A 28.8-GB/s 96-MB 3D-stacked SRAM is presented. A total of eight SRAM dies, designed in a 40-nm CMOS process, are vertically stacked and connected using an inductive coupling wireless link with a low-voltage NMOS push-pull transmitter that reduces the power of the link by 35% with a 0.4-V power supply. The SRAM utilizes an inverted bit insertion scheme that compensates for the degradation of the first transmitted bit, a coil termination scheme that aims to eliminate the ringing of 3D inductive coupling bus, and a 12:1 SerDes that minimizes power consumption and area overhead in inductive coupling channels. Low-power, large-capacity, 3-cycle latency 3D-stacked SRAM for a DNN accelerator is achieved with the combination of these techniques to serve as a replacement of 3D-stacked DRAM. The performance of the proposed 3D-SRAM is compared with HBM DRAM and achieves more than 50% lower energy consumption. The scaling scenario of the SRAM module is discussed in light of the scaling of the inductive coupling technology and logic process.
8 September 2020 in IEEE Solid-State Circuits Letters
An area-efficient low-power wireless inductive coupling interface for a 3D-stacked high-bandwidth memory is presented. A test chip was fabricated in a 40-nm CMOS technology and 1.5-Gbps 2.2-pJ/bit communication using a 0.23-V NMOS push-pull transmitter and 7-μm multi-layer coils was experimentally confirmed. The low-voltage transmitter reduces power dissipation by 55% compared to the conventional one in measurement. The proposed 7-μm multi-layer coils improve their mutual inductance by a factor of 4.5 compared to that of conventional single-layer ones. This paper discusses the multi-layer coil wiring topology based on simulation and measurement. Optimization of the link budget achieves 7.1% power reduction compared to the conventional structure. Using this interface, a memory module consisting of eight 8-μm-thick memory dies and a base die in a 40-nm CMOS process and communicating at 3.6 Gbps is extrapolated to achieve a 2.3-Tbps/mm2 area efficiency and 6.1-pJ/bit energy efficiency. We also estimate that a 41-Tbps/mm2 0.1-pJ/bit memory interface is achievable in a 7-nm CMOS process.