{"id":"https://openalex.org/W7134111884","doi":"https://doi.org/10.1109/jssc.2026.3663603","title":"Adelia: A 4-nm LLM Processing Unit With Streamlined Dataflow and Dual-Mode Parallelism for Maximizing Hardware Efficiency","display_name":"Adelia: A 4-nm LLM Processing Unit With Streamlined Dataflow and Dual-Mode Parallelism for Maximizing Hardware Efficiency","publication_year":2026,"publication_date":"2026-03-06","ids":{"openalex":"https://openalex.org/W7134111884","doi":"https://doi.org/10.1109/jssc.2026.3663603"},"language":null,"primary_location":{"id":"doi:10.1109/jssc.2026.3663603","is_oa":false,"landing_page_url":"https://doi.org/10.1109/jssc.2026.3663603","pdf_url":null,"source":{"id":"https://openalex.org/S83637746","display_name":"IEEE Journal of Solid-State Circuits","issn_l":"0018-9200","issn":["0018-9200","1558-173X"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Journal of Solid-State Circuits","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5047063086","display_name":"Sukbin Lim","orcid":"https://orcid.org/0000-0001-5743-4852"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Sukbin Lim","raw_affiliation_strings":["HyperAccel, Seoul, South Korea"],"raw_orcid":"https://orcid.org/0000-0001-5743-4852","affiliations":[{"raw_affiliation_string":"HyperAccel, Seoul, South Korea","institution_ids":[]}]},{"author_position":"middle","author":{"id":null,"display_name":"Jung-Hoon Kim","orcid":"https://orcid.org/0000-0003-4702-6489"},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Jung-Hoon Kim","raw_affiliation_strings":["HyperAccel, Seoul, South Korea","KAIST, Daejeon, South Korea"],"raw_orcid":"https://orcid.org/0000-0003-4702-6489","affiliations":[{"raw_affiliation_string":"HyperAccel, Seoul, South Korea","institution_ids":[]},{"raw_affiliation_string":"KAIST, Daejeon, South Korea","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5030449068","display_name":"Seungjae Moon","orcid":"https://orcid.org/0009-0002-5924-7000"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Seungjae Moon","raw_affiliation_strings":["HyperAccel, Seoul, South Korea"],"raw_orcid":"https://orcid.org/0009-0002-5924-7000","affiliations":[{"raw_affiliation_string":"HyperAccel, Seoul, South Korea","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5092947137","display_name":"Junseo Cha","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Junseo Cha","raw_affiliation_strings":["HyperAccel, Seoul, South Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"HyperAccel, Seoul, South Korea","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128296045","display_name":"Dongjin Seo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dongjin Seo","raw_affiliation_strings":["HyperAccel, Seoul, South Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"HyperAccel, Seoul, South Korea","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5041027404","display_name":"Junsu Kim","orcid":"https://orcid.org/0000-0001-9583-7978"},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Jongho Kim","raw_affiliation_strings":["HyperAccel, Seoul, South Korea","KAIST, Daejeon, South Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"HyperAccel, Seoul, South Korea","institution_ids":[]},{"raw_affiliation_string":"KAIST, Daejeon, South Korea","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5046228864","display_name":"Hunjong Lee","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hunjong Lee","raw_affiliation_strings":["HyperAccel, Seoul, South Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"HyperAccel, Seoul, South Korea","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100750649","display_name":"Jinwon Lee","orcid":"https://orcid.org/0000-0003-4810-1014"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jinwon Lee","raw_affiliation_strings":["HyperAccel, Seoul, South Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"HyperAccel, Seoul, South Korea","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5041027404","display_name":"Junsu Kim","orcid":"https://orcid.org/0000-0001-9583-7978"},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Joo-Young Kim","raw_affiliation_strings":["HyperAccel, Seoul, South Korea","KAIST, Daejeon, South Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"HyperAccel, Seoul, South Korea","institution_ids":[]},{"raw_affiliation_string":"KAIST, Daejeon, South Korea","institution_ids":["https://openalex.org/I157485424"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":9,"corresponding_author_ids":["https://openalex.org/A5047063086"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.37802143,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"61","issue":"4","first_page":"1513","last_page":"1525"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.565500020980835,"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":0.565500020980835,"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/T10904","display_name":"Embedded Systems Design Techniques","score":0.05849999934434891,"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/T11338","display_name":"Advancements in Photolithography Techniques","score":0.048500001430511475,"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/parallelism","display_name":"Parallelism (grammar)","score":0.5630000233650208},{"id":"https://openalex.org/keywords/dataflow","display_name":"Dataflow","score":0.5087000131607056},{"id":"https://openalex.org/keywords/unit","display_name":"Unit (ring theory)","score":0.38690000772476196},{"id":"https://openalex.org/keywords/parallel-processing","display_name":"Parallel processing","score":0.35429999232292175},{"id":"https://openalex.org/keywords/multi-core-processor","display_name":"Multi-core processor","score":0.3034999966621399}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.728600025177002},{"id":"https://openalex.org/C2781172179","wikidata":"https://www.wikidata.org/wiki/Q853109","display_name":"Parallelism (grammar)","level":2,"score":0.5630000233650208},{"id":"https://openalex.org/C96324660","wikidata":"https://www.wikidata.org/wiki/Q205446","display_name":"Dataflow","level":2,"score":0.5087000131607056},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.4318000078201294},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.39969998598098755},{"id":"https://openalex.org/C122637931","wikidata":"https://www.wikidata.org/wiki/Q118084","display_name":"Unit (ring theory)","level":2,"score":0.38690000772476196},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.376800000667572},{"id":"https://openalex.org/C106515295","wikidata":"https://www.wikidata.org/wiki/Q26806595","display_name":"Parallel processing","level":2,"score":0.35429999232292175},{"id":"https://openalex.org/C78766204","wikidata":"https://www.wikidata.org/wiki/Q555032","display_name":"Multi-core processor","level":2,"score":0.3034999966621399},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2955999970436096},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.2583000063896179}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/jssc.2026.3663603","is_oa":false,"landing_page_url":"https://doi.org/10.1109/jssc.2026.3663603","pdf_url":null,"source":{"id":"https://openalex.org/S83637746","display_name":"IEEE Journal of Solid-State Circuits","issn_l":"0018-9200","issn":["0018-9200","1558-173X"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Journal of Solid-State Circuits","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[{"score":0.7297838926315308,"display_name":"Affordable and clean energy","id":"https://metadata.un.org/sdg/7"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"The":[0,23],"proliferation":[1],"of":[2,25,44,54,59,99,219,228,239],"large":[3],"language":[4],"models":[5,10,28],"(LLMs)":[6],"as":[7,80],"cross-domain":[8],"foundation":[9],"is":[11],"fueled":[12],"by":[13,118,158,179],"aggressive":[14],"scaling":[15],"in":[16,101,210],"both":[17,121,144],"parameter":[18,161],"counts":[19],"and":[20,37,41,64,82,123,152,170,190,204,231],"inference-time":[21],"computation.":[22],"emergence":[24],"sophisticated":[26],"reasoning":[27],"further":[29],"accelerates":[30],"this":[31,116],"trend,":[32],"demanding":[33],"longer":[34],"context":[35],"windows":[36],"escalating":[38],"the":[39,51,102,197],"computational":[40],"memory":[42,122,139],"burdens":[43],"inference.":[45,208],"A":[46],"fundamental":[47],"challenge":[48,117],"arises":[49],"from":[50],"bimodal":[52,198],"nature":[53],"LLM":[55,110],"inference,":[56],"which":[57,87,165],"consists":[58],"a":[60,65,72,108,128,159,216,237],"compute-bound":[61],"prefill":[62],"phase":[63],"memory-bound":[66],"decode":[67],"phase.":[68],"This":[69,104],"duality":[70],"creates":[71],"significant":[73],"performance":[74,203],"bottleneck":[75],"for":[76,91,120,174],"conventional":[77],"architectures":[78],"such":[79],"GPUs":[81],"neural":[83],"processing":[84],"units":[85],"(NPUs),":[86],"are":[88],"typically":[89],"optimized":[90],"only":[92],"one":[93],"phase,":[94],"leading":[95],"to":[96,114],"severe":[97],"underutilization":[98],"resources":[100],"other.":[103],"article":[105],"introduces":[106],"Adelia,":[107],"novel":[109],"inference":[111,145],"accelerator":[112],"designed":[113],"resolve":[115],"co-optimizing":[119],"compute":[124,134],"efficiency.":[125],"Adelia":[126,200,214],"features":[127,149],"streamlined":[129],"dataflow":[130],"that":[131,185],"aligns":[132],"sustained":[133],"throughput":[135,218],"with":[136,224,236],"available":[137],"external":[138],"bandwidth,":[140],"maximizing":[141],"utilization":[142],"across":[143],"phases.":[146],"Its":[147],"architecture":[148],"powerful":[150],"matrix":[151],"vector":[153],"execution":[154],"engines":[155],"(VXEs)":[156],"fed":[157],"systolic":[160],"path":[162],"(SPP)":[163],"interconnect,":[164],"efficiently":[166],"multicasts":[167],"model":[168],"parameters":[169],"key\u2013value":[171],"(KV)":[172],"caches":[173],"spatial":[175],"reuse,":[176],"all":[177],"orchestrated":[178],"an":[180,225],"RISC":[181],"control":[182],"processor":[183],"(RCP)":[184],"provides":[186],"dynamic":[187],"instruction":[188],"dispatch":[189],"enables":[191],"dual-mode":[192],"parallelism.":[193],"By":[194],"holistically":[195],"addressing":[196],"bottleneck,":[199],"demonstrates":[201],"high":[202,217],"scalable":[205],"efficiency":[206],"during":[207],"Fabricated":[209],"Samsung\u2019s":[211],"4-nm":[212],"process,":[213],"achieves":[215],"512.6":[220],"Tokens/s":[221,233],"on":[222,234],"Phi-3-mini":[223],"end-to-end":[226],"runtime":[227,238],"5.993":[229],"s":[230],"65.72":[232],"LLaMA-30B":[235],"46.741":[240],"s.":[241]},"counts_by_year":[],"updated_date":"2026-03-28T06:11:35.319607","created_date":"2026-03-08T00:00:00"}
