检索结果-内蒙古大学图书馆

Performance, efficiency, and cost analysis of wafer-scale AI accelerators vs. single-chip GPUs

学校读者我要写书评

暂无评论

Device

Device 2025年

作者： Mihrimah Ozkan Lily Pompa Md Shaihan Bin Iqbal Yiu Chan Daniel Morales Zixun Chen Handing Wang Lusha Gao Sandra Hernandez Gonzalez Department of Electrical and Computer Engineering University of California Riverside Riverside CA USA Department of Bioengineering University of California Riverside Riverside CA USA Department of Materials Science and Engineering University of California Riverside Riverside CA USA

The exponential growth of artificial intelligence (AI) models, now reaching trillions of parameters, has revealed significant limitations in traditional single-chip graphics processing unit (GPU) architectures, particularly in scalability, energy efficiency, and computational throughput. Wafer-scale computing has emerged as a transformative paradigm, integrating multiple chiplets into a single monolithic wafer to deliver unprecedented performance and efficiency. Platforms such as the Cerebras Wafer-Scale Engine (WSE-3), with 4 trillion transistors and 900,000 cores, and Tesla’s Dojo, featuring 1.25 trillion transistors and 8,850 cores per training tile, exemplify the potential of wafer-scale AI accelerators to address the demands of large-scale AI workloads. This review provides a comprehensive comparative analysis of wafer-scale AI accelerators and single-chip GPUs, focusing on their relative performance, energy efficiency, and cost effectiveness in high-performance AI applications. Emerging technologies, such as TSMC’s chip-on-wafer-on-substrate (CoWoS), which promise to enhance computational density by up to 40 times, are also examined. In addition, this work discusses critical challenges, including fault tolerance, software optimization, and economic feasibility, offering insights into the trade-offs and synergies between these two hardware paradigms. Furthermore, emerging AI hardware trends, including three-dimensional (3D) integration, photonic chips, and advanced semiconductor materials, are also discussed. This review aims to inform the development of scalable and energy-efficient AI computing by evaluating their strengths and limitations. A future outlook outlines key advancements expected over the next 5 to 10 years, shaping the next generation of AI hardware.

关键词： AI AI training wafer-scale engine WSE chip technology GPU hardware performance DTI-5: Extend & Scale

Can Personalized Medicine Coexist with Health Equity? Examining the Cost Barrier and Ethical Implications

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Francisco, Kishi Kobe Yee Apuhin, Andrane Estelle Carnicer Tan, Myles Joshua Toledo Byers, Mickael Cavanaugh Maravilla, Nicholle Mae Amor Tan Karim, Hezerul Abdul AlDahoul, Nouar Biology Program College of Arts and Science University of St. La Salle Negros Occidental Bacolod City Philippines Department of Electrical and Computer Engineering Herbert Wertheim College of Engineering University of Florida GainesvilleFL United States Department of Epidemiology College of Public Health & Health Professions College of Medicine University of Florida GainesvilleFL United States Department of Natural Sciences College of Arts and Science University of St. La Salle Negros Occidental Bacolod City Philippines Department of Chemical Engineering College of Engineering and Technology University of St. La Salle Negros Occidental Bacolod City Philippines Department of Electronics Engineering College of Engineering and Technology University of St. La Salle Negros Occidental Bacolod City Philippines Yo-Vivo Corporation Negros Occidental Bacolod City Philippines Department of Civil and Coastal Engineering Herbert Wertheim College of Engineering University of Florida GainesvilleFL United States Faculty of Engineering Multimedia University Selangor Cyberjaya Malaysia Department of Computer Science Division of Science New York University Abu Dhabi Saadiyat Marina District Abu Dhabi United Arab Emirates

Personalized medicine (PM) promises to transform healthcare by providingtreatments tailored to individual genetic, environmental, and lifestylefactors. However, its high costs and infrastructure demands raise concernsabout exacerbating health disparities, especially between high-income countries(HICs) and low- and middle-income countries (LMICs). While HICs benefit fromadvanced PM applications through AI and genomics, LMICs often lack theresources necessary to adopt these innovations, leading to a wideninghealthcare divide. This paper explores the financial and ethical challenges ofPM implementation, with a focus on ensuring equitable access. It proposesstrategies for global collaboration, infrastructure development, and ethicalframeworks to support LMICs in adopting PM, aiming to prevent furtherdisparities in healthcare accessibility and outcomes. © 2024, CC BY.

