“…First, many works [16,22,52,59,74,77,78,82,106,109,110] show that the average DRAM access latency can be shortened by reducing DRAM access timings for particular memory locations that can tolerate faster accesses. This can be done safely because, although DRAM standards call for constant access timings across all memory locations, the minimum viable access timings that the hardware can support actually differ across memory locations due to factors such as heterogeneity in the circuit design [17,78] and manufacturing process variation in circuit components [16,22,52,74,263].…”