{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:LEPUH3LRFIXTB5OR22JYS7OHJQ","short_pith_number":"pith:LEPUH3LR","schema_version":"1.0","canonical_sha256":"591f43ed712a2f30f5d1d693897dc74c245222d973015ad300b790f68fc4b306","source":{"kind":"arxiv","id":"2512.16056","version":2},"attestation_state":"computed","paper":{"title":"MultiPath Memory Access: Breaking Host-GPU Bandwidth Bottlenecks in LLM Services","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"Multipath Memory Access routes host-GPU copies over unused server links to raise bandwidth 4.6x.","cross_cats":["cs.NI","cs.PF"],"primary_cat":"cs.DC","authors_text":"Chengguang Xu, Daoping Zhang, Feiqiang Sun, Feng Jin, Guo Chen, Junjie Chen, Lingfeng Tang, Peihao Huang, Yuxin Chen","submitted_at":"2025-12-18T00:45:00Z","abstract_excerpt":"Host-GPU data movement has become a latency-critical bottleneck in LLM serving, surfacing in common paths such as model-weight movement and KV cache offload/fetch. Today, each host-GPU copy is effectively confined to the PCIe path of the target GPU, even though modern multi-GPU servers contain additional PCIe links on peer GPUs and high bandwidth GPU interconnects. This leaves substantial intra-server I/O capacity unused. To address this issue, we present Multipath Memory Access (MMA), a software-defined multipath memory access system for host--GPU data transfer. To the best of our knowledge, "},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2512.16056","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.DC","submitted_at":"2025-12-18T00:45:00Z","cross_cats_sorted":["cs.NI","cs.PF"],"title_canon_sha256":"bee3c6b1a798f883f47117bc5ddb7809fff2573bae5d05669f3b5ee9bca3ad8c","abstract_canon_sha256":"1c276ad2095a7b3e426b7bf9defb888ab9216db718737a0b3515565e9a1c3757"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T02:44:32.142287Z","signature_b64":"eR1GunCgyG48Ik8to4qCzaMebyhYJy9plyCUfBH6x9wVHDP8Ex3+Tw+Bg9i4NUmHSdb5gGYrGqm3R3QB7BXLAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"591f43ed712a2f30f5d1d693897dc74c245222d973015ad300b790f68fc4b306","last_reissued_at":"2026-05-18T02:44:32.141856Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T02:44:32.141856Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MultiPath Memory Access: Breaking Host-GPU Bandwidth Bottlenecks in LLM Services","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"Multipath Memory Access routes host-GPU copies over unused server links to raise bandwidth 4.6x.","cross_cats":["cs.NI","cs.PF"],"primary_cat":"cs.DC","authors_text":"Chengguang Xu, Daoping Zhang, Feiqiang Sun, Feng Jin, Guo Chen, Junjie Chen, Lingfeng Tang, Peihao Huang, Yuxin Chen","submitted_at":"2025-12-18T00:45:00Z","abstract_excerpt":"Host-GPU data movement has become a latency-critical bottleneck in LLM serving, surfacing in common paths such as model-weight movement and KV cache offload/fetch. Today, each host-GPU copy is effectively confined to the PCIe path of the target GPU, even though modern multi-GPU servers contain additional PCIe links on peer GPUs and high bandwidth GPU interconnects. This leaves substantial intra-server I/O capacity unused. To address this issue, we present Multipath Memory Access (MMA), a software-defined multipath memory access system for host--GPU data transfer. To the best of our knowledge, "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"On an 8-GPU NVIDIA H20 server, MMA achieves 245 GB/s peak host-to-GPU bandwidth, a 4.62x improvement over native CUDA copies, and reduces TTFT for KV cache fetching by 1.14-2.38x and model wake-up/switching latency by 1.12-2.48x.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the additional relay paths through peer GPUs and high-bandwidth interconnects can be utilized with low enough overhead and without breaking CUDA stream semantics or application correctness under real LLM workloads.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"MMA routes host-GPU transfers over multiple available paths to deliver 4.62x higher peak bandwidth and lower latencies in LLM serving without hardware or driver changes.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Multipath Memory Access routes host-GPU copies over unused server links to raise bandwidth 4.6x.