Memory cores usually occupy a signif cant portion of the chip area of a complex system-on-chip. Thus, the yield of memory cores dominates the yield of the chip. Diagnosis and repair are two important techniques for memory yield improvement. In this paper, we propose two eff cient diagnosis algorithms for drowsy static random access memories (SRAMs). The f rst diagnosis algorithm, March D2, can be used to distinguish drowsy faults (DFs) from non-drowsy faults (NDFs). The second diagnosis algorithm, March D6, can distinguish different fault types of DFs. The March D2 and March D6 require 23N and (10log 2N+17)N Read/Write operations, respectively, for testing an N × W-bit drowsy SRAM.