{"id":"https://openalex.org/W7160318865","doi":"https://doi.org/10.48550/arxiv.2605.01938","title":"Cross-Layer Energy Analysis of Multimodal Training on Grace Hopper Superchips","display_name":"Cross-Layer Energy Analysis of Multimodal Training on Grace Hopper Superchips","publication_year":2026,"publication_date":"2026-05-03","ids":{"openalex":"https://openalex.org/W7160318865","doi":"https://doi.org/10.48550/arxiv.2605.01938"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.01938","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.01938","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.01938","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5135326321","display_name":"Mahmoud Ahmed","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ahmed, Mahmoud","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5087040072","display_name":"Sameh Abdulah","orcid":"https://orcid.org/0000-0002-8850-5753"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Abdulah, Sameh","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135375500","display_name":"Olatunji Ruwase","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ruwase, Olatunji","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5017294111","display_name":"Sam Ad\u00e9 Jacobs","orcid":"https://orcid.org/0000-0003-3425-5602"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jacobs, Sam Ade","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135318081","display_name":"Mathis Bode","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bode, Mathis","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135324235","display_name":"Mohamed Elhoseiny","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Elhoseiny, Mohamed","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5135349019","display_name":"David E. Keyes","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Keyes, David E.","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.4961000084877014,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.4961000084877014,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.18569999933242798,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T14347","display_name":"Big Data and Digital Economy","score":0.11580000072717667,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/efficient-energy-use","display_name":"Efficient energy use","score":0.532800018787384},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.5033000111579895},{"id":"https://openalex.org/keywords/scheduling","display_name":"Scheduling (production processes)","score":0.5012000203132629},{"id":"https://openalex.org/keywords/sequence","display_name":"Sequence (biology)","score":0.460999995470047},{"id":"https://openalex.org/keywords/training-set","display_name":"Training set","score":0.40540000796318054},{"id":"https://openalex.org/keywords/energy","display_name":"Energy (signal processing)","score":0.4041999876499176},{"id":"https://openalex.org/keywords/raw-data","display_name":"Raw data","score":0.3756999969482422},{"id":"https://openalex.org/keywords/frequency-scaling","display_name":"Frequency scaling","score":0.3723999857902527},{"id":"https://openalex.org/keywords/deep-learning","display_name":"Deep learning","score":0.3707999885082245}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8191999793052673},{"id":"https://openalex.org/C2742236","wikidata":"https://www.wikidata.org/wiki/Q924713","display_name":"Efficient energy use","level":2,"score":0.532800018787384},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.5220000147819519},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.5033000111579895},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.5012000203132629},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.460999995470047},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.42340001463890076},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4124999940395355},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.40540000796318054},{"id":"https://openalex.org/C186370098","wikidata":"https://www.wikidata.org/wiki/Q442787","display_name":"Energy (signal processing)","level":2,"score":0.4041999876499176},{"id":"https://openalex.org/C132964779","wikidata":"https://www.wikidata.org/wiki/Q2110223","display_name":"Raw data","level":2,"score":0.3756999969482422},{"id":"https://openalex.org/C157742956","wikidata":"https://www.wikidata.org/wiki/Q3237776","display_name":"Frequency scaling","level":3,"score":0.3723999857902527},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.3707999885082245},{"id":"https://openalex.org/C79974875","wikidata":"https://www.wikidata.org/wiki/Q483639","display_name":"Cloud computing","level":2,"score":0.3628999888896942},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.3612000048160553},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.35249999165534973},{"id":"https://openalex.org/C150899416","wikidata":"https://www.wikidata.org/wiki/Q1820378","display_name":"Transfer of