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"7a9b940f8b6997b6775ec9caeea74b659146f4f8fdfb5dfdef8a46921900e8ff"},"source":{"id":"2512.16056","kind":"arxiv","version":2},"verdict":{"id":"aa1d4163-20f0-4562-a277-c1fdf22739c5","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T21:54:28.884047Z","strongest_claim":"On an 8-GPU NVIDIA H20 server, MMA achieves 245 GB/s peak host-to-GPU bandwidth, a 4.62x improvement over native CUDA copies, and reduces TTFT for KV cache fetching by 1.14-2.38x and model wake-up/switching latency by 1.12-2.48x.","one_line_summary":"MMA routes host-GPU transfers over multiple available paths to deliver 4.62x higher peak bandwidth and lower latencies in LLM serving without hardware or driver changes.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the additional relay paths through peer GPUs and high-bandwidth interconnects can be utilized with low enough overhead and without breaking CUDA stream semantics or application correctness under real LLM workloads.","pith_extraction_headline":"Multipath Memory Access routes host-GPU copies over unused server links to raise bandwidth 4.6x."},"references":{"count":52,"sample":[{"doi":"","year":null,"title":"MMA code base will be opensourced after paper ac- cepted","work_id":"62fb66e6-930b-43a1-be90-57ee36b0ba4a","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"Advanced Micro Devices","work_id":"9be9b912-df33-4373-a31f-11476d1f55b2","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Advanced Micro Devices","work_id":"89774536-8edc-4183-ac14-ed01cf54940b","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"Deepspeed-inference: enabling efficient in- ference of transformer models at unprecedented scale","work_id":"461bf7aa-e1b9-4309-a540-e2905d823c31","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Qwen Technical Report","work_id":"bb1fd52f-6b2f-437c-9516-37bdf6eb9be8","ref_index":5,"cited_arxiv_id":"2309.16609","is_internal_anchor":true}],"resolved_work":52,"snapshot_sha256":"41433ff6ca49d50f8face4bafba8465b901f73a62d0613b2a08589f02906c2a9","internal_anchors":3},"formal_canon":{"evidence_count":2,"snapshot_sha256":"887e95af28021118d353108b58ceb78b080a7a286f5a7d98d9413c2e06346528"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2512.16056","created_at":"2026-05-18T02:44:32.141920+00:00"},{"alias_kind":"arxiv_version","alias_value":"2512.16056v2","created_at":"2026-05-18T02:44:32.141920+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2512.16056","created_at":"2026-05-18T02:44:32.141920+00:00"},{"alias_kind":"pith_short_12","alias_value":"LEPUH3LRFIXT","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"LEPUH3LRFIXTB5OR","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"LEPUH3LR","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LEPUH3LRFIXTB5OR22JYS7OHJQ","json":"https://pith.science/pith/LEPUH3LRFIXTB5OR22JYS7OHJQ.json","graph_json":"https://pith.science/api/pith-number/LEPUH3LRFIXTB5OR22JYS7OHJQ/graph.json","events_json":"https://pith.science/api/pith-number/LEPUH3LRFIXTB5OR22JYS7OHJQ/events.json","paper":"https://pith.science/paper/LEPUH3LR"},"agent_actions":{"view_html":"https://pith.science/pith/LEPUH3LRFIXTB5OR22JYS7OHJQ","download_json":"https://pith.science/pith/LEPUH3LRFIXTB5OR22JYS7OHJQ.json","view_paper":"https://pith.science/paper/LEPUH3LR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2512.16056&json=true","fetch_graph":"https://pith.science/api/pith-number/LEPUH3LRFIXTB5OR22JYS7OHJQ/graph.json","fetch_events":"https://pith.science/api/pith-number/LEPUH3LRFIXTB5OR22JYS7OHJQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LEPUH3LRFIXTB5OR22JYS7OHJQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LEPUH3LRFIXTB5OR22JYS7OHJQ/action/storage_attestation","attest_author":"https://pith.science/pith/LEPUH3LRFIXTB5OR22JYS7OHJQ/action/author_attestation","sign_citation":"https://pith.science/pith/LEPUH3LRFIXTB5OR22JYS7OHJQ/action/citation_signature","submit_replication":"https://pith.science/pith/LEPUH3LRFIXTB5OR22JYS7OHJQ/action/replication_record"}},"created_at":"2026-05-18T02:44:32.141920+00:00","updated_at":"2026-05-18T02:44:32.141920+00:00"}