Program
The times in the program below are Pacific Time (PT). We will keep updating the program as we get closer to the event date.
Conference Day 1: Wednesday, August 20
| Time | Title | Presenters |
|---|---|---|
| 09:20 | Introduction and Day 1 Welcome | General Chair: Artem Polyakov |
| 09:30 - 10:30 |
Keynote 1: Now in Focus: the Fifth, GenAI Epoch of Computing Infrastructure
[details] Session chair: Artem Polyakov |
Amin Vahdat, Google
|
| 10:30 - 10:45 |
Diamond Sponsor Talk: Lightmatter Transforming AI Infrastructure with the Power of 3D Photonics [details] |
Nick Harris |
| 10:45 - 11:00 |
Diamond Sponsor Talk: Cornelis Reimagining the Scale-Out Architecture for HPC and AI [details] |
Matt Williams |
| 11:00 - 12:00 |
Technical Paper Session A: Network Synchronization and Routing |
Session chair: Atiqul Mollah |
| bittide: Control Time, Not Flows | Martijn Bastiaan, Christiaan Baaij, Martin Izzard, Felix Klein, Sanjay Lall and Tammo Spalink | |
| Deadlock-free routing for Full-mesh networks without using Virtual Channels | Alejandro Cano Cos, Cristóbal Camarero, Carmen Martinez and Ramon Beivide | |
| 12:00 - 12:30 | Lunch | |
| 12:30 | Technical Paper Session B: Co-packaged Optics | Session chair: TBD |
| Accelerating Frontier MoE Training with 3D Integrated Optics | Taylor Groves, Mikhail Bernadskiy, Ho John Lee, Peter Carson, Thomas Graham and Eric Yeh | |
| Silicon Photonic Accelerated Memory Pooling for Efficient Compute Resource Allocation | Zhenguo Wu and Keren Bergman | |
| 13:30-13:45 |
Diamond Sponsor Talk: Cisco Managing the fabrics underpinning AI infrastructure [details] |
Will Eatherton |
| 13:45-14:00 |
Diamond Sponsor Talk: Nubis Overcoming Operational Barriers to AI Scale-Up with Co-Packaged Optics Integration [details] |
Peter Winzer |
| 14:00 - 15:30 |
Panel – LLM Token Economy: How is networking going to play in the age of agents?
[details] Session chair: Alan Benjamin |
Moderator: Peter Olcott (First Spark Ventures) Panelists: Phill Brown (Meta), Davor Capalija (Tenstorrent Inc.), Ethan Lockshin (Groq), Sid Sheth (d-Matrix) |
| 15:30 | The Head Bubba Memorial & HotI Closing Remarks Day 1 | Dan Pitt |
Conference Day 2: Thursday, August 21
| Time | Title | Presenters |
|---|---|---|
| 08:50 | Day 2 Welcome | General Chair: Artem Polyakov |
| 09:00 - 10:00 |
Keynote 2: Evolving inter-connect fabric and SDN in the cloud to meet the needs of AI
[details] Session chair: Dan Pitt |
Deepak Bansal (Microsoft)
|
| 10:00 - 10:15 |
Diamond Sponsor Talk: Avicena Paradigm Shift in AI Scale Up Clusters using microLED based Interconnects [details] |
Chris Pfistner |
| 10:15 - 10:25 |
Platinum Sponsor Talk: Arista Multiple Pathways to Energy Efficient Optics [details] |
Andreas Bechtolsheim |
| 10:25 - 10:30 |
Gold Sponsor Talk: GigaIO GigaIO: Empowering Every Accelerator to Lead the AI Inference Revolution [details] |
Alan Benjamin |
| 10:30 - 12:00 |
Technical Paper Session C: Interconnect Software – Runtimes & Workloads |
Session chair: Galen Shipman |
| Demystifying NCCL: An In-depth Analysis of GPU Communication Protocols and Algorithms | Zhiyi Hu, Siyuan Shen, Tommaso Bonato and Torsten Hoefler | |
| Characterizing Communication Patterns in Distributed LLM Inference | Lang Xu, Kaushik Kandadi Suresh, Quentin Anthony, Nawras Alnaasan and Dhabaleswar K Panda | |
| FPsPIN: An FPGA-based Open-Hardware Research Platform for Processing in the Network | Timo Schneider, Pengcheng Xu and Torsten Hoefler | |
| 12:00 - 12:30 | Lunch | |
| Invited Talk Session: Next Gen Interconnects | Session chair: TBD | |
| 12:30 - 13:00 | Ultra Ethernet for next-generation AI and HPC workloads [details] |
Torsten Hoefler |
| 13:00 - 13:30 | TBD | TBD |
| 13:30 - 13:40 |
Platinum Sponsor Talk: UALink Why UALink: A Protocol-Level Comparison for Rack-Scale AI [details] |
Sharada Yeluri |
| 13:40 - 13:50 |
Platinum Sponsor Talk: Celestial AI Photonic Fabric: Optical Scale-Up Networks for Accelerated Computing [details] |
Ravi Mahatme |
| 13:50 - 14:00 |
Platinum Sponsor Talk: Meta AI Infra at Meta [details] |
Wesley Bland |
| 14:00 - 15:00 | Technical Paper Session D: Scale-up Interconnects | Session chair: Letizia Giuliano |
| Micro-LED data interconnect for scale-up networks with record energy efficiency | Bardia Pezeshki, Alex Tselikov, Rob Kalman, Suresh Rangarajan, Ivan Huang, Jeff Pepper, Sarah Zou, Howard Rourke and Alasdair Fikouras | |
| On-Package Memory with Universal Chiplet Interconnect Express (UCIe): A Low Power, High Bandwidth, Low Latency and Low Cost Approach | Debendra Das Sharma, Swadesh Choudhary, Peter Onufryk and Robert Pelt | |
| 15:00 | HotI Awards | Ryan Grant |
| 15:10 | HotI Closing Remarks | Vice Chair: Sayan Ghosh |
Tutorial:
Conference Day 3: Friday, August 22
Track 1 (in parallel)
| Time | Title | Presenters |
|---|---|---|
| Morning Sessions | ||
| 08:30 - 12:00 | Principles and Practice of Scalable and Distributed Deep Neural Networks Training and Inference
[details] |
Dhabaleswar K. (DK) Panda, Nawras Alnaasan (The Ohio State University)
|
| 12:00 - 13:00 | Lunch | |
| Afternoon Session | ||
| 13:00 - 16:30 | High-Performance and Smart Networking Technologies for HPC and AI
[details] |
Dhabaleswar K. (DK) Panda, Benjamin Michalowicz (The Ohio State University)
|
Track 2 (in parallel)
| Time | Title | Presenters |
|---|---|---|
| Morning Sessions | ||
| 7:00 - 8:30 | Libfabric Communication Library
[details] |
Rajalaxmi Angadi, Alexia Ingerson (Intel Corporation)
|
| 08:30 - 12:00 | GPU Communication Libraries for Accelerating HPC and AI Applications
[details] |
Benjamin Glick, Pouya Kousha, Arnav Goel (NVIDIA), Andreas Herten (Jülich Supercomputing Centre)
|