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:20260202T201806Z
LOCATION:264
DTSTART;TZID=America/Chicago:20251116T144500
DTEND;TZID=America/Chicago:20251116T150000
UID:submissions.supercomputing.org_SC25_sess204_ws_ss103@linklings.com
SUMMARY:Run-time Energy-Efficiency Optimization for AI and HPC Workloads
DESCRIPTION:Gabriel Hautreux (LIRMM, Univ. Montpellier, CNRS; CINES) and A
 bdoulaye Gamatié and Gilles Sassatelli (LIRMM, Univ. Montpellier, CNRS)\n\
 nEffective power management is crucial for balancing high performance and 
 environmental impact in the exascale era, particularly for datacenters dom
 inated by massively parallel GPU systems due to the rise of AI. While many
  strategies rely on deep application knowledge, there is a growing need fo
 r application-agnostic approaches. We introduce a node-level power managem
 ent runtime designed for regular applications, featuring minimal overhead 
 and seamless deployment across any HPC/AI system. Our approach detects, at
  runtime, repetitive execution patterns via spectral analysis and then tra
 ces per-pattern energy consumption. A simple gradient-descent optimizer gr
 adually adjusts the GPU frequency until the least per-pattern energy (i.e.
 , maximum energy efficiency) is found. With this approach, we demonstrate 
 up to a 15% reduction in energy consumption for equivalent computational t
 asks, with no overhead and minimal impact on execution time. This solution
  has been validated across a diverse range of AI applications, and we disc
 uss the resulting energy savings.\n\nRecording: Livestreamed, Recorded\n\n
 Registration Category: Technical Program Reg Pass, Workshop Reg Pass\n\nSe
 ssion Chairs: Mike Woodacre (Hewlett Packard Enterprise (HPE)); Michèle We
 iland (EPCC, The University of Edinburgh; The University of Edinburgh); Fu
 miyoshi Shoji (RIKEN Center for Computational Science (R-CCS), Center for 
 Computational Science); Pekka Manninen (CSC - IT Center for Science; Unive
 rsity of Helsinki, Finland); James H. Rogers (Oak Ridge National Laborator
 y (ORNL)); and Cate Berard (US Department of Energy)\n\n
END:VEVENT
END:VCALENDAR