{"id":"https://openalex.org/W4409262453","doi":"https://doi.org/10.1109/tc.2025.3558579","title":"BE-NPU: A Bandwidth-Efficient Neural Processing Unit With Adaptive Processing Schemes for Reduced Off-Chip Bandwidth Demand","display_name":"BE-NPU: A Bandwidth-Efficient Neural Processing Unit With Adaptive Processing Schemes for Reduced Off-Chip Bandwidth Demand","publication_year":2025,"publication_date":"2025-04-08","ids":{"openalex":"https://openalex.org/W4409262453","doi":"https://doi.org/10.1109/tc.2025.3558579"},"language":"en","primary_location":{"id":"doi:10.1109/tc.2025.3558579","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tc.2025.3558579","pdf_url":null,"source":{"id":"https://openalex.org/S157670870","display_name":"IEEE Transactions on Computers","issn_l":"0018-9340","issn":["0018-9340","1557-9956","2326-3814"],"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 Transactions on Computers","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/A5080870367","display_name":"Yichuan Bai","orcid":"https://orcid.org/0000-0001-8895-1780"},"institutions":[{"id":"https://openalex.org/I881766915","display_name":"Nanjing University","ror":"https://ror.org/01rxvg760","country_code":"CN","type":"education","lineage":["https://openalex.org/I881766915"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Yichuan Bai","raw_affiliation_strings":["School of Electronic Science and Engineering, Nanjing University, Nanjing, China","School of Electronic Science and Engineering, Nanjing University, China"],"affiliations":[{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University, Nanjing, China","institution_ids":["https://openalex.org/I881766915"]},{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University, China","institution_ids":["https://openalex.org/I881766915"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100614073","display_name":"Xiaopeng Zhang","orcid":"https://orcid.org/0000-0003-0854-4742"},"institutions":[{"id":"https://openalex.org/I881766915","display_name":"Nanjing University","ror":"https://ror.org/01rxvg760","country_code":"CN","type":"education","lineage":["https://openalex.org/I881766915"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xiaopeng Zhang","raw_affiliation_strings":["School of Electronic Science and Engineering, Nanjing University, Nanjing, China","School of Electronic Science and Engineering, Nanjing University, China"],"affiliations":[{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University, Nanjing, China","institution_ids":["https://openalex.org/I881766915"]},{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University, China","institution_ids":["https://openalex.org/I881766915"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100422786","display_name":"Qian Wang","orcid":"https://orcid.org/0000-0002-2149-384X"},"institutions":[{"id":"https://openalex.org/I881766915","display_name":"Nanjing University","ror":"https://ror.org/01rxvg760","country_code":"CN","type":"education","lineage":["https://openalex.org/I881766915"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Qian Wang","raw_affiliation_strings":["School of Electronic Science and Engineering, Nanjing University, Nanjing, China","School of Electronic Science and Engineering, Nanjing University, China"],"affiliations":[{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University, Nanjing, China","institution_ids":["https://openalex.org/I881766915"]},{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University, China","institution_ids":["https://openalex.org/I881766915"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5050984267","display_name":"Yaqing Li","orcid":"https://orcid.org/0000-0002-8858-4666"},"institutions":[{"id":"https://openalex.org/I881766915","display_name":"Nanjing University","ror":"https://ror.org/01rxvg760","country_code":"CN","type":"education","lineage":["https://openalex.org/I881766915"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yaqing Li","raw_affiliation_strings":["School of Electronic Science and Engineering, Nanjing University, Nanjing, China","School of Electronic Science and Engineering, Nanjing University, China"],"affiliations":[{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University, Nanjing, China","institution_ids":["https://openalex.org/I881766915"]},{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University, China","institution_ids":["https://openalex.org/I881766915"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5084816122","display_name":"Yuan