关键词： Personalized medicine

Efficient thermo-optic micro-ring phase shifter made of PECVD silicon-rich amorphous silicon carbide

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Chang, Li-Yang Sunny Pappert, Steve Yu, Paul K.L. Materials Science and Engineering Program The Department of Mechanical and Aerospace Engineering University of California San Diego 9500 Gilman Drive La Jolla CA92093 United States Department of Electrical & Computer Engineering University of California San Diego 9500 Gilman Drive La Jolla CA92093 United States

In this work, the thermo-optic coefficient (TOC) of the silicon-rich amorphous silicon carbide (a-SiC) thin film deposited by plasma-enhanced chemical vapor deposition is characterized. We found that the TOC of the film increases as its silicon content increases. A more than three-fold improvement in the TOC is measured, reaching a TOC as high as 1.88 × 10−4 ℃−1, which is comparable to that of crystalline silicon. An efficient thermo-optic phase shifter has also been demonstrated by integrating the silicon-rich a-SiC micro-ring structure with a NiCr heater. Its tuning efficiency Pπ as low as 4.4 mW has been measured at an optical wavelength of 1550 nm. These findings make silicon-rich a-SiC a good material candidate for thermo-optic applications in photonic integrated circuits. Copyright © 2022, The Authors. All rights reserved.

关键词： Silicon carbide

All-antiferromagnetic electrically controlled memory on silicon featuring large tunneling magnetoresistance

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Shi, Jiacheng Lopez-Dominguez, Victor Arpaci, Sevdenur Sangwan, Vinod K. Mahfouzi, Farzad Kim, Jinwoong Athas, Jordan G. Hamdi, Mohammad Aygen, Can Phatak, Charudatta Carpentieri, Mario Jiang, Jidong S. Grayson, Matthew A. Kioussis, Nicholas Finocchio, Giovanni Hersam, Mark C. Amiri, Pedram Khalili Department of Electrical and Computer Engineering Northwestern University EvanstonIL60208 United States Universitat Jaume I Castellón12006 Spain Applied Physics Program Northwestern University EvanstonIL60208 United States Department of Materials Science and Engineering Northwestern University EvanstonIL60208 United States Department of Physics and Astronomy California State University Northridge NorthridgeCA91330 United States Materials Science Division Argonne National Laboratory LemontIL60439 United States Department of Electrical and Information Engineering Politecnico di Bari Bari Italy Department of Mathematical and Computer Sciences Physical Sciences and Earth Sciences University of Messina Messina98166 Italy Department of Chemistry Northwestern University EvanstonIL60208 United States

Antiferromagnetic (AFM) materials are a pathway to spintronic memory and computing devices with unprecedented speed, energy efficiency, and bit density. Realizing this potential requires AFM devices with simultaneous electrical writing and reading of information, which are also compatible with established silicon-based manufacturing. Recent experiments have shown tunneling magnetoresistance (TMR) readout in epitaxial AFM tunnel junctions. However, these TMR structures were not grown using a silicon-compatible deposition process, and controlling their AFM order required external magnetic fields. Here we show three-terminal AFM tunnel junctions based on the noncollinear antiferromagnet PtMn3, sputter-deposited on silicon. The devices simultaneously exhibit electrical switching using electric currents, and electrical readout by a large room-temperature TMR effect. First-principles calculations explain the TMR in terms of the momentum-resolved spin-dependent tunneling conduction in tunnel junctions with noncollinear AFM electrodes. Copyright © 2023, The Authors. All rights reserved.

