Networks-on-chip (NoC) have emerged to tackle different on-chip communication challenges and can satisfy different demands in terms of performance, cost and reliability. Currently, interconnects based on metal are reaching performance limits given relentless technology scaling. In particular, a performance bottleneck has emerged due to the demands for communication in terms of bandwidth for multicasting and broadcasting. As a result, various state-of-the-art architectures have been proposed as alternatives and emerging interconnects including the use of optics or radio frequency (RF). This article presents a comprehensive survey of these various interconnect fabrics, and discusses their current and future potentials and obstacles as well. This article aims to drive the research community to achieve a better utilization of the merits of on-chip interconnects and addresses the challenges involved. New interconnect technologies, such as optical interconnect, wireless NoC (WiNoC), RF transmission lines (RF-I) and surface wave interconnects (SWI), are discussed, evaluated and compared. Consequently, these emerging interconnects can continue to provide the cost efficiency and performance that are highly demanded for future many-core processors and high performance computing.

As scaling becomes increasingly difficult, 3D integration has emerged as a viable alternative to achieve the requisite bandwidth and power efficiency challenges. However mechanical stress induced by the through silicon vias (TSV) is one of the key constraints in the 3D flow that must be controlled in order to preserve the integrity of front end devices. For the first time an extended and comprehensive study is given for the stress induced by single- and arrayed TSVs and its impact on both analog and digital FEOL devices and circuits. This work provides a complete experimental assessment and quantifies the stress distribution and its effect on front end devices. By using a combined experimental and theoretical approach we provide a framework that will enable stress aware design and the right definition of keep out zone and ultimately save valuable silicon area.

Energy efficiency continues to be the core design challenge for artificial intelligence (AI) hardware designers. In this paper, we propose a new AI hardware architecture targeting Internet of Things applications. The architecture is founded on the principle of learning automata, defined using propositional logic. The logic-based underpinning enables low-energy footprints as well as high learning accuracy during training and inference, which are crucial requirements for efficient AI with long operating life. We present the first insights into this new architecture in the form of a custom-designed integrated circuit for pervasive applications. Fundamental to this circuit is systematic encoding of binarized input data fed into maximally parallel logic blocks. The allocation of these blocks is optimized through a design exploration and automation flow using field programmable gate array-based fast prototypes and software simulations. The design flow allows for an expedited hyperparameter search for meeting the conflicting requirements of energy frugality and high accuracy. Extensive validations on the hardware implementation of the new architecture using single- and multi-class machine learning datasets show potential for significantly lower energy than the existing AI hardware architectures. In addition, we demonstrate test accuracy and robustness matching the software implementation, outperforming other state-of-the-art machine learning algorithms. This article is part of the theme issue ‘Advanced electromagnetic non-destructive evaluation and smart monitoring’.

Accurate measurement of edge time differences down to 10 ps or less is required for tests of timing in digital systems. A circuit is described that is aimed at reliably amplifying these time differences by a factor between 3 and 10 before measurement to enable greater accuracy.

Inference at-the-edge using embedded machine learning models is associated with challenging trade-offs between resource metrics, such as energy and memory footprint, and the performance metrics, such as computation time and accuracy. In this work, we go beyond the conventional Neural Network based approaches to explore Tsetlin Machine (TM), an emerging machine learning algorithm, that uses learning automata to create propositional logic for classification. We use algorithm-hardware co-design to propose a novel methodology for training and inference of TM. The methodology, called REDRESS, comprises independent TM training and inference techniques to reduce the memory footprint of the resulting automata to target low and ultra-low power applications. The array of Tsetlin Automata (TA) holds learned information in the binary form as bits: $\lbrace 0,1\rbrace$ , called excludes and includes, respectively. REDRESS proposes a lossless TA compression method, called the include-encoding, that stores only the information associated with includes to achieve over 99% compression. This is enabled by a novel computationally minimal training procedure, called the Tsetlin Automata Re-profiling, to improve the accuracy and increase the sparsity of TA to reduce the number of includes, hence, the memory footprint. Finally, REDRESS includes an inherently bit-parallel inference algorithm that operates on the optimally trained TA in the compressed domain, that does not require decompression during runtime, to obtain high speedups when compared with the state-of-the-art Binary Neural Network (BNN) models. In this work, we demonstrate that using REDRESS approach, TM outperforms BNN models on all design metrics for five benchmark datasets viz. MNIST, CIFAR2, KWS6, Fashion-MNIST and Kuzushiji-MNIST. When implemented on an STM32F746G-DISCO microcontroller, REDRESS obtained speedups and energy savings ranging 5-5700× compared with different BNN models.