Memory Macro Research Articles

Multi-Ported GC-eDRAM Bitcell with Dynamic Port Configuration and Refresh Mechanism

Embedded memories occupy an increasingly dominant part of the area and power budgets of modern systems-on-chips (SoCs). Multi-ported embedded memories, commonly used by media SoCs and graphical processing units, occupy even more area and consume higher power due to larger memory bitcells. Gain-cell eDRAM is a high-density alternative for multi-ported operation with a small silicon footprint. However, conventional gain-cell memories have limited data availability, as they require periodic refresh operations to maintain their data. In this paper, we propose a novel multi-ported gain-cell design, which provides up-to N read ports and M independent write ports (NRMW). In addition, the proposed design features a configurable mode of operation, supporting a hidden refresh mechanism for improved memory availability, as well as a novel opportunistic refresh port approach. An 8kbit memory macro was implemented using a four-transistor bitcell with four ports (2R2W) in a 28 nm FD-SOI technology, offering up-to a 3× reduction in bitcell area compared to other dual-ported SRAM memory options, while also providing 100% memory availability, as opposed to conventional dynamic memories, which are hindered by limited availability.

An Energy-Efficient Bayesian Neural Network Accelerator With CiM and a Time-Interleaved Hadamard Digital GRNG Using 22-nm FinFET

Bayesian neural networks (BNNs) have been proposed to address the problems of overfitting and overconfident decision making, common in conventional neural networks (NNs), due to their ability to model and express uncertainty in their predictions. However, BNNs require multiple inference passes to produce the necessary posterior distributions used to generate these highly desirable uncertainty estimates. As such, BNNs require not only an efficient, high-performance multiply-accumulation (MAC) operation but also an efficient Gaussian random number generator (GRNG) with high-quality statistics. In this article, an NN accelerator chip, leveraging a multi-bit analog compute-in-memory (CiM) static random-access memory (SRAM) macro, with a tightly coupled and highly efficient GRNG scheme, is presented in the Intel 22FFL process. The CiM macro achieves a peak energy efficiency of 32.2 TOP/sW, with 8-bit precision, while ensuring accurate on-chip matrix-vector multiplications (MVMs) with a computation error less than 0.5%. The variable precision GRNG achieves a peak throughput of 7.31 GSamp/s for an energy efficiency of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\sim$</tex-math> </inline-formula> 1 TSamp/J. Overall, our proposed system achieves a peak energy efficiency of 1170 GOP/s/W, a 35–133 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\times$</tex-math> </inline-formula> improvement over the state-of-the-art BNN accelerators, with 98.14% accuracy for the MNIST dataset.

3D SRAM Macro Design in 3D Nanofabric Process Technology

In this paper, we introduce a novel design of a 3D static random-access memory (SRAM) macro in a 3D Nanofabric process technology. The 3D Nanofabric technology is based on enabling the processing of N stack of identical layers simultaneously regardless of the number of stacked layers which consequently reduces the fabrication cost as well as the footprint of SRAM macros. To enable simultaneous patterning of stacked layers, 3D Nanofabric requires circuit topology and layout that rely on a single layer where the device channel, poly, and metal wires are all embedded without any other crossing than the gate on top of the device channel. Accordingly, we modify the layouts of the conventional SRAM bit-cell and periphery circuits which are complex and contain several metal crossings. Furthermore, we propose a new overall organization of the 3D SRAM macro that incorporates a stack of multiple identical layers each consisting of an equal size 2D array of bit-cells and the periphery circuits. We show that the proposed 3D Nanofabric SRAM macro offers 71.2% footprint gain and 36.3% read access speed improvement compared to equal size 2D SRAM macro in 3 nm FinFET.

