HOTI - Hot Interconnects Symposium Courses

Building Custom AI Infrastructure with NVLink Fusion

Discover how to build custom AI infrastructure using NVLink Fusion technology for high-performance computing and accelerated workloads in this technical presentation.

• YouTube
• 28 minutes
• Self-Paced
• Free Video

Principles and Practice of Scalable and Distributed Deep Neural Networks

Discover principles and practices for building scalable distributed deep neural networks in this comprehensive 3.5-hour tutorial from Ohio State University experts.

• YouTube
• 3 hours 30 minutes
• Self-Paced
• Free Video

GPU Communication Libraries for Accelerating HPC and AI Applications

Discover GPU communication libraries to accelerate high-performance computing and AI applications through hands-on techniques and optimization strategies.

• YouTube
• 2 hours 38 minutes
• Self-Paced
• Free Video

Accelerating Frontier MoE Training with 3D Integrated Optics

Discover how 3D integrated optics can accelerate Mixture of Experts (MoE) model training for frontier AI systems through advanced interconnect technologies.

• YouTube
• 23 minutes
• Self-Paced
• Free Video

Characterizing Communication Patterns in Distributed LLM Inference

Explore communication patterns and performance characteristics in distributed large language model inference systems through detailed analysis and optimization strategies.

• YouTube
• 33 minutes
• Self-Paced
• Free Video

Demystifying NCCL - An In-Depth Analysis of GPU Communication Protocols and Algorithms

Dive into NCCL's GPU communication protocols and algorithms through expert analysis from ETH Zurich and NVIDIA researchers, covering optimization strategies for distributed computing.

• YouTube
• 29 minutes
• Self-Paced
• Free Video

Libfabric Communication Library Tutorial

Discover Intel's Libfabric library for high-performance network communication in HPC environments through hands-on examples and implementation strategies.

• YouTube
• 1 hour 46 minutes
• Self-Paced
• Free Video

Ultra Ethernet for Next Generation AI and HPC Workloads

Discover how Ultra Ethernet revolutionizes AI and HPC workloads with advanced networking solutions for next-generation computing infrastructure.

• YouTube
• 32 minutes
• Self-Paced
• Free Video

Deadlock Free Routing for Full Mesh Networks Without Using Virtual Channels

Discover deadlock-free routing techniques for full mesh networks that eliminate the need for virtual channels in high-performance interconnect systems.

• YouTube
• 24 minutes
• Self-Paced
• Free Video

Bittide - Control Time Not Flows

Discover Bittide's innovative approach to controlling time rather than data flows in high-performance interconnect systems through expert analysis and technical insights.

• YouTube
• 21 minutes
• Self-Paced
• Free Video

Silicon Photonic Accelerated Memory Pooling for Efficient Compute Resource Allocation

Explore silicon photonic technology for accelerated memory pooling and optimized compute resource allocation in modern data centers.

• YouTube
• 35 minutes
• Self-Paced
• Free Video

On Package Memory with Universal Chiplet Interconnect Express UCIe

Explore on-package memory solutions and Universal Chiplet Interconnect Express (UCIe) technology with industry experts from Intel and AMD in this technical symposium session.

• YouTube
• 30 minutes
• Self-Paced
• Free Video

High Performance and Smart Networking Technologies for HPC and AI

Explore cutting-edge networking technologies optimized for high-performance computing and artificial intelligence applications in this comprehensive 3-hour tutorial.

• YouTube
• 3 hours 3 minutes
• Self-Paced
• Free Video

Now in Focus - The Fifth GenAI Epoch of Computing Infrastructure

Explore the transformative fifth epoch of computing infrastructure driven by generative AI and its revolutionary impact on modern technology systems.

• YouTube
• 1 hour 5 minutes
• Self-Paced
• Free Video

Evolving Interconnect Fabric and SDN in the Cloud to Meet the Needs of AI

Discover how interconnect fabric and software-defined networking evolve in cloud environments to support demanding AI workloads and infrastructure requirements.

• YouTube
• 1 hour 1 minute
• Self-Paced
• Free Video