Technology

Analog Plus 3D Optics to Accelerate AI inference and Combinatorial Optimization (Microsoft, Cambridge)

Source

SemiEngineering

Published

Sep 12, 2025

TL;DR

AI Generated

Researchers from Microsoft Research, Barclays, and University of Cambridge have published a technical paper introducing an analog optical computer (AOC) that combines analog electronics and 3D optics to accelerate AI inference and combinatorial optimization in a single platform. The AOC utilizes a rapid fixed-point search to avoid energy-intensive digital conversions and enhance noise robustness, enabling it to implement compute-bound neural models and advanced optimization approaches. Through case studies in image classification, regression, medical imaging, and financial transactions, the AOC demonstrates the benefits of co-designing hardware and abstraction for more efficient computing. This technology, built with scalable consumer-grade components, offers a promising path for faster and sustainable AI and optimization innovation.

Read Full Article

Disaggregating LLM Inference: Inside the SambaNova Intel Heterogeneous Compute Blueprint

SambaNova Systems and Intel have introduced a blueprint for heterogeneous inference that optimizes modern large language model (LLM) workloads by utilizing specialized hardware for different phases of inference: GPUs for prefill, SambaNova RDUs for decode, and Intel Xeon 6 CPUs for agentic tools and orchestration. This approach addresses the complexity of agentic AI systems with varying compute demands. By isolating tasks onto specific hardware, the architecture improves efficiency, scalability, and cost-effectiveness. The design reflects a shift towards specialized compute fabrics and better supports the evolving landscape of AI reasoning systems.

SemiWiki•

1 week ago

Intel and SambaNova team up on heterogenous AI inference platform — different hardware performs different workloads

Intel and SambaNova have collaborated on a new heterogeneous inference platform that utilizes different hardware components for various AI workloads. The platform leverages AI accelerators or GPUs for prefill, SambaNova's SN50 RDU for decoding, and Xeon 6 processors for agent-related operations and workload distribution. This architecture aims to compete with Nvidia by offering a scalable solution for enterprises and cloud operators, set to be available in the second half of 2026. The collaboration emphasizes the performance benefits of Xeon 6 processors and their compatibility with existing data center infrastructures.

Tom's Hardware•

3 weeks ago

Nvidia's new CPX GPU aims to change the game in AI inference — how the debut of cheaper and cooler GDDR7 memory could redefine AI inference infrastructure

Nvidia has introduced the Rubin CPX GPU, designed to enhance AI inference by focusing on the context phase with specialized hardware and 128GB of GDDR7 memory. The Rubin CPX aims to optimize long-context inference processing, enabling more efficient and cost-effective AI infrastructure. Nvidia's Dynamo software orchestration layer intelligently manages inference workloads across different GPUs in a disaggregated system, streamlining the process for developers. Companies like Cursor, Runway, and Magic are already planning to integrate Rubin CPX into their AI workflows for various applications. This shift in AI infrastructure represents a new paradigm, optimizing hardware resources for improved efficiency and scalability in AI processing.

Tom's Hardware•

8 months ago

Research Bits: Sept. 30

Researchers from various institutions have developed a hybrid memory system that combines ferroelectric capacitors and memristors into a single, CMOS-compatible memory stack for adaptive local training and inference of artificial neural networks. A magnetic transistor with built-in memory has been created by MIT and University of Chemistry and Technology Prague, allowing for improved signal strength and faster information retrieval. Additionally, researchers from the University of Gothenburg and Chalmers University of Technology have built light-powered gears on a micrometer scale for on-chip motors, enabling new possibilities for microscale mechanics and potential medical applications.

SemiEngineering•