PIMCA: A Programmable In-Memory Computing Accelerator for Energy-Efficient DNN Inference

This article presents a programmable in-memory computing accelerator (PIMCA) for low-precision (1–2 b) deep neural network (DNN) inference. The custom 10T1C bitcell in the in-memory computing (IMC) macro has four additional transistors and one capacitor to perform capacitive-coupling-based multiply and accumulation (MAC) in analog-mixed-signal (AMS) domain. A macro containing <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$256\ttimes 128$</tex-math> </inline-formula> bitcells can simultaneously activate all the rows, and as a result, it can perform a matrix-vector multiplication (VMM) in one cycle. PIMCA integrates 108 of such IMC static random-access memory (SRAM) macros with the custom six-stage pipeline and the custom instruction set architecture (ISA) for instruction-level programmability. The results of IMC macros are fed to a single-instruction-multiple-data (SIMD) processor for other computations such as partial sum accumulation, max-pooling, activation functions, etc. To effectively use the IMC and SIMD datapath, we customize the ISA especially by adding hardware loop support, which reduces the program size by up to 73%. The accelerator is prototyped in a 28-nm technology, and integrates a total of 3.4-Mb IMC SRAM and 1.5-Mb off-the-shelf activation SRAM, demonstrating one of the largest IMC accelerators to date. It achieves the system-level energy efficiency of 437 TOPS/W and the peak throughput of 49 TOPS at the 42-MHz clock frequency and 1-V supply for the VGG9 and the ResNet-18 on the CIFAR-10 dataset.

BP-SCIM: A Reconfigurable 8T SRAM Macro for Bit-Parallel Searching and Computing In-Memory

This work presents BP-SCIM: a reconfigurable 8T static random access memory (SRAM) macro for bit-parallel searching and computing in-memory (CIM). The decoupled read/write ports of the employed 8T SRAM bit-cell eliminate read disturbance during search and CIM operations. BP-SCIM can support both in-memory Boolean logic and arithmetic operations. Novel CIM-friendly algorithms and peripheral circuits are proposed to reduce the latency of complex arithmetic operations such as multiplication and division. In addition, BP-SCIM can be configured as either a binary content-addressable memory (CAM) or a ternary CAM for fast searching. A 256 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\times$</tex-math> </inline-formula> 64 BP-SCIM test chip was implemented in 65-nm CMOS technology. The 8-bit addition and 8-bit multiplication operations can achieve the maximum energy efficiency of 3.11 TOPS/W and 0.17 TOPS/W, respectively at 0.7 V supply. For the binary CAM search operation, BP-SCIM can achieve the minimum energy consumption of 0.91 fJ/bit/search at 87 MHz and 0.8 V supply.

A 2941-TOPS/W Charge-Domain 10T SRAM Compute-in-Memory for Ternary Neural Network

In this paper, we present a 10T SRAM compute-in memory (CiM) macro to process the multiplication-accumulation (MAC) operations between ternary-inputs and binary-weights. In the proposed 10T SRAM bitcell, the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">charge-domain</i> analog computations are employed to improve the noise tolerance of bit-line (BL) signals where the MAC results are represented in CiM. Parallel processing of 3 different analog levels for ternary input activations is also performed in the proposed single 10T bitcell. To reduce the analog-to-digital converter (ADC) bit-resolutions without sacrificing deep neural network (DNN) accuracies, a confined-slope non-uniform integration (CS-NUI) ADC is proposed, which can provide layer-wise adaptive quantization for multiple different layers with different MAC distributions. In addition, by sharing the ADC reference voltage generator in every single column of SRAM array, the ADC area is effectively reduced with improved energy efficiencies of CiM. The <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$256\times64.10$</tex-math> </inline-formula> T SRAM CiM macro with the proposed charge-sharing scheme and CS-NUI ADCs has been implemented using 28nm CMOS process. The silicon measurement results show that the proposed CiM shows the accuracies of 98.66% and 88.48% with MNIST dataset on MLP, and CIFAR-10 dataset on VGGNet-7, respectively, with the energy efficiency of 2941-TOPS/W and the area efficiency of 59.584-TOPS/mm <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$^{2}$</tex-math> </inline-formula> .

