BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20250626T234543Z
LOCATION:B312-B313A
DTSTART;TZID=America/New_York:20241120T110000
DTEND;TZID=America/New_York:20241120T113000
UID:submissions.supercomputing.org_SC24_sess497_gb104@linklings.com
SUMMARY:Breaking the Molecular Dynamics Timescale Barrier Using a Wafer-Sc
 ale System
DESCRIPTION:Kylee Santos (Cerebras Systems), Stan Moore (Sandia National L
 aboratories), Tomas Oppelstrup (Lawrence Livermore National Laboratory (LL
 NL)), Amirali Sharifian and Ilya Sharapov (Cerebras Systems), Aidan Thomps
 on (Sandia National Laboratories), Delyan Z. Kalchev (Cerebras Systems), D
 anny Perez (Los Alamos National Laboratory (LANL)), Robert Schreiber (Cere
 bras Systems), Scott Pakin (Los Alamos National Laboratory (LANL)), Edgar 
 A. Leon (Lawrence Livermore National Laboratory (LLNL)), James H. Laros II
 I (Sandia National Laboratories), Michael James (Cerebras Systems), and Si
 vasankaran Rajamanickam (Sandia National Laboratories)\n\nMolecular dynami
 cs (MD) simulations have transformed our understanding of the nanoscale, d
 riving breakthroughs in materials science, computational chemistry, and se
 veral other fields, including biophysics and drug design. Even on exascale
  supercomputers, however, runtimes are excessive for systems and timescale
 s of scientific interest. Here, we demonstrate strong scaling of MD simula
 tions on the Cerebras Wafer Scale Engine. By dedicating a processor core f
 or each simulated atom, we demonstrate a 457-fold improvement in timesteps
  per second versus the Frontier GPU-based exascale platform, along with a 
 large improvement in timesteps per unit energy. Reducing every year of run
 time to less than a day unlocks currently inaccessible timescales of slow 
 microstructure transformation processes that are critical for understandin
 g material behavior and function. Our dataflow algorithm runs embedded-ato
 m method (EAM) simulations at rates over 699k timesteps per second for pro
 blems with up to 800k atoms. This demonstrated performance is unprecedente
 d for general-purpose processing cores.\n\nRegistration Category: Tech Pro
 gram Reg Pass\n\nSession Chair: Amanda Randles (Duke University)\n\n
END:VEVENT
END:VCALENDAR