To continue the advances in nanoelectronics in scaling while fulfilling the demand of high computing such as Artificial intelligence (AI), Machine Learning (ML), internet-of-things (IoT) and hybrid systems; reducing the power consumption and boosting the performance has become critical in the semiconductor community. The traditional Moore’s law scaling technology, traditionally used in materials, devices, and systems, may no longer guarantee the computational demand of the conventional von-Neumann architecture. It has become crucial for the electronics and systems era to subvert the bottleneck in current CPU architecture i.e. von-Neumann architecture. To tackle the new computing paradigms which subvert the memory wall in von Neumann bottleneck, the new computing configurations e.g. neuromorphic computing, edge computing, in-memory computing are attracting a considerable amount of attention. In this era, non-volatile memory technology using emerging new materials and device physics for implementation of neuromorphic systems and data-centric computing are promising ways to achieve the goals and requirements of next generation energy-efficient high-performance computing applications.
Emerging memory can be categorized by switching mechanism in materials and device physics, such as ferroelectric random access memory (FeRAM), resistive random access memory (RRAM), phase-change memory (PCM) and Magnetic Random Access Memory (MRAM), which show great promise for next generation storage and computational applications. Meanwhile, dynamic switch devices are considered the candidate for the new computational applications, such as ferroelectric field-effectransistor (FeFET), tunneling FET (TFET), negative capacitance FET (NCFET), selectors etc; especially in the crossbar memory array applications.
In addition to renovation on devices, it is essential to enable the superior potential of computationally heavy ML algorithms, such as deep neural networks, developing specialized computing hardware, circuit, system, and application processing units (APUs). The introduction of emerging memory technologies opens the opportunity to revisit the conventional computing concepts and move towards unconventional computing paradigms e.g. near-memory/in-memory computing. In-memory computing has been explored through various technological solutions, from mainstream static random access memory (SRAM) and dynamic random access memory (DRAM) to emerging solutions such as embedded dynamic random access memory (eDRAM) and NAND/NOR FLASH memory. Beyond these digital technologies, in-memory computing based on emerging memory technologies has shown shown to be promising for next generation energy-efficient hardware by offering ultra-scalable low-power devices that can provide highly parallel vector matrix multiplications (VMMs).
The aim of this Research Topic is to collate interdisciplinary research on hardware/software co-solutions on the basis of materials, devices, circuits, and systems to develop efficient components, based on next-generation memory technologies. The topics of interest in this collection include but are not limited to:
- Materials enable energy-efficient switching
- Materials enable the electrical signal interfacing between biological materials and CMOS
- Devices to perform memory characteristics efficiently
- CMOS-compatible materials and devices for memory and computing technology
- Reliability solutions toward emerging memory
- Multilevel state applications and analog behaviors in memory with distinguishable distribution
- Emerging memory simulation models (including devices, circuits and systems)
- Novel hardware/software solutions for efficient inference and training on memory crossbar arrays
- Conceptual designs for efficient implementation of ML applications
- Input and output peripheral circuits for RS-based VMM accelerators (digital-to-analog convertors (DACs), analog-to-digital convertors (ADCs), trans-impedance amplifiers (TIAs), etc.
Topic Editor Yao-Feng Chang is employed by company Intel Corp. The rest of the Topic Editors declare no conflicts of interest.
To continue the advances in nanoelectronics in scaling while fulfilling the demand of high computing such as Artificial intelligence (AI), Machine Learning (ML), internet-of-things (IoT) and hybrid systems; reducing the power consumption and boosting the performance has become critical in the semiconductor community. The traditional Moore’s law scaling technology, traditionally used in materials, devices, and systems, may no longer guarantee the computational demand of the conventional von-Neumann architecture. It has become crucial for the electronics and systems era to subvert the bottleneck in current CPU architecture i.e. von-Neumann architecture. To tackle the new computing paradigms which subvert the memory wall in von Neumann bottleneck, the new computing configurations e.g. neuromorphic computing, edge computing, in-memory computing are attracting a considerable amount of attention. In this era, non-volatile memory technology using emerging new materials and device physics for implementation of neuromorphic systems and data-centric computing are promising ways to achieve the goals and requirements of next generation energy-efficient high-performance computing applications.
Emerging memory can be categorized by switching mechanism in materials and device physics, such as ferroelectric random access memory (FeRAM), resistive random access memory (RRAM), phase-change memory (PCM) and Magnetic Random Access Memory (MRAM), which show great promise for next generation storage and computational applications. Meanwhile, dynamic switch devices are considered the candidate for the new computational applications, such as ferroelectric field-effectransistor (FeFET), tunneling FET (TFET), negative capacitance FET (NCFET), selectors etc; especially in the crossbar memory array applications.
In addition to renovation on devices, it is essential to enable the superior potential of computationally heavy ML algorithms, such as deep neural networks, developing specialized computing hardware, circuit, system, and application processing units (APUs). The introduction of emerging memory technologies opens the opportunity to revisit the conventional computing concepts and move towards unconventional computing paradigms e.g. near-memory/in-memory computing. In-memory computing has been explored through various technological solutions, from mainstream static random access memory (SRAM) and dynamic random access memory (DRAM) to emerging solutions such as embedded dynamic random access memory (eDRAM) and NAND/NOR FLASH memory. Beyond these digital technologies, in-memory computing based on emerging memory technologies has shown shown to be promising for next generation energy-efficient hardware by offering ultra-scalable low-power devices that can provide highly parallel vector matrix multiplications (VMMs).
The aim of this Research Topic is to collate interdisciplinary research on hardware/software co-solutions on the basis of materials, devices, circuits, and systems to develop efficient components, based on next-generation memory technologies. The topics of interest in this collection include but are not limited to:
- Materials enable energy-efficient switching
- Materials enable the electrical signal interfacing between biological materials and CMOS
- Devices to perform memory characteristics efficiently
- CMOS-compatible materials and devices for memory and computing technology
- Reliability solutions toward emerging memory
- Multilevel state applications and analog behaviors in memory with distinguishable distribution
- Emerging memory simulation models (including devices, circuits and systems)
- Novel hardware/software solutions for efficient inference and training on memory crossbar arrays
- Conceptual designs for efficient implementation of ML applications
- Input and output peripheral circuits for RS-based VMM accelerators (digital-to-analog convertors (DACs), analog-to-digital convertors (ADCs), trans-impedance amplifiers (TIAs), etc.
Topic Editor Yao-Feng Chang is employed by company Intel Corp. The rest of the Topic Editors declare no conflicts of interest.