Configurable and High-Throughput CIM SRAM for Boolean Logic Operation With 321 GOPS/kb and 164395.6 GOPS/mm2

In this letter, we propose a twin-operand computing method based on transforming a static random-access memory (SRAM) macro that can be used for image processing. The proposed cross-coupled 10T cell structure can be configured into a dual 4T structure, and the computational results can be stored in situ, thus significantly improving energy efficiency. Furthermore, unlike previous designs, the proposed computationin-memory (CIM) macro allows independent operands, where two operands are stored in dual 4T. Not only can it provide computational functions, but it can also double the storage density. The measurement results of a 55 nm CMOS prototype chip show that it achieves the highest throughput of 321 GOPS/Kb, 164395.6 GOPS/mm2, and 364.2 TOPS/W at 1.2 V.

Pseudo-Static Gain Cell of Embedded DRAM for Processing-in-Memory in Intelligent IoT Sensor Nodes.

This paper presents a pseudo-static gain cell (PS-GC) with extended retention time for an embedded dynamic random-access memory (eDRAM) macro for analog processing-in-memory (PIM). The proposed eDRAM cell consists of a two-transistor (2T) gain cell with a pseudo-static leakage compensation that maintains stored data without charge loss issue. Hence, the PS-GC can offer unlimited retention time in the same manner as static RAM (SRAM). Due to the extended retention time, bulky capacitors in conventional eDRAM are no longer needed, thereby, improving the area efficiency of eDRAM-based analog PIMs. The active leakage compensation of the PS-GC can effectively hold stored data even in a deep-submicron process that show significant leakage current. Therefore, the PS-GC can accelerate write-access time and read-access time without concern of increased leakage current. The proposed gain cell and its 64 × 64 eDRAM macro were implemented in a 28 nm CMOS process. The bitcell of the proposed gain cell has 0.79- and 0.58-times the area of those of 6T SRAM and 8T STAM, respectively. The post-layout simulation results demonstrate that the eDRAM maintains the pseudo-static operation with unlimited retention time successfully under wide range variations of process, voltage and temperature. At the operating frequency of 667 MHz, the eDRAM macro achieved an operating voltage range from 0.9 to 1.2 V and operating temperature range from −25 to 85 °C regardless of the process variation. The post-layout simulated write-access time and read-access time were below 0.3 ns at an operating temperature of 85 °C. The PS-GC consumes a static power of 2.2 nW/bit at an operating temperature of 25 °C.

A 22-nm 1-Mb 1024-b Read Data-Protected STT-MRAM Macro With Near-Memory Shift-and-Rotate Functionality and 42.6-GB/s Read Bandwidth for Security-Aware Mobile Device

The development of security-aware mobile devices using wide-input–output (IO) nonvolatile memory (NVM) is hindered by high peak current, large area overhead for high read bandwidth (BWR), and considerable energy consumption for data movement between NVM and logic blocks. Furthermore, data stored in NVM are vulnerable to reverse-engineering attacks. This work presents a high BWR security-aware near-memory-computing spin-transfer torque magnetic random-access memory (STT-MRAM) macro using a multi-bit current-mode sense amplifier (MB-CSA) to reduce peak current and energy consumption for wide-IO access, a near-memory shift-and-rotate functionality (NSRF) in conjunction with the MB-CSA to reduce area overhead and enable the completion of read and logic operations within a single cycle, and a reverse-engineering-proof XOR–based memory data protector to protect data stored in NVM against reverse-engineering attacks. A 1-Mb 1024-b read STT-MRAM macro with data protector fabricated using foundry embedded 22-nm STT-MRAM. This work achieved 42.67 GB/s for BWR and 0.23 pJ/b. Inclusion of the NSRF circuit reduced area overhead by 33.3% while increasing latency by only 170 ps.

