IXPUG PVC Users Group at SC24
Date/Time: Tuesday, Nov 19, 2024 3:30 to 5:00 PM
Location: Westin Peachtree, Chastain Room, Level 6 (210 Peachtree St NW, Atlanta, GA 30303). Open to all.
Registration: Please fill out the brief registration form if you'd like to attend and indicate if you'd like to give a lightning talk. Open to all.
Event Description: Those working with or interested in the Intel® Data Center GPU Max Series (a.k.a. Ponte Vecchio; PVC) will gather to share experiences with PVC and plans for deployment in the third meeting of the PVC User Group. Light snacks and beverage service will be provided.
Agenda:
All times are shown in ET / Atlanta, GA Time, UTC/GMT -5 hours. Event details are subject to change. Register at: https://events.cels.anl.gov/e/IXPUG-PVC-SC24
Tuesday, November 19
3:30-3:35 p.m. Welcome David Martin, Argonne Leadership Computing Facility, Argonne National Laboratory
3:35-3:55 p.m. Keynote Premanand Sakarda, Intel Corporation
Performance at Scale for HPC and AI Applications on the Aurora Supercomputer System
System level modeling and optimizations are critical to deliver performance at scale for supercomputers. On Aurora, as node count increases, the performance across compute, memory, and network subsystems are the primary driving forces for realized performance for HPC and AI applications. Performance for HPL and HPL MxP (aka HPL AI) depends on the single node performance, scaling efficiency, and application-level algorithmic improvements to effectively utilize the hardware resources. We will provide insights into the latest performance results on select benchmarks and applications across machine learning, data analytics, and simulation applications.
3:55-4:15 p.m. Keynote Thomas Applencourt, Argonne Leadership Computing Facility, Argonne National Laboratory
Micro-Benchmarking on PVC
Intel Data Center GPU Max 1550, known as Ponte Vecchio (PVC), is a new Intel GPU architecture for high-performance computing. It is the basis of two systems on the June 2024 Top 500 list, Dawn (\#51) and Aurora (\#2). This work provides micro-benchmarking data on PVCs from which application developers may benefit, shows how the micro-benchmarking results are indicative of mini-app performance on PVC, and demonstrates real applications on large-scale Intel GPU systems. We quantify the obtainable performance from PVC systems through micro-benchmarking fundamental architectural properties. We evaluate the performance of four mini-apps with known performance characteristics, and two full applications, comparing performance on a node of Aurora and Dawn with a node of NVIDIA H100 GPUs and a node of AMD MI250 GPUs. We show the figure-of-merit of the mini-apps on a single PVC ranges from 0.6--1.8X the performance of an H100, and 0.8--7.5X of a MI250.
4:15-4:45 p.m. Lightning Talks
- Amit Ruhela, Texas Advanced Computing Center, The University of Texas at Austin
- Md "Wasi" Rahman, Intel Corporation
Intel® SHMEM Experience on Intel® Tiber™ AI Cloud
This talk will present the details on how to get access to PVC instances on Intel® Tiber™ AI Cloud and share findings on enabling Intel® SHMEM on these instances. An example will follow to illustrate a benchmark run utilizing all 16 PVC tiles with Xe Link fabric for scale-up operations.
- Kacper Kornet, University of Cambridge
4:45-5:00 p.m. Refreshments and Discussion
5:00 p.m. Organization and Next Meeting
Questions? General questions should be sent to This email address is being protected from spambots. You need JavaScript enabled to view it.