despite significant progress in computer vision technology, the development of realistic, real-time scene rendering via neural radiation fields remains hindered by high computational demands and time-intensive renderi...
详细信息
ISBN:
(纸本)9798350386851;9798350386844
despite significant progress in computer vision technology, the development of realistic, real-time scene rendering via neural radiation fields remains hindered by high computational demands and time-intensive rendering processes. This limitation makes such systems impractical for applications requiring rapid feedback. The study proposes an immersive navigation system utilizing 3d Gaussian splatters to facilitate high-quality dynamic rendering. This system seamlessly integrates various technologies, including Unity3d, ChatGPT, Whisper, and voice generators, and is equipped with the HTC VIVE Pro, offering users an unparalleled exploration experience. With potential applications in museums, art galleries, shopping malls, and gaming, this system holds the prospect of unveiling new levels of realism and interactivity in virtual environments.
3d heterogeneous integration has become increasingly important as it allows continued package system scaling in combination with transistor scaling to increase density and performance. As solder-based interconnect rea...
详细信息
ISBN:
(纸本)9798350375985;9798350375992
3d heterogeneous integration has become increasingly important as it allows continued package system scaling in combination with transistor scaling to increase density and performance. As solder-based interconnect reaches its pitch scaling limit around 10 mu m, hybrid bonding interconnect (HBI) is the next generation interconnect technology to continue the pitch scaling journey beyond 10 mu m. HBI offers unique capabilities that enable ultra-high interconnect density and bandwidth, low interconnect power consumption and latency, and various new 3d packaging architectures. The benefit of having HBI is crucial for future AI and high performance computing (HPC) systems. In this paper, we present a systematic study of chip-to-wafer hybrid bonding (C2WHB) down to 3 mu m pitch and below. Starting with the discussion about challenges for the placement accuracy requirement with symmetric paddesign to enable sub-3 mu m pitch chip-to-wafer HBI. We show excellent bonding quality, electrical yield and reliability are achieved with comprehensive optimization between design, process control and tool/material development. Finally, we cover 3d heterogeneous integration architectures with C2W HB.
The impact of informed high level decisions on the performance-power efficiency of semiconductors is well known. From the context of heterogeneous integration for 3d IC, new innovations are needed which close the loop...
详细信息
ISBN:
(纸本)9798350387186;9798350387179
The impact of informed high level decisions on the performance-power efficiency of semiconductors is well known. From the context of heterogeneous integration for 3d IC, new innovations are needed which close the loop from architecture to device;from multiphysics to high level abstraction;from performance to security. Such considerations will be presented.
As 3ddie-on-Wafer and Wafer-on-Wafer stacking technologies mature, integrationdensity will increase to provide new opportunities for complex homogeneous or heterogeneous integration of logic circuits across multiple...
详细信息
ISBN:
(纸本)9798350387186;9798350387179
As 3ddie-on-Wafer and Wafer-on-Wafer stacking technologies mature, integrationdensity will increase to provide new opportunities for complex homogeneous or heterogeneous integration of logic circuits across multiple materials and through multiple interconnect materials and technologies. This will create opportunities for new advancements in design, verification and analyses tools anddesign flows, test strategies and test planning, andsystem level practices.
As systems grow in size and complexity, the challenge of comprehensively maintaining and understanding their structure also increases. Utilizing a Model-Basedsystems Engineering (MBSE) approach can be beneficial in a...
详细信息
data-driven control, which embraces artificial intelligence, machine learning, and experience-based inferencing architectures, has gained significant interest for its ability to provide robust optimization in model-fr...
详细信息
ISBN:
(纸本)9798350355376;9798350355369
data-driven control, which embraces artificial intelligence, machine learning, and experience-based inferencing architectures, has gained significant interest for its ability to provide robust optimization in model-free, nonlinear, and time-varying paradigms. Traditional systems, such as the haptic paddle, used to communicate systemdynamics principles in undergraduate curricula, have yet to be adapted to the memory and processing requirements of data-driven control. In this work, we present a modular, open-source 3d printable friction-driven haptic paddle design, building on the designs proposed by the community, using commercial components and simple microelectronic packaging, to enable robust data-driven control for integration in undergraduate education. We make use of the RP2040 microcontroller, a small light-weight logic platform capable of fast online computation and robust memory storage for onboarddata-driven control. To validate our design, we first develop an experimental model of the physical dynamics that shows that our 3d printed friction drive is comparable with friction driven paddles and capstan-cable driven paddles. Further, we demonstrate the utility of our design in explicating data-driven control by presenting the development of basic machine learning and reinforcement learning architectures for online, model-free robust control in the presence of time-variable plant dynamics in a trajectory tracking task that is well suited for implementation in undergraduate and introductory graduate systemdynamics and controls curricula.
