{"id":"https://openalex.org/W7225575601","doi":"https://doi.org/10.5281/zenodo.24053461","title":"Energy per FLOP Is Not Constant: Measured Dispersion Across FLOP-Matched Training Workloads","display_name":"Energy per FLOP Is Not Constant: Measured Dispersion Across FLOP-Matched Training Workloads","publication_year":2026,"publication_date":"2026-11-08","ids":{"openalex":"https://openalex.org/W7225575601","doi":"https://doi.org/10.5281/zenodo.24053461"},"language":"en","primary_location":{"id":"doi:10.5281/zenodo.24053461","doi":"https://doi.org/10.5281/zenodo.24053461","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24053461","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.5281/zenodo.24053461","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5119187960","display_name":"Sai Praneeth Reddy Dhadi","orcid":"https://orcid.org/0009-0002-1272-4666","observed_orcids":["https://orcid.org/0009-0002-1272-4666"]},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"DHADI, SAI PRANEETH REDDY","raw_affiliation_strings":["Atlas AI Labs"],"raw_orcid":"https://orcid.org/0009-0002-1272-4666","affiliations":[{"raw_affiliation_string":"Atlas AI Labs","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":["https://openalex.org/A5119187960"],"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":0.9702000021934509,"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":0.9702000021934509,"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/T14098","display_name":"Energy Efficiency in Computing","score":0.0203000009059906,"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/T12238","display_name":"Green IT and Sustainability","score":0.0052999998442828655,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/energy-measurement","display_name":"energy measurement","score":0.5630000233650208},{"id":"https://openalex.org/keywords/gpu-power-consumption","display_name":"GPU power consumption","score":0.6930000185966492},{"id":"https://openalex.org/keywords/on-chip-sensors","display_name":"on-chip sensors","score":0.6740000247955322},{"id":"https://openalex.org/keywords/dense-layers","display_name":"dense layers","score":0.5849999785423279},{"id":"https://openalex.org/keywords/convolution","display_name":"convolution","score":0.7149999737739563},{"id":"https://openalex.org/keywords/memory-traffic","display_name":"memory traffic","score":0.8130000233650208},{"id":"https://openalex.org/keywords/laptop-hardware","display_name":"laptop hardware","score":0.32199999690055847},{"id":"https://openalex.org/keywords/reproducibility","display_name":"reproducibility","score":0.6460000276565552}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6924999952316284},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.5800999999046326},{"id":"https://openalex.org/C3826847","wikidata":"https://www.wikidata.org/wiki/Q188768","display_name":"FLOPS","level":2,"score":0.5015000104904175},{"id":"https://openalex.org/C186370098","wikidata":"https://www.wikidata.org/wiki/Q442787","display_name":"Energy (signal processing)","level":2,"score":0.47029998898506165},{"id":"https://openalex.org/C2778565505","wikidata":"https://www.wikidata.org/wiki/Q2207566","display_name":"Spec#","level":2,"score":0.4481000006198883},{"id":"https://openalex.org/C2780008327","wikidata":"https://www.wikidata.org/wiki/Q3962","display_name":"Laptop","level":2,"score":0.42559999227523804},{"id":"https://openalex.org/C189783530","wikidata":"https://www.wikidata.org/wiki/Q352090","display_name":"CPU cache","level":3,"score":0.392300009727478},{"id":"https://openalex.org/C49154492","wikidata":"https://www.wikidata.org/wiki/Q5300","display_name":"Central processing unit","level":2,"score":0.3709999918937683},{"id":"https://openalex.org/C2742236","wikidata":"https://www.wikidata.org/wiki/Q924713","display_name":"Efficient energy use","level":2,"score":0.36820000410079956},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.36149999499320984},{"id":"https://openalex.org/C2778476105","wikidata":"https://www.wikidata.org/wiki/Q628539","display_name":"Workload","level":2,"score":0.36059999465942383},{"id":"https://openalex.org/C157742956","wikidata":"https://www.wikidata.org/wiki/Q3237776","display_name":"Frequency scaling","level":3,"score":0.3221000134944916},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.3172000050544739},{"id":"https://openalex.org/C2780165032","wikidata":"https://www.wikidata.org/wiki/Q16869822","display_name":"Energy consumption","level":2,"score":0.31610000133514404},{"id":"https://openalex.org/C156972235","wikidata":"https://www.wikidata.org/wiki/Q1443434","display_name":"Instructions per cycle","level":3,"score":0.3160000145435333},{"id":"https://openalex.org/C99821215","wikidata":"https://www.wikidata.org/wiki/Q1136583","display_name":"Swap (finance)","level":2,"score":0.31119999289512634},{"id":"https://openalex.org/C43364308","wikidata":"https://www.wikidata.org/wiki/Q8799","display_name":"Byte","level":2,"score":0.2921000123023987},{"id":"https://openalex.org/C178693496","wikidata":"https://www.wikidata.org/wiki/Q911691","display_name":"Clock rate","level":3,"score":0.29179999232292175},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.29030001163482666},{"id":"https://openalex.org/C2780385302","wikidata":"https://www.wikidata.org/wiki/Q367158","display_name":"Protocol (science)","level":3,"score":0.28290000557899475},{"id":"https://openalex.org/C68043766","wikidata":"https://www.wikidata.org/wiki/Q267416","display_name":"Static random-access memory","level":2,"score":0.2818000018596649},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.2728999853134155},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.2711000144481659},{"id":"https://openalex.org/C163258240","wikidata":"https://www.wikidata.org/wiki/Q25342","display_name":"Power (physics)","level":2,"score":0.27090001106262207},{"id":"https://openalex.org/C74193536","wikidata":"https://www.wikidata.org/wiki/Q574844","display_name":"Kernel (algebra)","level":2,"score":0.26989999413490295},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.2630000114440918},{"id":"https://openalex.org/C19499675","wikidata":"https://www.wikidata.org/wiki/Q232207","display_name":"Monte Carlo method","level":2,"score":0.2547999918460846}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.5281/zenodo.24053461","doi":"https://doi.org/10.5281/zenodo.24053461","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24053461","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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"},{"id":"doi:10.5281/zenodo.24053462","doi":"https://doi.org/10.5281/zenodo.24053462","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24053462","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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"}],"best_oa_location":{"id":"doi:10.5281/zenodo.24053461","doi":"https://doi.org/10.5281/zenodo.24053461","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24053461","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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":0.4000000059604645,"display_name":"Affordable and clean energy","id":"https://metadata.un.org/sdg/7"}],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2948541941","https://openalex.org/W2281172090","https://openalex.org/W7162626485","https://openalex.org/W7203680217","https://openalex.org/W6907500292","https://openalex.org/W3006451002","https://openalex.org/W2119722822","https://openalex.org/W2542146864","https://openalex.org/W7080443232","https://openalex.org/W7143981818"],"abstract_inverted_index":{"The":[0,94,228],"number":[1],"of":[2,31,135,190,198],"floating-point":[3],"operations":[4],"(FLOPs)":[5],"a":[6,13,39,78,107,133,219,232,237],"workload":[7],"performs":[8],"is":[9],"widely":[10],"used":[11],"as":[12],"proxy":[14,24],"for":[15,183],"the":[16,54,81,88,115,126,150,199,203,210,216,248,250,255,259],"energy":[17,68,121,163,174],"it":[18],"consumes.":[19],"This":[20],"article":[21],"tests":[22],"that":[23,74,258],"by":[25,106,132,223],"direct":[26],"physical":[27],"measurement.":[28,113],"Five":[29],"families":[30,131,167],"neural-network":[32],"training":[33],"step,":[34],"namely":[35],"dense":[36],"layers,":[37],"convolutions,":[38],"recurrent":[40],"cell,":[41],"embedding":[42],"lookups,":[43],"and":[44,61,87,100,104,128,149,186,212,246,254],"elementwise":[45],"updates":[46],"with":[47,175],"sparse":[48],"products,":[49],"were":[50,102],"each":[51],"matched":[52],"to":[53,141],"same":[55],"FLOP":[56,123],"count":[57],"within":[58],"two":[59],"percent":[60],"executed":[62],"on":[63,166,209,215,266],"one":[64],"laptop":[65],"computer":[66],"while":[67],"was":[69],"read":[70],"from":[71,169,244],"on-chip":[72],"sensors":[73],"had":[75],"first":[76],"passed":[77],"load-response":[79],"test:":[80],"CPU":[82,119,155,211],"package":[83,120,156],"through":[84,91],"Intel":[85],"RAPL,":[86],"GPU":[89,151,217],"board":[90,152],"NVIDIA":[92],"NVML.":[93],"hypotheses,":[95],"decision":[96],"rules,":[97],"sample":[98],"size":[99],"analysis":[101],"preregistered":[103,204],"fixed":[105,220],"cryptographic":[108],"digest":[109],"before":[110],"any":[111],"confirmatory":[112,238],"In":[114],"primary":[116],"session,":[117],"net":[118],"per":[122],"differed":[124],"between":[125],"most":[127],"least":[129],"efficient":[130],"factor":[134],"94.43":[136],"(95%":[137],"block-bootstrap":[138],"interval":[139],"78.06":[140],"119);":[142],"an":[143],"independent":[144],"earlier":[145],"session":[146],"gave":[147,153],"93.58,":[148],"26.05.":[154],"power":[157],"varied":[158],"little":[159],"across":[160],"families,":[161],"so":[162,257],"followed":[164],"runtime:":[165],"withheld":[168],"fitting,":[170],"runtime":[171],"alone":[172],"predicted":[173],"19.85%":[176],"mean":[177],"absolute":[178],"percentage":[179],"error,":[180],"against":[181],"63.72%":[182],"FLOPs":[184],"alone,":[185],"adding":[187],"shape-derived":[188],"estimates":[189],"memory":[191],"traffic":[192],"did":[193],"not":[194],"improve":[195],"prediction.":[196],"None":[197],"three":[200],"sessions":[201],"satisfied":[202],"stability":[205],"criterion,":[206],"because":[207],"contention":[208],"clock":[213],"transitions":[214],"moved":[218],"reference":[221],"kernel":[222],"more":[224],"than":[225,236],"ten":[226],"percent.":[227],"study":[229],"therefore":[230],"reports":[231],"replicated":[233],"measurement":[234,252],"rather":[235],"verdict.":[239],"It":[240],"documents":[241],"every":[242],"deviation":[243],"protocol":[245],"releases":[247],"instrument,":[249],"raw":[251],"windows":[253],"analysis,":[256],"result":[260],"can":[261],"be":[262],"reproduced":[263],"or":[264],"refuted":[265],"other":[267],"hardware.":[268]},"counts_by_year":[],"updated_date":"2026-11-10T11:39:43.021295","created_date":"2026-11-09T00:00:00"}