{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:NFLM4SX5IA765JME7F77P47YAX","short_pith_number":"pith:NFLM4SX5","schema_version":"1.0","canonical_sha256":"6956ce4afd403feea584f97ff7f3f805d99fdd2ad6e4e5e22e6f1ea189c19676","source":{"kind":"arxiv","id":"2608.04169","version":1},"attestation_state":"computed","paper":{"title":"On Design Principles for Efficient Heterogeneous DRAM-PIM-GPU Systems","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.ET"],"primary_cat":"cs.AR","authors_text":"Corey Lammie, Hadjer Benmeziane, Irem Boybat, William Andrew Simon","submitted_at":"2026-08-04T19:26:54Z","abstract_excerpt":"Heterogeneous DRAM-based processing-in-memory (PIM)-GPU systems promise significant efficiency gains for decode-phase large language model (LLM) inference, particularly in long-output generation, yet current design practices overlook critical factors that determine real-world performance. Through systematic evaluation of diverse architectures and workloads (OPT-7B/70B, Mamba2-2.7B/70B), we reveal three fundamental design principles: (i) static power consumption (DRAM leakage, refresh, and GPU idle power) can dominate the efficiency calculus, causing dynamic-only models to overestimate tokens/s"},"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":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2608.04169","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AR","submitted_at":"2026-08-04T19:26:54Z","cross_cats_sorted":["cs.ET"],"title_canon_sha256":"736f91220ac03ad8db085febdcb4b6fe09fde910e54c25c55de796dfef3c7175","abstract_canon_sha256":"f4130310ea172759142451db05ab463b350cdcdb0dd434fddb441903529e8530"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-06T01:30:45.338547Z","signature_b64":"PEhppNU8nZSKbNzrCNSNqnbKc1YAb5vpOPrNHAlAN/8JbOCn7NSrBTSKL+GSljjqHc1vcdsIWtyDITHjhPbZCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6956ce4afd403feea584f97ff7f3f805d99fdd2ad6e4e5e22e6f1ea189c19676","last_reissued_at":"2026-08-06T01:30:45.337106Z","signature_status":"signed_v1","first_computed_at":"2026-08-06T01:30:45.337106Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"On Design Principles for Efficient Heterogeneous DRAM-PIM-GPU Systems","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.ET"],"primary_cat":"cs.AR","authors_text":"Corey Lammie, Hadjer Benmeziane, Irem Boybat, William Andrew Simon","submitted_at":"2026-08-04T19:26:54Z","abstract_excerpt":"Heterogeneous DRAM-based processing-in-memory (PIM)-GPU systems promise significant efficiency gains for decode-phase large language model (LLM) inference, particularly in long-output generation, yet current design practices overlook critical factors that determine real-world performance. Through systematic evaluation of diverse architectures and workloads (OPT-7B/70B, Mamba2-2.7B/70B), we reveal three fundamental design principles: (i) static power consumption (DRAM leakage, refresh, and GPU idle power) can dominate the efficiency calculus, causing dynamic-only models to overestimate tokens/s"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.04169","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2608.04169/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"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":"2608.04169","created_at":"2026-08-06T01:30:45.338926+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.04169v1","created_at":"2026-08-06T01:30:45.338926+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.04169","created_at":"2026-08-06T01:30:45.338926+00:00"},{"alias_kind":"pith_short_12","alias_value":"NFLM4SX5IA76","created_at":"2026-08-06T01:30:45.338926+00:00"},{"alias_kind":"pith_short_16","alias_value":"NFLM4SX5IA765JME","created_at":"2026-08-06T01:30:45.338926+00:00"},{"alias_kind":"pith_short_8","alias_value":"NFLM4SX5","created_at":"2026-08-06T01:30:45.338926+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NFLM4SX5IA765JME7F77P47YAX","json":"https://pith.science/pith/NFLM4SX5IA765JME7F77P47YAX.json","graph_json":"https://pith.science/api/pith-number/NFLM4SX5IA765JME7F77P47YAX/graph.json","events_json":"https://pith.science/api/pith-number/NFLM4SX5IA765JME7F77P47YAX/events.json","paper":"https://pith.science/paper/NFLM4SX5"},"agent_actions":{"view_html":"https://pith.science/pith/NFLM4SX5IA765JME7F77P47YAX","download_json":"https://pith.science/pith/NFLM4SX5IA765JME7F77P47YAX.json","view_paper":"https://pith.science/paper/NFLM4SX5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.04169&json=true","fetch_graph":"https://pith.science/api/pith-number/NFLM4SX5IA765JME7F77P47YAX/graph.json","fetch_events":"https://pith.science/api/pith-number/NFLM4SX5IA765JME7F77P47YAX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NFLM4SX5IA765JME7F77P47YAX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NFLM4SX5IA765JME7F77P47YAX/action/storage_attestation","attest_author":"https://pith.science/pith/NFLM4SX5IA765JME7F77P47YAX/action/author_attestation","sign_citation":"https://pith.science/pith/NFLM4SX5IA765JME7F77P47YAX/action/citation_signature","submit_replication":"https://pith.science/pith/NFLM4SX5IA765JME7F77P47YAX/action/replication_record"}},"created_at":"2026-08-06T01:30:45.338926+00:00","updated_at":"2026-08-06T01:30:45.338926+00:00"}