A 36.2 dB High SNR and PVT/Leakage-Robust eDRAM Computing-In-Memory Macro With Segmented BL and Reference Cell Array

Computing-in-memory (CIM) shows high energy-efficiency through the analog DNN computation inside the memory macros. However, as the DNN size increases, the energy-efficiency of CIM is reduced by external memory access (EMA). One of the promising solutions is eDRAM based CIM to increase memory capacity with a high density cell. Although the eDRAM-CIM has a higher density than the SRAM-CIM, it suffers from both poor robustness and a low signal-to-noise ratio (SNR). In this brief, the energy-efficient eDRAM-CIM macro is proposed while improving computational robustness and SNR with three key features: 1) High SNR voltage-based accumulation with segmented BL architecture (SBLA), resulting in 17.1 dB higher SNR, 2) canceling PVT/leakage-induced error with common-mode error canceling (CMEC) circuit, resulting in 51.4&#x0025; PVT variation reduction and 51.4&#x0025; refresh power reduction, 3) a ReLU-based zero-gating ADC (ZG-ADC), resulting in ADC power reduction up to 58.1&#x0025;. According to these new features, the proposed eDRAM-CIM macro achieves 81.5-to-115.0 TOPS/W energy-efficiency with 209-to-295 <inline-formula> <tex-math notation="LaTeX">$\mu \text{W}$ </tex-math></inline-formula> power consumption when 4b <inline-formula> <tex-math notation="LaTeX">$\times $ </tex-math></inline-formula> 4b MAC operation is performed with 250 MHz core frequency. The proposed macro also achieves 91.52&#x0025; accuracy at the CIFAR-10 object classification dataset (ResNet-20) without accuracy drop even with PVT variation.

Estimation of minimum operating voltage in fully depleted SOI SRAM cells using gamma distribution

The minimum operating voltage (V min) of every cell on a 32 kb fully depleted SOI static random-access memory (SRAM) macro is successfully measured. Competing V min distribution models, which include the gamma and log-normal distributions, are approximated using the generalized gamma distribution. It is found that the V min of the cells follows the gamma distribution. This finding gives a simple method to estimate the worst V min of an SRAM macro by measuring a few samples and performing linear extrapolation from the gamma distribution.

In-Memory Computing on Resistive RAM Systems Using Majority Operation

Memristors can be used to build nonvolatile memory systems with in-memory computing (IMC) capabilities. A number of prior works demonstrate the design of an IMC-capable memory macro using a memristor crossbar. However, read disturbance limits the use of such memory systems built using a 0-transistor, 1-RRAM (0T1R) structure that suffers from the sneak path problem. In this paper, we introduce a scheme for both memory and logic operations using the 1-transistor, 1-RRAM (1T1R) memristor crossbar, which effectively mitigates the read disturbance problem. The memory array is designed using nMOS transistors and the VTEAM memristor model. The peripheral circuitry like decoders, voltage multiplexers, and sense amplifiers is designed using a 45[Formula: see text]nm CMOS technology node. We introduce a mapping technique to realize arbitrary logic functions using Majority (MAJ) gate operations in the 1T1R crossbar. Through extensive experimentation on benchmark functions, it has been found that the proposed mapping method gives an improvement of 65% or more in terms of the number of time steps required, and 59% or more in terms of energy consumption as compared to some of the recent methods.

SecNVM: Power Side-Channel Elimination Using On-Chip Capacitors for Highly Secure Emerging NVM

Emerging nonvolatile memories (NVMs), such as resistive RAM (RRAM) and spin-transfer-torque RAM (STTRAM), present exciting opportunities for data storage applications and offer improved access speeds, retention times, power consumption, and scalability. However, these technologies leak the Hamming weight of data through power side-channel during read and write operations. We propose a technique leveraging on-chip capacitor and voltage regulator (VR) that powers the NVM read/write operations. The side-channel leakage is eliminated due to the isolation of memory array from the external power supply during read/write operations. The residual charge on capacitor bank is discarded safely to prevent information leakage during capacitor recharging. The VR ensures a steady voltage during the entire read/write operations even though the capacitor discharges. The design presents a performance (instructions per cycle) degradation of 0.53%-1.2% under parsec and splash-2 benchmarks and incurs an area overhead of ~ 3.54×10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-5</sup> % and an energy overhead of ~ 3.05 ×10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-5</sup> % for a 4-Mb RRAM memory array. For a 64-bit word, the design improves security by 2.7 × 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">19</sup> × to 2 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">64</sup> ×. SecNVM should be used in small security-critical memory macros to limit the overhead. SecNVM is generic and could protect any security module such as encryption engines, against power side-channel attacks.

