Ponte Vecchio Across the Atlantic: Single-Node Benchmarking of Two Intel GPU Systems

Thomas Applencourt; Aditya Sadawarte; Servesh Muralidharan; Colleen Bertoni; JaeHyuk Kwack; Ye Luo; Esteban Rangel; John Tramm; Yasaman Ghadar; Arjen Tamerus; Chris Edsall; Tom Deakin

doi:10.1109/SCW63240.2024.00184

Ponte Vecchio Across the Atlantic: Single-Node Benchmarking of Two Intel GPU Systems

Thomas Applencourt, Aditya Sadawarte, Servesh Muralidharan, Colleen Bertoni, JaeHyuk Kwack, Ye Luo, Esteban Rangel, John Tramm, Yasaman Ghadar, Arjen Tamerus, Chris Edsall, Tom Deakin

School of Computer Science

Research output: Chapter in Book/Report/Conference proceeding › Conference Contribution (Conference Proceeding)

145 Downloads (Pure)

Abstract

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.

Original language	English
Title of host publication	Proceedings of the 15th IEEE International Workshop on Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems
Publisher	Association for Computing Machinery (ACM)
Pages	1430-1442
Number of pages	13
ISBN (Electronic)	979-8-3503-5554-3
DOIs	https://doi.org/10.1109/SCW63240.2024.00184
Publication status	Published - 26 Nov 2024
Event	15th IEEE International Workshop on Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems, PMBS 2024 - Atlanta, United States Duration: 18 Nov 2024 → 18 Nov 2024 https://pmbs-workshop.github.io/

Conference

Conference	15th IEEE International Workshop on Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems, PMBS 2024
Abbreviated title	PMBS 2024
Country/Territory	United States
City	Atlanta
Period	18/11/24 → 18/11/24
Internet address	https://pmbs-workshop.github.io/

Access to Document

10.1109/SCW63240.2024.00184

Full-text PDF (accepted manuscript)
This is the accepted author manuscript (AAM) of the article which has been made Open Access under the University of Bristol's Scholarly Works Policy. The final published version (Version of Record) can be found on the publisher's website. The copyright of any third-party content, such as images, remains with the copyright holder.
Accepted author manuscript, 261 KBLicence: CC BY

Handle.net

Persistent link

Cite this

Applencourt, T., Sadawarte, A., Muralidharan, S., Bertoni, C., Kwack, J., Luo, Y., Rangel, E., Tramm, J., Ghadar, Y., Tamerus, A., Edsall, C., & Deakin, T. (2024). Ponte Vecchio Across the Atlantic: Single-Node Benchmarking of Two Intel GPU Systems. In Proceedings of the 15th IEEE International Workshop on Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems (pp. 1430-1442). Association for Computing Machinery (ACM). https://doi.org/10.1109/SCW63240.2024.00184

@inproceedings{054216f81b774dfebca1f7e3e15ceabd,

title = "Ponte Vecchio Across the Atlantic: Single-Node Benchmarking of Two Intel GPU Systems",

abstract = "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.",

author = "Thomas Applencourt and Aditya Sadawarte and Servesh Muralidharan and Colleen Bertoni and JaeHyuk Kwack and Ye Luo and Esteban Rangel and John Tramm and Yasaman Ghadar and Arjen Tamerus and Chris Edsall and Tom Deakin",

year = "2024",

month = nov,

day = "26",

doi = "10.1109/SCW63240.2024.00184",

language = "English",

pages = "1430--1442",

booktitle = "Proceedings of the 15th IEEE International Workshop on Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems",

publisher = "Association for Computing Machinery (ACM)",

address = "United States",

note = "15th IEEE International Workshop on Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems, PMBS 2024, PMBS 2024 ; Conference date: 18-11-2024 Through 18-11-2024",

url = "https://pmbs-workshop.github.io/",

}

Applencourt, T, Sadawarte, A, Muralidharan, S, Bertoni, C, Kwack, J, Luo, Y, Rangel, E, Tramm, J, Ghadar, Y, Tamerus, A, Edsall, C & Deakin, T 2024, Ponte Vecchio Across the Atlantic: Single-Node Benchmarking of Two Intel GPU Systems. in Proceedings of the 15th IEEE International Workshop on Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems. Association for Computing Machinery (ACM), pp. 1430-1442, 15th IEEE International Workshop on Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems, PMBS 2024, Atlanta, Georgia, United States, 18/11/24. https://doi.org/10.1109/SCW63240.2024.00184

Ponte Vecchio Across the Atlantic: Single-Node Benchmarking of Two Intel GPU Systems. / Applencourt, Thomas; Sadawarte, Aditya; Muralidharan, Servesh et al.
Proceedings of the 15th IEEE International Workshop on Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems. Association for Computing Machinery (ACM), 2024. p. 1430-1442.

Research output: Chapter in Book/Report/Conference proceeding › Conference Contribution (Conference Proceeding)

TY - GEN

T1 - Ponte Vecchio Across the Atlantic

T2 - 15th IEEE International Workshop on Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems, PMBS 2024

AU - Applencourt, Thomas

AU - Sadawarte, Aditya

AU - Muralidharan, Servesh

AU - Bertoni, Colleen

AU - Kwack, JaeHyuk

AU - Luo, Ye

AU - Rangel, Esteban

AU - Tramm, John

AU - Ghadar, Yasaman

AU - Tamerus, Arjen

AU - Edsall, Chris

AU - Deakin, Tom

PY - 2024/11/26

Y1 - 2024/11/26

N2 - 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.

AB - 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.

U2 - 10.1109/SCW63240.2024.00184

DO - 10.1109/SCW63240.2024.00184

M3 - Conference Contribution (Conference Proceeding)

SP - 1430

EP - 1442

BT - Proceedings of the 15th IEEE International Workshop on Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems

PB - Association for Computing Machinery (ACM)

Y2 - 18 November 2024 through 18 November 2024

ER -

Applencourt T, Sadawarte A, Muralidharan S, Bertoni C, Kwack J, Luo Y et al. Ponte Vecchio Across the Atlantic: Single-Node Benchmarking of Two Intel GPU Systems. In Proceedings of the 15th IEEE International Workshop on Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems. Association for Computing Machinery (ACM). 2024. p. 1430-1442 doi: 10.1109/SCW63240.2024.00184

Ponte Vecchio Across the Atlantic: Single-Node Benchmarking of Two Intel GPU Systems

Abstract

Conference

Access to Document

Handle.net

Fingerprint

Cite this