BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260202T201805Z
LOCATION:267
DTSTART;TZID=America/Chicago:20251117T153000
DTEND;TZID=America/Chicago:20251117T160000
UID:submissions.supercomputing.org_SC25_sess198_ws_pmbsf108@linklings.com
SUMMARY:Characterizing the Impact of GPU Power Management on an Exascale S
 ystem
DESCRIPTION:Mariana Costa (UFRGS); Antigoni Georgiadou and James B. White 
 III (Oak Ridge National Laboratory (ORNL)); Bruno Villasenor Alvarez and J
 ordà Polo (Advanced Micro Devices, Inc. (AMD)); Woong Shin (Oak Ridge Nati
 onal Laboratory (ORNL)); Philippe O. A. Navaux (UFRGS); Bronson Messer (Oa
 k Ridge National Laboratory (ORNL)); and Arthur Lorenzon (Federal Universi
 ty of Rio Grande do Sul, Brazil)\n\nAs GPU-accelerated high-performance co
 mputing (HPC) systems approach exascale performance, controlling energy co
 nsumption without compromising throughput is essential. Architectures such
  as the AMD MI250X-based Frontier supercomputer provide runtime mechanisms
  like frequency and power capping, enabling energy tuning without modifyin
 g application code. Although both target energy reduction, they operate vi
 a distinct hardware control paths and influence workloads differently. We 
 present a comprehensive evaluation of these strategies on a leadership-cla
 ss system using diverse HPC proxy applications representative of productio
 n workloads. Our study analyzes performance–energy trade-offs across multi
 ple capping levels, node counts (1 and 32), and application profiles. Resu
 lts show that frequency capping generally achieves higher energy efficienc
 y and scalability, with gains of up to 13.2% without performance loss, whi
 le power capping is more effective for single-node runs or bursty GPU util
 ization. We also provide practical guidelines to help system administrator
 s and users balance energy efficiency and performance in large-scale scien
 tific workloads.\n\nRecording: Livestreamed, Recorded\n\nRegistration Cate
 gory: Technical Program Reg Pass, Workshop Reg Pass\n\nSession Chairs: Ste
 ven A. Wright (University of York, England); Simon Hammond (National Nucle
 ar Security Administration (NNSA)); and Sascha Hunold (Technical Universit
 y of Vienna)\n\n
END:VEVENT
END:VCALENDAR