CAP-RAM: A Charge-Domain In-Memory Computing 6T-SRAM for Accurate and Precision-Programmable CNN Inference

A compact, accurate, and bitwidth-programmable in-memory computing (IMC) static random-access memory (SRAM) macro, named CAP-RAM, is presented for energy-efficient convolutional neural network (CNN) inference. It leverages a novel charge-domain multiply-and-accumulate (MAC) mechanism and circuitry to achieve superior linearity under process variations compared to conventional IMC designs. The adopted semi-parallel architecture efficiently stores filters from multiple CNN layers by sharing eight standard 6T SRAM cells with one charge-domain MAC circuit. Moreover, up to six levels of bit-width of weights with two encoding schemes and eight levels of input activations are supported. A 7-bit charge-injection SAR (ciSAR) analog-to-digital converter (ADC) getting rid of sample and hold (S&H) and input/reference buffers further improves the overall energy efficiency and throughput. A 65-nm prototype validates the excellent linearity and computing accuracy of CAP-RAM. A single 512x128 macro stores a complete pruned and quantized CNN model to achieve 98.8% inference accuracy on the MNIST data set and 89.0% on the CIFAR-10 data set, with a 573.4-giga operations per second (GOPS) peak throughput and a 49.4-tera operations per second (TOPS)/W energy efficiency.

Low-Cost Side-Channel Secure Standard 6T-SRAM-Based Memory With a 1% Area and Less Than 5% Latency and Power Overheads

Side-channel attacks constitute a concrete threat to IoT systems-on-a-chip (SoCs). Embedded memories implemented with 6T SRAM macrocells often dominate the area and power consumption of these SoCs. Regardless of the computational platform, the side-channel sensitivity of low-hierarchy cache memories can incur significant overhead to protect the memory content (i.e., data encryption, data masking, etc.). In this manuscript, we provide a silicon proof of the effectiveness of a low cost side-channel attack protection that is embedded within the memory macro to achieve a significant reduction in information leakage. The proposed solution incorporates low-cost impedance randomization units, which are integrated into the periphery of a conventional 6T SRAM macro in fine-grain memory partitions, providing possible protection against electromagnetic adversaries. Various blocks of unprotected and protected SRAM macros were designed and fabricated in a 55 nm test-chip. The protected ones had little as 1% area overhead and less than 5% performance and power penalties compared to a conventional SRAM design. To evaluate the security of the proposed solution, we applied a robust mutual information metric and an adaptation to the memory context to enhance this evaluation framework. Assessment of the protected memory demonstrated a significant information leakage reduction from 8 bits of information exposed after only 100 cycles of attack to less than ~1.5 bits of mutual information after 160K traces. The parametric nature of the protection mechanisms are discussed while specifying the proposed design parameters. Overall, the proposed methodology enables designs with higher security-level at a minimal cost.

A 0.506-pJ 16-kb 8T SRAM With Vertical Read Wordlines and Selective Dual Split Power Lines

This article presents an 8T static random access memory (SRAM) macro with vertical read wordline (RWL) and selective dual split power (SDSP) lines techniques. The proposed vertical RWL reduces dynamic energy consumption during read operation by charging and discharging only selected read bitlines (RBLs). The data-aware SDSP technique combined with vertical write bitlines enhances both the write margin (WM) and the static noise margin (SNM). A 16-kb SRAM test chip fabricated in 65-nm CMOS technology demonstrates the minimum energy consumption of 0.506 pJ at 0.4 V and the minimum operating voltage of 0.26 V.

