Text-to-motion models are AI systems that generate human motion sequences directly from natural language descriptions, serving as key enablers for immersive virtual avatars and interactive digital humans in AR/VR ecosystems. However, state-of-the-art text-to-motion diffusion models suffer from substantial computational costs due to their iterative nature, making them ill-suited for …
The rapid growth of machine learning (ML) workloads, particularly in computer vision applications, has significantly increased computational and energy demands in modern electronic systems, motivating the use of hardware accelerators to offload processing from general-purpose processors. Despite advances in computationally efficient ML models, achieving energy-efficient inference on resource-constrained edge devices …
This letter presents a 2-transistor (2T) bipolar embedded resistive RAM (eRRAM) MACRO fabricated in a 28-nm high-k metal gate (HKMG) process for multitime programmable (MTP) applications. To overcome the scaling bottlenecks of traditional embedded Flash, this work utilizes an extra-mask-free, pure front-end-of-line (FEOL) integration, offering a robust solution for automotive …
This article presents MITTA, the first silicon-proven transformer accelerator optimized for multi-task inference across both natural language processing (NLP) and image processing domains. MITTA accelerates a task-sharing algorithm that minimizes sub-task computation by reusing both activations and weights from a shared base task, requiring only sparse delta computation for sub-tasks. …
True random number generators (TRNGs) are critical for hardware security, providing unpredictable entropy for cryptographic applications. Spin-transfer torque magnetic tunnel junction (STT-MTJ) devices offer a promising entropy source due to their low-power consumption, nonvolatility, and stochastic switching behavior. This work presents an MTJ-based TRNG that produces three independent bit streams. …
This article presents a 2–18 GHz high-efficiency CMOS nonuniform distributed power amplifier (NDPA) with a novel reconfigurable inductive termination technique for ultra-broadband efficiency enhancement. First, the inherent drawback of the degrading efficiency with growing frequency in a conventional non-reconfigurable NDPA architecture with multi-octave bandwidth is studied. A simple and effective reconfigurable …
A hybrid ring oscillator (RO)/ arbiter physical unclonable function (PUF) is implemented in a 28-nm CMOS, where two competing ROs accumulate a sufficiently large phase difference exceeding a predefined deadzone (DZ). The resulting eye-opening arbiter (EOA) architecture enables a prediction of PUF bit stability over temperature change (from −40 °C to $125~^{\…
This article presents an injection-locked clock multiplier (ILCM) achieving the low-reference spur (spur ${}_{\mathrm {REF}}$ ) with minimal overhead of a calibrator. To remove the dominant sources of frequency error, which are frequency drift ( $f_{\mathrm {DF}}$ ), phase offset ( $\varPhi _{\mathrm {OS}}$ ), and injection-induced phase error ( $\varPhi _{\mathrm {INJ}}$ ), the ILCM …
This letter presents a novel hardware accelerator compatible with <3- $\mu $ m pitch 3-D Cu-Cu hybrid bonding interconnect (HBI) technology, particularly designed to efficiently execute multihead attention (MHA) of encoder transformer models. We present an accelerator that addresses performance losses due to low precision models by incorporating specialized hardware optimizations …
