliability issues, occurs when a cell is repeatedly read. A conventional solution for read disturbance errors is periodically scrubbing the cells. However, this method requires read counters to count the number of reads per word. This paper proposes an on-demand scrubbing solution that does not require read counters, which significantly reduces resource overhead. The proposed method observes the number of errors in a word using error-correcting code. If the number of errors is larger than a pre-defined threshold, scrubbing is performed to fix the errors. The proposed method removes nearly 1GB of hardware overhead required by read counters, and fixes more than 99.99% of read disturbance errors.