Du","orcid":"https://orcid.org/0000-0002-5316-619X"},"institutions":[{"id":"https://openalex.org/I881766915","display_name":"Nanjing University","ror":"https://ror.org/01rxvg760","country_code":"CN","type":"education","lineage":["https://openalex.org/I881766915"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yuan Du","raw_affiliation_strings":["School of Electronic Science and Engineering, Nanjing University, Nanjing, China","School of Electronic Science and Engineering, Nanjing University, China"],"affiliations":[{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University, Nanjing, China","institution_ids":["https://openalex.org/I881766915"]},{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University, China","institution_ids":["https://openalex.org/I881766915"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5039665206","display_name":"Li Du","orcid":"https://orcid.org/0000-0003-2687-6978"},"institutions":[{"id":"https://openalex.org/I881766915","display_name":"Nanjing University","ror":"https://ror.org/01rxvg760","country_code":"CN","type":"education","lineage":["https://openalex.org/I881766915"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Li Du","raw_affiliation_strings":["School of Electronic Science and Engineering, Nanjing University, Nanjing, China","School of Electronic Science and Engineering, Nanjing University, China"],"affiliations":[{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University, Nanjing, China","institution_ids":["https://openalex.org/I881766915"]},{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University, China","institution_ids":["https://openalex.org/I881766915"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":6,"corresponding_author_ids":["https://openalex.org/A5080870367"],"corresponding_institution_ids":["https://openalex.org/I881766915"],"apc_list":null,"apc_paid":null,"fwci":4.5405,"has_fulltext":false,"cited_by_count":2,"citation_normalized_percentile":{"value":0.92779633,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":98,"max":99},"biblio":{"volume":"74","issue":"7","first_page":"2376","last_page":"2388"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10904","display_name":"Embedded Systems Design Techniques","score":0.9107999801635742,"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/T10904","display_name":"Embedded Systems Design Techniques","score":0.9107999801635742,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/nat","display_name":"Nat","score":0.45828312635421753},{"id":"https://openalex.org/keywords/physics","display_name":"Physics","score":0.44972705841064453},{"id":"https://openalex.org/keywords/bandwidth","display_name":"Bandwidth (computing)","score":0.43218743801116943},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.2992187738418579},{"id":"https://openalex.org/keywords/telecommunications","display_name":"Telecommunications","score":0.2717134356498718},{"id":"https://openalex.org/keywords/quantum-mechanics","display_name":"Quantum mechanics","score":0.06726992130279541}],"concepts":[{"id":"https://openalex.org/C182516595","wikidata":"https://www.wikidata.org/wiki/Q376660","display_name":"Nat","level":2,"score":0.45828312635421753},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.44972705841064453},{"id":"https://openalex.org/C2776257435","wikidata":"https://www.wikidata.org/wiki/Q1576430","display_name":"Bandwidth (computing)","level":2,"score":0.43218743801116943},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.2992187738418579},{"id":"https://openalex.org/C76155785","wikidata":"https://www.wikidata.org/wiki/Q418","display_name":"Telecommunications","level":1,"score":0.2717134356498718},{"id":"https://openalex.org/C62520636","wikidata":"https://www.wikidata.org/wiki/Q944","display_name":"Quantum mechanics","level":1,"score":0.06726992130279541}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/tc.2025.3558579","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tc.2025.3558579","pdf_url":null,"source":{"id":"https://openalex.org/S157670870","display_name":"IEEE Transactions on Computers","issn_l":"0018-9340","issn":["0018-9340","1557-9956","2326-3814"],"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 Transactions on