Samba: Simple Hybrid State Space Models for Language Modeling
Dive into the mechanics of Samba, a hybrid state space model built on Mamba that enables efficient unlimited context language modeling for advanced AI applications.
Decoding on Graphs: Empowering LLMs with Knowledge Graphs Through Well-Formed Chains
Discover how to enhance LLM capabilities by integrating Knowledge Graphs through the Decoding on Graphs framework, focusing on graph-aware constrained decoding and well-formed reasoning chains.
Perfect Reasoning for AI Agents Using ReasonAgain - Symbolic Code Implementation
Discover how to enhance AI agent reasoning through symbolic programming, exploring ReasonAgain's methodology for improving LLM mathematical and logical capabilities across multiple coding languages.
Conversational AI-Enabled User Interface for HPC Tools
Discover how AI-powered conversational interfaces can simplify complex HPC tools, making performance analysis more accessible through natural language commands and improving productivity for both novice and expert users.
Adapting the Data Center to New AI Trends
Explore how AI trends reshape data center infrastructure, focusing on high-density server racks, liquid cooling solutions, and innovative approaches to meet evolving computational demands.
Generative AI and the Outsized Role of Serdes in Data Center Performance
Explore how AI's data center evolution puts interconnect and serdes technology at the forefront of scalability, examining power efficiency challenges and performance impacts across multiple scales.
Next Steps in AI Technology Adoption - Data Center Challenges and Solutions
Explore key challenges and solutions for AI technology adoption in data centers, from power management to cooling systems and global deployment strategies.
CXL Shared Memory Technology for Accelerating AI Cluster Performance
Discover how CXL shared memory technology revolutionizes AI cluster performance through Gismo software, enabling faster data access and reduced memory usage for large-scale AI frameworks.
Telemetry-Based Load Balancing of AI/ML Workloads in Self-Healing Networks
Discover how to optimize AI/ML workload performance through telemetry-based load balancing, network self-healing, and in-band monitoring for efficient Ethernet fabric architecture deployment.
Grand Teton - Meta's Next Generation AI Hardware Platform for Training and Inference
Discover Meta's next-gen AI hardware platform Grand Teton, exploring its modular system design, training capabilities, and innovative solutions for modern AI infrastructure challenges.
Make Quantum Computing Useful - Performance Management Software Solutions
Discover how AI-based quantum error suppression technology maximizes hardware performance and enables 1000X improvement in quantum computing systems through Fire Opal software.
AI Networking Market Overview - Implications for Data Center Infrastructure
Explore the unique characteristics of AI backend networks and their crucial role in connecting accelerated servers for large language models and generative AI workloads.
The Next Wave of Computing with Optical I/O - Scaling AI and Cloud Computing Performance
Discover how optical I/O and photonic solutions revolutionize data center performance, enabling higher bandwidth and lower latency for AI, HPC, and cloud computing through advanced chiplet technology.
CXL 2.0 Switch Enabling Composable Memory Architecture in AI and HPC Computing
Explore the role of CXL 2.0 switching technology in enabling flexible memory architectures for advanced AI and high-performance computing systems.
Scalability of Memory Subsystem in AI Storage Systems - IO Profiling and Performance Analysis
Explore IO profiling in AI systems, examining pipeline stages, metadata costs, and filesystem latency impacts on GPU performance, with solutions for optimizing storage systems.
Modeling Methodology and Tools for Hardware-Software Codesign
Explore methodologies and tools for optimizing AI systems through hardware-software co-design, focusing on emerging challenges in large language models and generative AI architectures.
Architectural Challenges and Innovation for Compute Infrastructure Co-Design in Generative AI
Explore architectural solutions for generative AI computing challenges, focusing on heterogeneous computing infrastructure and chiplet-based systems for improved energy efficiency.
The Future of AI/ML Innovation Through Row-Scale Disaggregation
Explore how Composable Disaggregated Infrastructure revolutionizes system architecture for AI/ML applications through innovative network fabric design and extended PCIe capabilities.
BBCube 3D - Heterogeneous 3D Integration for TB/s Bandwidth in AI and HPC Applications
Discover groundbreaking BBCube 3D technology for heterogeneous integration, offering enhanced bandwidth and energy efficiency through innovative stacking processes and TSV implementation for AI/HPC applications.
Advanced Packaging Technologies for Chiplet Adoption and Memory Integration in HPC and AI Applications
Explore cutting-edge packaging technologies and chiplet design innovations for next-generation AI and HPC systems, including D2D protocols, hybrid copper bonding, and photonic integration.
Open Infrastructure Platforms to Usher in the Age of Generative AI
Explore Meta's infrastructure innovations for GenAI, from MTIA chips to Grand Teton systems, covering deployment strategies, liquid cooling advances, and future data center challenges.