关键词： Silicon

Representing and extending ensembles of parsimonious evolutionary histories with a directed acyclic graph

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Dumm, Will Barker, Mary Howard-Snyder, William DeWitt, William S. Matsen, Frederick A. Computational Biology Program Fred Hutchinson Cancer Research Center SeattleWA United States Howard Hughes Medical Institute Computational Biology Program Fred Hutchinson Cancer Research Center SeattleWA United States Paul G. Allen School of Computer Science and Engineering University of Washington SeattleWA United States Department of Electrical Engineering and Computer Sciences University of California BerkeleyCA United States Department of Genome Sciences University of Washington SeattleWA United States Department of Statistics University of Washington SeattleWA United States

In many situations, it would be useful to know not just the best phylogenetic tree for a given data set, but the collection of high-quality trees. This goal is typically addressed using Bayesian techniques, however, current Bayesian methods do not scale to large data sets. Furthermore, for large data sets with relatively low signal one cannot even store every good tree individually, especially when the trees are required to be bifurcating. In this paper, we develop a novel object called the "history subpartition directed acyclic graph" (or "history sDAG" for short) that compactly represents an ensemble of trees with labels (e.g. ancestral sequences) mapped onto the internal nodes. The history sDAG can be built efficiently and can also be efficiently trimmed to only represent maximally parsimonious trees. We show that the history sDAG allows us to find many additional equally parsimonious trees, extending combinatorially beyond the ensemble used to construct it. We argue that this object could be useful as the "skeleton" of a more complete uncertainty *** Codes 92-08 (Primary) 92B10, 92-04 (Secondary) Copyright © 2023, The Authors. All rights reserved.

关键词： Phylogenetic

Beyond Bounds on Light Scattering with Complex Frequency Excitations

学校读者我要写书评

暂无评论

Physical Review Letters 2022年第20期129卷 203601-203601页

作者： Seunghwi Kim Sergey Lepeshov Alex Krasnok Andrea Alù Photonics Initiative Advanced Science Research Center City University of New York New York New York 10031 USA DTU Electro Department of Electrical and Photonics Engineering Technical University of Denmark Ørsteds Plads 343 Kgs. Lyngby DK-2800 Denmark Department of Electrical and Computer Engineering Florida International University Miami Florida 33174 USA Physics Program Graduate Center City University of New York New York New York 10016 USA

Light scattering is one of the most established wave phenomena in optics, lying at the heart of light-matter interactions and of crucial importance for nanophotonic applications. Passivity, causality, and energy conservation imply strict bounds on the degree of control over scattering from small particles, with implications on the performance of many optical devices. Here, we demonstrate that these bounds can be surpassed by considering excitations at complex frequencies, yielding extreme scattering responses as tailored nanoparticles reach a quasi-steady-state regime. These mechanisms can be used to engineer light scattering of nanostructures beyond conventional limits for noninvasive sensing, imaging, and nanoscale light manipulation.

关键词： Metamaterials Mie scattering

Subdiffusive wave transport and weak localization transition in three-dimensional stealthy hyperuniform disordered systems

学校读者我要写书评

暂无评论

Physical Review B 2022年第6期105卷 064204-064204页

作者： F. Sgrignuoli S. Torquato L. Dal Negro Department of Electrical and Computer Engineering & Photonics Center Boston University 8 Saint Mary's Street Boston Massachusetts 02215 USA Department of Chemistry Department of Physics Princeton Institute for the Science and Technology of Materials and Program in Applied and Computational Mathematics Princeton University Princeton New Jersey 08540 USA Division of Material Science and Engineering Boston University 15 Saint Mary's Street Brookline Massachusetts 02446 USA Department of Physics Boston University 590 Commonwealth Avenue Boston Massachusetts 02215 USA

The purpose of this work is to understand the fundamental connection between structural correlations and light localization in three-dimensional (3D) open scattering systems of finite size. We numerically investigate the transport of vector electromagnetic waves scattered by resonant electric dipoles spatially arranged in 3D space by stealthy hyperuniform disordered point patterns. Three-dimensional stealthy hyperuniform disordered systems are engineered with different structural correlation properties determined by their degree of stealthiness χ. Such fine control of exotic states of amorphous matter enables the systematic design of optical media that interpolate in a tunable fashion between uncorrelated random structures and crystalline materials. By solving the electromagnetic multiple scattering problem using Green's matrix spectral method, we establish a transport phase diagram that demonstrates a distinctive transition from a diffusive to a weak localization regime beyond a critical scattering density that depends on χ. The transition is characterized by studying the Thouless number and the spectral statistics of the scattering resonances. In particular, by tuning the χ parameter, we demonstrate large spectral gaps and suppressed subradiant proximity resonances, facilitating light localization. Moreover, consistently with previous studies, our results show a region of the transport phase diagram where the investigated scattering systems become transparent. Our work provides a systematic description of the transport and weak localization properties of light in stealthy hyperuniform structures and motivates the engineering of photonic systems with enhanced light-matter interactions for applications to both classical and quantum devices.