Embedded 1-Mb ReRAM-Based Computing-in- Memory Macro With Multibit Input and Weight for CNN-Based AI Edge Processors

Computing-in-memory (CIM) based on embedded nonvolatile memory is a promising candidate for energy-efficient multiply-and-accumulate (MAC) operations in artificial intelligence (AI) edge devices. However, circuit design for NVM-based CIM (nvCIM) imposes a number of challenges, including an area-latency-energy tradeoff for multibit MAC operations, pattern-dependent degradation in signal margin, and small read margin. To overcome these challenges, this article proposes the following: 1) a serial-input non-weighted product (SINWP) structure; 2) a down-scaling weighted current translator (DSWCT) and positive–negative current-subtractor (PN-ISUB); 3) a current-aware bitline clamper (CABLC) scheme; and 4) a triple-margin small-offset current-mode sense amplifier (TMCSA). A 55-nm 1-Mb ReRAM-CIM macro was fabricated to demonstrate the MAC operation of 2-b-input, 3-b-weight with 4-b-out. This nvCIM macro achieved $T_{\text {MAC}}= 14.6$ ns at 4-b-out with peak energy efficiency of 53.17 TOPS/W.

An Energy-Efficient Normally Off Microcontroller With 880-nW Standby Power, 1 Clock System Backup, and 4.69-$\mu$ s Wakeup Featuring 60-nm CAAC-IGZO FETs

This letter presents a microcontroller unit (MCU) that employs a $c$ -axis aligned crystalline indium–gallium–zinc oxide semiconductor FET (CAAC-IGZO FET) with an extremely low off-state current below the zepto-ampere (10−21 A) level. This MCU based on IGZO-FETs has logic circuits and memory macros that can retain data even when the power supply is turned off, and internal voltage regulators that store the reference voltage. A prototype chip has been fabricated with a monolithic process that combines an IGZO process (embedded in BEOL) and an Si CMOS process; the chip has a standby power of 880 nW, a system backup time of 21 ns with 44 fJ/bit, and a wakeup time of $4.69~\mu \text{s}$ . The fabricated chip has demonstrated that the highly energy-efficient normally off computing that establishes low-power operations in Internet of Things and edge AI computing can be achieved with IGZO technology.

Slicing FIFOs for On-Chip Memory Bandwidth Exhaustion

Within an MPSoC environment, the delivery of data from one computation stage to the next is often the bottleneck of the overall system performance. Especially in high-speed communication where at least two actors (producer and consumer) need access to the same data, the memory interface is usually the limiting factor. Therefore, many solutions like network on chips, clusters of shared memory, memory hierarchy, and so on have been proposed to transfer data from one actor to another. All solutions face the common problem that, at the lowest layer, multiple actors need simultaneous access to a hardware memory block. To avoid using dual port memory macros—which are expensive in terms of chip area—a hardware design is proposed where the communication through First-In-First-Out-Buffers (FIFOs) is transparently sliced to a set of altering memory banks. This allows simultaneous memory accesses to the same FIFO while using only single port memory macros, thus reducing on-chip area and access delay. The proposed FIFO controller balances the load of multiple FIFOs evenly over independent memory regions to maximize system performance.

A 28-nm 1R1W Two-Port 8T SRAM Macro With Screening Circuitry Against Read Disturbance and Wordline Coupling Noise Failures

We demonstrate a 1-read/1-write two-port (2P) embedded static random access memory macro based on 8T SRAM bitcell with an effective scheme for design of testability. To achieve a smaller macro area, a differential sense amplifier is introduced to read out the data, where the reference voltage for reading 0/1 data is generated by an unselected bitcell array. In addition, we propose a screening test circuit for read disturbance and wordline coupling noise. A 512-kbit 2P SRAM macro using 28-nm high-K/metal gate bulk CMOS technology has been designed, confirming experimentally that the worst minimum operation voltage ( $V_{\text {min}}$ ) can be reproduced by our test circuit. Bit density of 3.16 Mb/mm2 was achieved.