Computers","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":31,"referenced_works":["https://openalex.org/W1772650917","https://openalex.org/W2194775991","https://openalex.org/W2289252105","https://openalex.org/W2508457857","https://openalex.org/W2734572653","https://openalex.org/W2764207251","https://openalex.org/W2773339846","https://openalex.org/W2898136110","https://openalex.org/W2935849356","https://openalex.org/W2963163009","https://openalex.org/W2996367417","https://openalex.org/W3004127905","https://openalex.org/W3013082411","https://openalex.org/W3084937280","https://openalex.org/W3095870016","https://openalex.org/W3157657667","https://openalex.org/W3160243752","https://openalex.org/W3199252062","https://openalex.org/W3205656656","https://openalex.org/W3210597831","https://openalex.org/W3216623900","https://openalex.org/W4205884968","https://openalex.org/W4294690581","https://openalex.org/W4311410588","https://openalex.org/W4312646647","https://openalex.org/W4360831791","https://openalex.org/W4388894005","https://openalex.org/W4392114295","https://openalex.org/W4400811482","https://openalex.org/W6793164127","https://openalex.org/W6847370542"],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2935759653","https://openalex.org/W3105167352","https://openalex.org/W54078636","https://openalex.org/W2954470139","https://openalex.org/W2383430300","https://openalex.org/W2355747264","https://openalex.org/W1501425562","https://openalex.org/W2902782467","https://openalex.org/W3084825885"],"abstract_inverted_index":{"Existing":[0],"neural":[1,18],"processing":[2,130,144],"units":[3],"(NPUs)":[4],"mainly":[5],"focus":[6],"on":[7],"the":[8,47,52,94,100,114,134,153,169,182],"optimized":[9],"multiply-accumulate":[10],"(MAC)":[11],"arrays":[12],"for":[13,40,81,104],"efficient":[14],"inference":[15,138],"of":[16,46,55,65,107,118,139,206,211],"convolutional":[17],"networks":[19],"(CNNs).":[20],"However,":[21],"off-chip":[22,36,66],"data":[23,67],"transmission":[24,68],"usually":[25],"keeps":[26],"NPUs":[27,82],"waiting":[28],"during":[29,137],"CNN":[30],"inference,":[31],"causing":[32],"up":[33],"to":[34,71,83,112,132,167,192],"38.4GB/s":[35],"bandwidth":[37,53],"(OCB)":[38],"demand":[39,117,136],"mobile":[41],"AI":[42],"devices.":[43],"And":[44],"none":[45],"previous":[48],"benchmarks":[49],"quantitatively":[50],"evaluate":[51],"efficiency":[54],"different":[56,72,85,119,140],"NPU":[57,120,124,172,196],"architectures.":[58,121],"In":[59],"addition,":[60],"CNNs":[61,86],"exhibit":[62],"distinct":[63],"characteristics":[64],"when":[69],"applied":[70],"fields,":[73],"and":[74,149,163,175,186,208],"it":[75],"has":[76],"become":[77],"a":[78],"challenging":[79],"task":[80],"support":[84],"efficiently":[87],"with":[88,128,199,218,230],"reasonable":[89],"OCB":[90,116,135],"demand.":[91],"To":[92],"address":[93],"aforementioned":[95],"issues,":[96],"this":[97],"paper":[98],"proposes":[99],"Bandwidth-Peak":[101],"Performance":[102],"Ratio":[103],"n":[105],"percentages":[106],"ideal":[108],"frame":[109],"rate":[110],"(BPPR-n%)":[111],"demonstrate":[113],"normalized":[115],"A":[122],"bandwidth-efficient":[123],"(BE-NPU)":[125],"is":[126,158,216],"introduced":[127,159],"adaptive":[129,143],"schemes":[131,145,166],"reduce":[133],"CNNs.":[141],"The":[142,213],"include":[146],"both":[147],"instruction-level":[148,154],"thread-level":[150,183],"schemes.":[151],"For":[152,181],"scheme,":[155,184],"decoupled":[156],"execute/access":[157],"into":[160],"depth-first":[161],"(DF)":[162],"layer-first":[164],"(LF)":[165],"improve":[168,194],"concurrency":[170],"between":[171],"calculation":[173],"(CAL)":[174],"direct":[176],"memory":[177],"access":[178],"(DMA)":[179],"instructions.":[180],"DF":[185],"LF":[187],"threads":[188],"are":[189],"hybridly":[190],"processed":[191],"further":[193],"overall":[195],"efficiency.":[197],"Compared":[198],"state-of-the-art":[200],"works,":[201],"BE-NPU":[202,223],"achieves":[203],"48.1%~80.6%":[204],"reduction":[205,210],"BPPR-80%":[207],"67.0%~95.1%":[209],"BPPR-95%.":[212],"proposed":[214],"architecture":[215],"synthesized":[217],"TSMC":[219],"28nm":[220],"technology":[221],"node.":[222],"utilizes":[224],"14.3%":[225],"additional":[226],"logic":[227],"gates":[228],"compared":[229],"baseline":[231],"implementation.":[232]},"counts_by_year":[{"year":2026,"cited_by_count":2}],"updated_date":"2026-03-27T05:58:40.876381","created_date":"2025-04-09T00:00:00"}
