Technology

Edge AI Safety: Agentic AI Architecture That Leverages 3D To Integrate A Dedicated Safety Layer (Princeton, HKUST, NC State Univ.)

Source

SemiEngineering

Published

Nov 17, 2025

TL;DR

AI Generated

Researchers from Princeton University, Hong Kong University of Science and Technology, and North Carolina State University published a technical paper titled “3D Guard-Layer: An Integrated Agentic AI Safety System for Edge Artificial Intelligence.” The paper proposes an AI safety architecture leveraging 3D integration to enhance safety in edge AI systems. By dynamically learning and mitigating attacks against AI systems, this architecture aims to improve system reliability and performance while minimizing costs. The system utilizes the advantages of co-location with edge computing hardware to monitor, detect, and proactively address threats to AI systems.

Read Full Article

Intel VP claims up to 30% of CPU performance is untapped by modern games — software optimization is critical to unlocking full potential of hybrid CPUs

Intel's VP Robert Hallock emphasizes that up to 30% of CPU performance in modern games remains untapped due to software optimization issues with hybrid CPUs. Hallock discusses the impact of disabling E-cores on Intel CPUs for better performance, noting that both E-cores and P-cores contribute equally to overall operation. He highlights the importance of software optimization in maximizing CPU performance, mentioning Intel's binary optimization feature in Arrow Lake chips as an example. Hallock suggests that software optimization is crucial for extracting more performance from existing silicon, with potential for significant gains in efficiency.

Tom's Hardware•

6 days ago

SemiEngineering

Can Edge AI Keep Up?

Experts discuss the challenges of keeping edge AI architectures adaptable while maintaining power, performance, and area targets. The cadence for model updates varies by application, with some industries experiencing rapid changes while others remain more static. Heterogeneous architectures and robust software/compiler toolchains are crucial for balancing adaptability with efficiency. The discussion includes insights from industry leaders at Arm, Cadence, Expedera, Mixel, Quadric, Rambus, Siemens EDA, and Synopsys on the evolving landscape of AI model development and hardware design.

SemiEngineering•

1 week ago

Apex Gaming PCs recalls nearly 18,000 power supplies over missing safety labels — the fix is a warning sticker sent by mail, units are entirely safe

Apex Gaming PCs is recalling nearly 18,000 power supplies due to missing safety labels that identify electrical shock and electrocution hazards. The affected units were sold between December 2021 and February 2026, both as standalone components and pre-installed in gaming desktops. The recall is not due to a hardware defect, as the power supplies are safe to use, but lack necessary warning labels. Customers will receive warning stickers by mail or can request a replacement PSU. The company advises users to stop using the recalled PSUs and not to open or service them.

Tom's Hardware•

1 week ago

Speculation: Silicon’s Most Expensive Compulsion

Dr. Thang Tran, the Founder and CTO of Simplex Micro, discusses the significant silicon area and power dedicated to speculative execution in modern high-performance processors. He introduces Time-Based Scheduling (TBS) as an alternative architecture that eliminates the need for speculation by dispatching instructions when their data is ready. TBS focuses on vector processing units (VPUs) for AI workloads, where speculation can lead to wasted work across data elements. By removing speculation overhead, TBS allows for more efficient allocation of resources like execution units, cache capacity, and core count. The article highlights the drawbacks of speculation, the benefits of TBS, and the potential impact on various workload types.

SemiWiki•

2 weeks ago