Circuit and Architecture Co-Design of STT-RAM for High Performance and Low Energy

Spin-Transfer Torque Random Access Memory (STT-RAM) has been proved a promising emerging nonvolatile memory technology suitable for many applications such as cache mem- ory of CPU. Compared with other conventional memory technology, STT-RAM offers many attractive features such as nonvolatility, fast random access speed and extreme low leakage power.\ud However, STT-RAM is still facing many challenges. First of all, programming STT-RAM is a stochastic process due to random thermal fluctuations, so the write errors are hard to avoid. Secondly, the existing STT-RAM cell designs can be used for only single-port accesses, which limits the memory access bandwidth and constraints the system performance. Finally, while other memory technology supports multi-level cell (MLC) design to boost the storage density, adopting MLC to STT-RAM brings many disadvantages such as requirement for large transistor and low access speed. In this work, we proposed solutions on both circuit and architecture level to address these challenges.\ud For the write error issues, we proposed two probabilistic methods, namely write-verify- rewrite with adaptive period (WRAP) and verify-one-while-writing (VOW), for performance improvement and write failure reduction.\ud For dual-port solution, we propose the design methods to support dual-port accesses for STT-RAM. The area increment by introducing an additional port is reduced by leveraging the shared source-line structure. Detailed analysis on the performance/reliability degrada- tion caused by dual-port accesses is performed, and the corresponding design optimization is provided.\ud To unleash the potential of MLC STT-RAM cache, we proposed a new design through a cross-layer co-optimization. The memory cell structure integrated the reversed stacking of magnetic junction tunneling (MTJ) for a more balanced device and design trade-off. In architecture development, we presented an adaptive mode switching mechanism: based on application’s memory access behavior, the MLC STT-RAM cache can dynamically change between low latency SLC mode and high capacity MLC mode.\ud Finally, we present a 4Kb test chip design which can support different types and sizes of MTJs. A configurable sensing solution is used in the test chip so that it can support wide range of MTJ resistance. Such test chip design can help to evaluate various type of MTJs in the future