While 3d X-ray Tools or X-ray Microscopes (XRM) have been the traditional nondestructive imaging tools in Failure Analysis (FA) labs, they normally involve smaller packages or small sections cut out of a package or PC...
详细信息
ISBN:
(纸本)9798350360615;9798350360608
While 3d X-ray Tools or X-ray Microscopes (XRM) have been the traditional nondestructive imaging tools in Failure Analysis (FA) labs, they normally involve smaller packages or small sections cut out of a package or PCB if high resolution better than 1 um is desired. They are seldom used for intact board level FA nor is a tool of choice for inline metrology for defect inspection in wafer level integration. As sample size increases to several cm, such as 2.5d/3d Packages, Multi Chiplet Modules (MCM), PCBs or a 300 mm wafer, high resolution (< 1 um) imaging is not practical nor in many cases achievable. The acquisition time takes too long (typically hours);images are noisy and are filled with beam hardening artifacts (with bands and streaks). In this paper we will revisit the use of X-rays for high resolution (sub-micron) board level FA and in line metrology for next generation wafer level integration using a novel 3d X-ray system that can achieve sub-micron resolution on large PCBs to 300 mm wafers in as little as a few minutes to evaluate failures or defects in 3d.
Next-generation semiconductor hardware technologies andsystemintegration serve as the physical foundation in the pursuit of ubiquitous machine intelligence, with unprecedented requirements in energy efficiency, perf...
详细信息
ISBN:
(纸本)9798350323481
Next-generation semiconductor hardware technologies andsystemintegration serve as the physical foundation in the pursuit of ubiquitous machine intelligence, with unprecedented requirements in energy efficiency, performance, cost effectiveness, and security. Here, we provide an overview of emerging technologies with an emphasis on 3dsystemintegration, anddiscuss on cross-layer designs for memory-centric computing in the 3d era.
This paper presents an investigation of system-onChip (SoC) communication latency optimization for 3dsystemintegration and highlights the role of architectural modifications to maximize the Power, Performance, &...
详细信息
ISBN:
(纸本)9798350330991;9798350331004
This paper presents an investigation of system-onChip (SoC) communication latency optimization for 3dsystemintegration and highlights the role of architectural modifications to maximize the Power, Performance, & Area (PPA) benefits. An instance of a highly configurable RISC-V SoC is implemented using similar to 2nm nanosheet technology anddifferent 3d stacking options using design flow from sign-off tools. The proposed implementation targets performance optimization for different 3d partitioning scenarios: Memory-on-Logic (MoL) & Logicon-Logic (LoL). We target 2-die 3d Integrated Circuits (3dIC) with high density 3d interconnect using Face-to-Face (F2F) hybrid bonding (similar to 1 mu m), and 3-die stack, as Face-to-Back (F2B) on top of F2F. Our analysis of the 16-core SoC instance shows that the proposed architectural optimizations bring a significant reduction of 4 pipeline stages in the design hierarchy at a marginal cost of 9% effective frequency loss when implemented in 3d in comparison to the baseline 2d architecture. Further, going from 2d to 3d allows more than 40% total system wire-length reduction & 10% less cell area, resulting in 20% power savings. These findings hold promise for further explorations on manycore SoC instances (256 & more) facing system interconnect challenges.
In the era of big data and large models, automatic annotating functions for multi-modal data are of great significance for real-world AI-driven applications, such as autonomous driving and embodied AI. Unlike traditio...
ISBN:
(纸本)9798350384581;9798350384574
In the era of big data and large models, automatic annotating functions for multi-modal data are of great significance for real-world AI-driven applications, such as autonomous driving and embodied AI. Unlike traditional closed-set annotation, open-vocabulary annotation is essential to achieve human-level cognition capability. However, there are few open-vocabulary auto-labeling systems for multi-modal 3ddata. In this paper, we introduce OpenAnnotate3d, an open-source open-vocabulary auto-labeling system that can automatically generate 2d masks, 3d masks, and3d bounding box annotations for vision and point clouddata. Our system integrates the chain-of-thought capabilities of Large Language Models (LLMs) and the cross-modality capabilities of vision-language models (VLMs). To the best of our knowledge, OpenAnnotate3d is one of the pioneering works for open-vocabulary multi-modal 3d auto-labeling. We conduct comprehensive evaluations on both public and in-house real-worlddatasets, which demonstrate that the system significantly improves annotation efficiency compared to manual annotation while providing accurate open-vocabulary auto-annotating results.
暂无评论