{"id":"https://openalex.org/W7165110164","doi":"https://doi.org/10.1145/3787109.3815290","title":"MRAM-MoE: Efficient Inference of Mixture-of-Experts LLMs with MRAM Chiplet-based Accelerators","display_name":"MRAM-MoE: Efficient Inference of Mixture-of-Experts LLMs with MRAM Chiplet-based Accelerators","publication_year":2026,"publication_date":"2026-06-18","ids":{"openalex":"https://openalex.org/W7165110164","doi":"https://doi.org/10.1145/3787109.3815290"},"language":null,"primary_location":{"id":"doi:10.1145/3787109.3815290","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3787109.3815290","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the Great Lakes Symposium on VLSI 2026","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://doi.org/10.1145/3787109.3815290","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5044295544","display_name":"M Ashiqur Rahman","orcid":null},"institutions":[{"id":"https://openalex.org/I82497590","display_name":"Auburn University","ror":"https://ror.org/02v80fc35","country_code":"US","type":"education","lineage":["https://openalex.org/I82497590"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Md Tanjimur Rahman","raw_affiliation_strings":["Electrical and computer Engineering, Auburn University, Auburn, AL, Auburn, USA"],"raw_orcid":"https://orcid.org/0009-0007-9107-0336","affiliations":[{"raw_affiliation_string":"Electrical and computer Engineering, Auburn University, Auburn, AL, Auburn, USA","institution_ids":["https://openalex.org/I82497590"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5108983614","display_name":"Md Asef","orcid":null},"institutions":[{"id":"https://openalex.org/I82497590","display_name":"Auburn University","ror":"https://ror.org/02v80fc35","country_code":"US","type":"education","lineage":["https://openalex.org/I82497590"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Md Asef","raw_affiliation_strings":["Electrical and computer Engineering, Auburn University, Auburn, AL, Auburn, USA"],"raw_orcid":"https://orcid.org/0009-0005-7325-3134","affiliations":[{"raw_affiliation_string":"Electrical and computer Engineering, Auburn University, Auburn, AL, Auburn, USA","institution_ids":["https://openalex.org/I82497590"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5036369307","display_name":"Mehdi Sadi","orcid":"https://orcid.org/0000-0002-0468-7810"},"institutions":[{"id":"https://openalex.org/I82497590","display_name":"Auburn University","ror":"https://ror.org/02v80fc35","country_code":"US","type":"education","lineage":["https://openalex.org/I82497590"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Mehdi Sadi","raw_affiliation_strings":["Electrical and Computer Engineering, Auburn University, Auburn, AL, Auburn, AL, USA"],"raw_orcid":"https://orcid.org/0000-0002-0468-7810","affiliations":[{"raw_affiliation_string":"Electrical and Computer Engineering, Auburn University, Auburn, AL, Auburn, AL, USA","institution_ids":["https://openalex.org/I82497590"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I82497590"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.67494357,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"502","last_page":"508"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.4339999854564667,"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":0.4339999854564667,"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/T14347","display_name":"Big Data and Digital Economy","score":0.10239999741315842,"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"}},{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.09700000286102295,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/magnetoresistive-random-access-memory","display_name":"Magnetoresistive random-access memory","score":0.920199990272522},{"id":"https://openalex.org/keywords/latency","display_name":"Latency (audio)","score":0.44929999113082886},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.4422000050544739},{"id":"https://openalex.org/keywords/memory-bandwidth","display_name":"Memory bandwidth","score":0.42649999260902405},{"id":"https://openalex.org/keywords/reduction","display_name":"Reduction (mathematics)","score":0.4196000099182129},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.3856000006198883},{"id":"https://openalex.org/keywords/simd","display_name":"SIMD","score":0.3804999887943268},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.367900013923645}],"concepts":[{"id":"https://openalex.org/C46891859","wikidata":"https://www.wikidata.org/wiki/Q1061546","display_name":"Magnetoresistive random-access memory","level":3,"score":0.920199990272522},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7184000015258789},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.4717000126838684},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.44929999113082886},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.4422000050544739},{"id":"https://openalex.org/C188045654","wikidata":"https://www.wikidata.org/wiki/Q17148339","display_name":"Memory bandwidth","level":2,"score":0.42649999260902405},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.4196000099182129},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.3856000006198883},{"id":"https://openalex.org/C150552126","wikidata":"https://www.wikidata.org/wiki/Q339387","display_name":"SIMD","level":2,"score":0.3804999887943268},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.367900013923645},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.33799999952316284},{"id":"https://openalex.org/C2776257435","wikidata":"https://www.wikidata.org/wiki/Q1576430","display_name":"Bandwidth (computing)","level":2,"score":0.33079999685287476},{"id":"https://openalex.org/C107598950","wikidata":"https://www.wikidata.org/wiki/Q259864","display_name":"Microarchitecture","level":2,"score":0.3199999928474426},{"id":"https://openalex.org/C98986596","wikidata":"https://www.wikidata.org/wiki/Q1143031","display_name":"Semiconductor memory","level":2,"score":0.30140000581741333},{"id":"https://openalex.org/C123657996","wikidata":"https://www.wikidata.org/wiki/Q12271","display_name":"Architecture","level":2,"score":0.2978000044822693},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.2890999913215637},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.2863999903202057},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.28610000014305115},{"id":"https://openalex.org/C46362747","wikidata":"https://www.wikidata.org/wiki/Q173431","display_name":"CMOS","level":2,"score":0.2808000147342682},{"id":"https://openalex.org/C2779602883","wikidata":"https://www.wikidata.org/wiki/Q15544750","display_name":"Memory architecture","level":2,"score":0.27950000762939453},{"id":"https://openalex.org/C189930140","wikidata":"https://www.wikidata.org/wiki/Q1112878","display_name":"CAS latency","level":4,"score":0.2766000032424927},{"id":"https://openalex.org/C93389723","wikidata":"https://www.wikidata.org/wiki/Q7247313","display_name":"Process variation","level":3,"score":0.2759999930858612},{"id":"https://openalex.org/C163258240","wikidata":"https://www.wikidata.org/wiki/Q25342","display_name":"Power (physics)","level":2,"score":0.26660001277923584},{"id":"https://openalex.org/C153247305","wikidata":"https://www.wikidata.org/wiki/Q835713","display_name":"Memory address","level":3,"score":0.2590000033378601},{"id":"https://openalex.org/C182365436","wikidata":"https://www.wikidata.org/wiki/Q50701","display_name":"Variable (mathematics)","level":2,"score":0.2563999891281128}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3787109.3815290","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3787109.3815290","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the Great Lakes Symposium on VLSI 2026","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1145/3787109.3815290","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3787109.3815290","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the Great Lakes Symposium on VLSI 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[{"score":0.5140886306762695,"id":"https://metadata.un.org/sdg/7","display_name":"Affordable and clean energy"}],"awards":[{"id":"https://openalex.org/G734514311","display_name":"CAREER: Optimizing the Next Frontiers of Chiplet-based Designs in Advanced Packaging","funder_award_id":"2441694","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":19,"referenced_works":["https://openalex.org/W1975904824","https://openalex.org/W2042076387","https://openalex.org/W2112388836","https://openalex.org/W3005785620","https://openalex.org/W3196387322","https://openalex.org/W4249654426","https://openalex.org/W4280564961","https://openalex.org/W4388739149","https://openalex.org/W4393592481","https://openalex.org/W4399485941","https://openalex.org/W4402436905","https://openalex.org/W4406178148","https://openalex.org/W4408894550","https://openalex.org/W4414198165","https://openalex.org/W4414198920","https://openalex.org/W4414974080","https://openalex.org/W4416004263","https://openalex.org/W4416429485","https://openalex.org/W4416429515"],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2],"(LLMs)":[3],"require":[4],"substantial":[5],"compute":[6,82],"and":[7,15,34,39,93,101,135,138,148,159,190],"memory":[8,13,79,125],"resources":[9],"during":[10],"inference,":[11],"making":[12],"bandwidth":[14],"capacity":[16],"key":[17],"bottlenecks.":[18],"Mixture-of-Experts":[19],"(MoE)":[20],"architectures":[21],"reduce":[22],"computational":[23],"cost":[24],"through":[25],"sparse":[26],"expert":[27,37],"activation":[28],"but":[29],"introduce":[30],"challenges":[31],"in":[32,156,162],"storing":[33],"accessing":[35],"large":[36],"weights":[38],"Key\u2013Value":[40],"(KV)":[41],"caches,":[42],"demanding":[43],"custom":[44],"hardware":[45],"accelerators":[46],"for":[47,58,104,128,142,169],"efficient":[48],"inference.":[49,61],"This":[50],"work":[51],"proposes":[52],"MRAM-MOE,":[53],"a":[54,123,153],"MRAM":[55,70,145,188],"chiplet-based":[56],"architecture":[57,64],"accelerating":[59],"MoE":[60,96,129],"The":[62],"MRAM-MOE":[63],"uses":[65],"variable":[66],"retention":[67],"banks":[68],"of":[69,95],"cells":[71],"as":[72,75,77],"stacked":[73,187],"chiplets,":[74],"well":[76],"on-chip":[78,194],"on":[80,109],"SIMD":[81],"dies.":[83],"A":[84],"comprehensive":[85],"performance":[86],"modeling":[87],"framework":[88],"analyzes":[89],"the":[90,110,113,143,170,179],"FLOPs,":[91],"throughput,":[92],"latency":[94,137],"operations":[97],"across":[98],"both":[99],"prefill":[100],"decode":[102],"phases":[103],"five":[105],"representative":[106],"models.":[107],"Based":[108],"analyzed":[111],"latencies,":[112],"STT-MRAM":[114,172],"thermal":[115],"stability":[116],"factor":[117],"(\u0394)":[118],"is":[119],"tuned":[120],"to":[121,178],"implement":[122],"multi-retention":[124],"model":[126,173],"optimized":[127,171],"access":[130],"patterns.":[131],"Reliability,":[132],"area,":[133],"capacity,":[134],"read/write":[136],"energy":[139],"are":[140],"evaluated":[141],"customized":[144],"under":[146],"process":[147],"temperature":[149],"variations.":[150],"Results":[151],"illustrate":[152],"66.7%":[154],"reduction":[155,161],"dynamic":[157],"power":[158],"95.3%":[160],"die":[163],"area":[164],"at":[165],"14":[166],"nm":[167],"technology":[168],"(\u0394":[174],"=":[175],"27.3)":[176],"compared":[177],"conventional":[180],"SRAM,":[181],"while":[182],"supporting":[183],"two":[184,191],"1":[185],"GB":[186],"chiplets":[189],"70":[192],"MB":[193],"MRAMs.":[195]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-06-19T00:00:00"}