关键词： Anderson localization Light-matter interaction Mesoscopics Photonics Scattering theory Weak localization Metamaterials

The 2020 power blackout in the brazilian state of amapá: Lessons learned and opportunities

学校读者我要写书评

暂无评论

The 2020 power blackout in the brazilian state of amapá: Le...

2021 IEEE PES Innovative Smart Grid Technologies Conference - Latin America, ISGT Latin America 2021

作者： Do Prado, Josue Campos Santos, Marcos Vinicius Ramos, Lucas Feksa Canha, Luciane Neves Konzen, Gabriel School of Engineering Computer Science Washington State University VancouverWA98686 United States School of Engineering Newcastle University Newcastle Upon TyneNE1 7RU United Kingdom Federal University of Rondônia Department of Electrical Engineering Porto Velho Brazil Graduate Program in Electrical Engineering Federal University of Santa Maria Santa Maria Brazil Division of Energy Economics and Environmental Studies Rio de Janeiro Brazil

ISBN: (纸本)9781665444217

The Brazilian state of Amapa=-notable for having 70% of its area consisting of unexplored Amazonian forests-experienced a massive power blackout in November 2020. An internal short-circuit in a transformer led to an explosion followed by a fire at the Macapa substation. This event left about 90% of the Amapa's population without power for several days in the middle of the COVID-19 pandemic, disrupting the lives of hundreds of thousands of people and raising concerns from several electricity sector stakeholders. This paper discusses select lessons learned from the power blackout and presents future opportunities to improve power system reliability and resilience for the grids in Amapá, and other similar Brazilian states, with a special focus on the increasingly important role of of distributed energy resources and smart grid technologies. © 2021 IEEE.

关键词： Outages

Plug-and-play stability for intracortical brain-computer interfaces: a one-year demonstration of seamless brain-to-text communication 23

学校读者我要写书评

暂无评论

Plug-and-play stability for intracortical brain-computer int...

Proceedings of the 37th International Conference on Neural Information Processing Systems

作者： Chaofei Fan Nick Hahn Foram Kamdar Donald Avansino Guy H. Wilson Leigh Hochberg Krishna V. Shenoy Jaimie M. Henderson Francis R. Willett Department of Computer Science Stanford University Department of Neurosurgery Stanford University Howard Hughes Medical Institute at Stanford University Department of Neuroscience Stanford University School of Engineering and Carney Institute for Brain Science Brown University and VA RR & D Center for Neurorestoration and Neurotechnology VA Providence Healthcare System and Center for Neurotechnology and Neurorecovery Department of Neurology Massachusetts General Hospital Harvard Medical School Bio-X Program Stanford University and Department of Neurobiology Stanford University and Department of Bioengineering Stanford University and Wu Tsai Neurosciences Institute Stanford University and Howard Hughes Medical Institute at Stanford University and Department of Electrical Engineering Stanford University Department of Neurosurgery Stanford University and Wu Tsai Neurosciences Institute Stanford University

Intracortical brain-computer interfaces (iBCIs) have shown promise for restoring rapid communication to people with neurological disorders such as amyotrophic lateral sclerosis (ALS). However, to maintain high performance over time, iBCIs typically need frequent recalibration to combat changes in the neural recordings that accrue over days. This requires iBCI users to stop using the iBCI and engage in supervised data collection, making the iBCI system hard to use. In this paper, we propose a method that enables self-recalibration of communication iBCIs without interrupting the user. Our method leverages large language models (LMs) to automatically correct errors in iBCI outputs. The self-recalibration process uses these corrected outputs ("pseudo-labels") to continually update the iBCI decoder online. Over a period of more than one year (403 days), we evaluated our Continual Online Recalibration with Pseudo-labels (CORP) framework with one clinical trial participant. CORP achieved a stable decoding accuracy of 93.84% in an online handwriting iBCI task, significantly outperforming other baseline methods. Notably, this is the longest-running iBCI stability demonstration involving a human participant. Our results provide the first evidence for long-term stabilization of a plug-and-play, high-performance communication iBCI, addressing a major barrier for the clinical translation of iBCIs.

关键词：