Two-dimensional van der Waals (2D vdW) ferromagnets possess outstanding scalability, controllable ferromagnetism, and out-of-plane anisotropy, enabling the compact spintronics-based non-volatile in-memory computing (nv-IMC) that promises to tackle the memory wall bottleneck issue. Here, by employing the intriguing room-temperature ferromagnetic characteristics of emerging 2D Fe3GeTe2 with the dissimilar electronic structure of the two spin-conducting channels, we report on a new type of non-volatile spin-orbit torque (SOT) magnetic tunnel junction (MTJ) device based on Fe3GeTe2/MgO/Fe3GeTe2 heterostructure, which demonstrates the uni-polar and high-speed field-free magnetization switching by adjusting the ratio of field-like torque to damping-like torque coefficient in the free layer. Compared to the conventional 2T1M structure, the developed 3-transistor-2-MTJ (3T2M) cell is implemented with the complementary data storage feature and the enhanced sensing margin of 201.4% (from 271.7 mV to 547.2 mV) and 276% (from 188.2 mV to 520 mV) for reading “1” and “0”, respectively. Moreover, superior to the traditional CoFeB-based MTJ memory cell counterpart, the 3T2M crossbar array architecture can be executed for AND/NAND, OR/NOR Boolean logic operation with a fast latency of 24 ps and ultra-low power consumption of 2.47 fJ/bit. Such device to architecture design with elaborated micro-magnetic and circuit-level simulation results shows great potential for realizing high-performance 2D material-based compact SOT magnetic random-access memory, facilitating new applications of highly reliable and energy-efficient nv-IMC.
Probabilistic computing is an emerging computational paradigm that uses probabilistic circuits to efficiently solve optimization problems such as invertible logic, where traditional digital computations are difficult to solve. This paper proposes a true random number generator (TRNG) based on resistive random-access memory (RRAM), which is combined with an activation function implemented by a piecewise linear function to form a standard p-bit cell, one of the most important parts of a p-circuit. A p-bit multiplexing strategy is also applied to reduce the number of p-bits and improve resource utilization. To verify the superiority of the proposed probabilistic circuit, we implement the invertible p-circuit on a field-programmable gate array (FPGA), including AND gates, full adders, multi-bit adders, and multipliers. The results of the FPGA implementation show that our approach can significantly save the consumption of hardware resources.
1 October 2020 in IEEE Journal of Solid-State Circuits
By virtue of hardware parallelism, ternary content addressable memory (TCAM) is attractive for low-latency search for packet forwarding in routers for network communications in the upcoming fifth-generation (5G) era. However, the demand for high-density TCAM encounters remarkable costs in silicon area and power consumption. In this work, a 16-kb nonvolatile ternary content addressable memory (nvTCAM) test chip based on resistive memory (ReRAM) is demonstrated in 28-nm process with two techniques to deal with the aforementioned issues. First, the crossbar array with 2-diode-2-ReRAM (2D2R) nvTCAM cell is proposed to realize >3x improvement in storage density. The back-end-of-line integration of both diode and ReRAM resistor also allows for further 3-D stacking to realize larger storage density. Second, the machine learning concept is exploited to realize intelligent search operation. K-means clustering is employed to allocate the entry storage and then the search of destination IP address can be targeted to a specific bank for low power. The evaluation shows >70% reduction in search energy with 2% overhead in silicon area for bank count of four. The test chip also achieves a match delay of 2 ns under nominal operating conditions.
12 March 2021 in IEEE Transactions on Circuits and Systems II: Express Briefs
In this work, a ReRAM-based energy-efficient CIM accelerator is presented with two techniques for edge AI applications. Firstly, a circuit-algorithm co-design scheme is proposed to realize fully analog processing, which improves the energy efficiency and the throughput of neural network. To deal with the I-V nonlinearity of ReRAM, a nonlinear-aware training algorithm is proposed to improve the network accuracy. Secondly, a 1T2R cell is proposed to replace previous 2T2R for weight storage with 35% area saving. For evaluation, a neural network with two fully connected layers and one ReLU layer is built for the MNIST dataset. The error rate can be reduced by >46% and the energy efficiency is 99 TOPS/W@200 MHz, 2.6X improvement over the digital method.
26 October 2022 in IEEE Transactions on Circuits and Systems I: Regular Papers
Computing-in-memory (CIM) relieves the Von Neumann bottleneck by storing the weights of neural networks in memory arrays. However, two challenges still exist, hindering the efficient acceleration of convolutional neural networks (CNN) in artificial intelligence (AI) edge devices. Firstly, the activations for sliding window (SW) operations in CNN still bring high memory access pressure. This can be alleviated by increasing the SW parallelism, but simple array replication suffers from poor array utilization and large peripheral circuits overhead. Secondly, the partial sums from individual CIM arrays, which are usually accumulated to obtain the final sum, introduce large latency due to enormous shift-and-add operations. Moreover, high-resolution ADCs are also needed to reduce the quantization error of partial sums, further increasing the hardware costs. In this paper, a hardware-efficient CIM accelerator, ARBiS, is proposed with improved activation reusability and bit-scalable matrix-vector-multiplication (MVM) for CNN acceleration in AI edge applications. The cyclic-shift weight duplication exploits a third dimension of receptive field (RF) depth for SW weight mapping to reduce the memory accesses of activations, improving the array utilization. The parasitic-capacitance charge sharing is employed to realize high-precision analog MVM in order to reduce the ADC cost. Compared with conventional architectures, ARBiS with parallel processing of 9 SW operations achieves 56.6% $\sim $ 58.8% alleviation of memory access pressure. Meanwhile, ARBiS configured with 8-bit ADCs saves 92.53% $\sim $ 94.53% ADC energy consumption. An ARBiS accelerator is evaluated to realize a computational efficiency (CE) of 10.28 (10.43) TOPS/mm $^{2}$ , an energy efficiency (EE) of 91.19 (112.36) TOPS/W with 8-bit (4-bit) ADCs, achieving 11.4 $\sim $ 11.7 $\times$ (11.6 $\sim $ 11.8 $\times $ ), 1.1 $\sim $ 3.3 $\times$ (1.4 $\sim $ 4 $\times $ ) improvements over state-of-the-art works, respectively.
In this work, a memristive spike-based computing in memory (CIM) system with adaptive neuron (MSPAN) is proposed to realize energy-efficient remote arrhythmia detection with high accuracy in edge devices by software and hardware co-design. A multi-layer deep integrative spiking neural network (DiSNN) is first designed with an accuracy of 93.6% in 4-class ECG classification tasks. Then a memristor-based CIM architecture and the corresponding mapping method are proposed to deploy the DiSNN. By evaluation, the overall system achieves an accuracy of over 92.25% on the MIT-BIH dataset while the area is 3.438 mm2 and the power consumption is 0.178 μJ per heartbeat at a clock frequency of 500 MHz. These results reveal that the proposed MSPAN system is promising for arrhythmia detection in edge devices.