learning","level":2,"score":0.3522000014781952},{"id":"https://openalex.org/C18555067","wikidata":"https://www.wikidata.org/wiki/Q8375051","display_name":"Joint (building)","level":2,"score":0.33559998869895935},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.33239999413490295},{"id":"https://openalex.org/C557945733","wikidata":"https://www.wikidata.org/wiki/Q389772","display_name":"Data transmission","level":2,"score":0.31049999594688416},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.30979999899864197},{"id":"https://openalex.org/C121687571","wikidata":"https://www.wikidata.org/wiki/Q4677630","display_name":"Activity recognition","level":2,"score":0.29350000619888306},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.2896000146865845},{"id":"https://openalex.org/C2780165032","wikidata":"https://www.wikidata.org/wiki/Q16869822","display_name":"Energy consumption","level":2,"score":0.28290000557899475},{"id":"https://openalex.org/C93996380","wikidata":"https://www.wikidata.org/wiki/Q44127","display_name":"Server","level":2,"score":0.2775000035762787},{"id":"https://openalex.org/C58489278","wikidata":"https://www.wikidata.org/wiki/Q1172284","display_name":"Data set","level":2,"score":0.2662000060081482},{"id":"https://openalex.org/C139807058","wikidata":"https://www.wikidata.org/wiki/Q352374","display_name":"Adaptation (eye)","level":2,"score":0.26080000400543213},{"id":"https://openalex.org/C68387754","wikidata":"https://www.wikidata.org/wiki/Q7271585","display_name":"Schedule","level":2,"score":0.2563000023365021},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.25600001215934753}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.01938","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.01938","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.01938","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.01938","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[{"score":0.9049112796783447,"id":"https://metadata.un.org/sdg/7","display_name":"Affordable and clean energy"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Multimodal":[0],"deep":[1],"learning":[2,6],"models":[3],"enable":[4],"joint":[5],"across":[7],"heterogeneous":[8,74,224],"data":[9,39,93,139],"sources,":[10],"including":[11],"text,":[12],"images,":[13],"and":[14,23,29,56,88,106,122,141,148,182,200,212],"video,":[15],"but":[16],"their":[17],"rapid":[18],"scaling":[19],"introduces":[20],"significant":[21],"memory":[22],"communication":[24,57],"bottlenecks.":[25],"As":[26],"model":[27],"sizes":[28],"sequence":[30,180,201],"lengths":[31],"increase,":[32],"training":[33,111,221],"performance":[34,107],"becomes":[35],"increasingly":[36],"impacted":[37],"by":[38,138],"movement":[40,140],"rather":[41,143],"than":[42,144],"computation.":[43],"Frameworks":[44],"such":[45,76],"as":[46,77],"DeepSpeed":[47],"mitigate":[48],"these":[49,60,162],"challenges":[50],"through":[51],"CPU":[52],"offloading,":[53],"activation":[54],"checkpointing,":[55],"optimizations.":[58],"However,":[59],"techniques":[61],"introduce":[62],"additional":[63],"system":[64],"activity,":[65],"which":[66],"may":[67],"affect":[68],"energy":[69,105,133,208],"efficiency.":[70],"Meanwhile,":[71],"tightly":[72],"integrated":[73],"architectures,":[75],"the":[78,117],"NVIDIA":[79],"Grace":[80],"Hopper":[81],"(GH200)":[82],"superchip,":[83],"provide":[84],"high-bandwidth":[85,126,194],"CPU-GPU":[86,127,195],"interconnects":[87,196],"unified":[89],"memory,":[90],"thereby":[91],"reducing":[92],"transfer":[94],"overhead.":[95],"In":[96],"this":[97],"work,":[98],"we":[99,164],"present":[100],"a":[101,166,204],"cross-layer":[102],"analysis":[103],"of":[104,168],"trade-offs":[108],"in":[109],"multimodal":[110,220],"on":[112,161,222],"GH200":[113],"systems,":[114],"explicitly":[115],"characterizing":[116],"interactions":[118],"between":[119],"application,":[120],"runtime,":[121],"hardware":[123],"layers.":[124],"Leveraging":[125],"interconnects,":[128],"our":[129],"results":[130,190],"show":[131],"that":[132,149,173,192],"efficiency":[134],"is":[135],"primarily":[136],"governed":[137],"overlap":[142],"raw":[145],"compute":[146],"utilization,":[147],"configurations":[150],"optimized":[151],"for":[152,158,171,218],"runtime":[153,210],"are":[154],"not":[155],"necessarily":[156],"optimal":[157],"energy.":[159],"Based":[160],"findings,":[163],"distill":[165],"set":[167],"actionable":[169],"guidelines":[170,217],"practitioners":[172],"demonstrate":[174,191],"how":[175],"to":[176,185],"balance":[177,206],"offloading":[178,198],"strategies,":[179],"parallelism,":[181,202],"hardware-aware":[183],"scheduling":[184],"achieve":[186],"energy-efficient":[187],"training.":[188],"Our":[189],"leveraging":[193],"enables":[197],"strategies":[199],"achieving":[203],"strong":[205],"among":[207],"efficiency,":[209],"performance,":[211],"computational":[213],"throughput,":[214],"providing":[215],"practical":[216],"efficient":[219],"modern":[223],"systems.":[225]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-06T00:00:00"}