{
 "_doc": "One result as verifycorelabs.com prints it: the headline, the problem and the buyer cut (never reworded) from the lab's plain-English account of the result, the one figure a buyer is shown and its limit, and the lab's current claim and limits verbatim, copied at the commit named in source. A field is left out, never reworded, when it uses a word this site keeps for its own process or names a file this site does not serve (not_reprinted).",
 "slug": "reorder-memory-mechanism",
 "company": "AxiomLimit",
 "order": 5,
 "rank": 2,
 "served_receipt": "inference-limits/results/reorder-memory-mechanism.json",
 "lab_file_older": false,
 "headline": [
  "In a computer simulation of an AI-cluster network, a receiving network card keeps track of",
  "that arrive out of order with a small note of fixed size instead of a buffer that grows with the data in flight"
 ],
 "qualification": [
  "that beats cards that buffer (the simulator's default setting) by thousands of times, but beats the existing fixed-size designs (AMD STrack, Meta CTS) on memory by less than two times — often by tens of bytes — and its real edge over them is that it keeps far more throughput when traffic is sprayed across many paths."
 ],
 "why": [
  "When AI-cluster traffic is sprayed across many network paths,",
  "arrive out of order; a receiving card that buffers them needs on-chip memory that grows with the data in flight"
 ],
 "buyer": [
  "makers of network-interface cards and switch chips for AI data-centre Ethernet, and large cloud operators that design their own cluster networks"
 ],
 "figure": "1.965x against faithful bounded-state prior art on the identical grid",
 "figure_from": "claim",
 "limit": "netsim simulation, not silicon.",
 "limit_from": "lab scope",
 "plain": "1.965x",
 "limfig": "1.443x",
 "quotes": {
  "geomean": "giving a 3,886x robust geomean over the closed-loop grid",
  "clamp": "The 62,208x hero divides by exactly the 16 B ratio_floor clamp; the defensible twin is hero_large_message at 39,637x with a 96 B denominator"
 },
 "claim": "Bounded descriptor state scales O(1) where the faithful buffering field scales Theta(W), giving a 3,886x robust geomean over the closed-loop grid — and 1.965x against faithful bounded-state prior art on the identical grid.",
 "limits": "Simulation, seeds [1,2,3], generated_utc 2026-06-27. Four sibling instruments are NO-SUM and must never be added: 122,880x, 42,111x, 24,320x, 62,208x. The 62,208x hero divides by exactly the 16 B ratio_floor clamp; the defensible twin is hero_large_message at 39,637x with a 96 B denominator. Three published n for the same geomean reconcile as 960 all rows / 768 wall-payer rows / 756 nonzero. FIX-I open: 122,880x still unlabelled across 17 files. STrack was excluded by name from the escape sweep until 2026-08-19; thresholds were NOT moved to accommodate it.",
 "not_reprinted": [],
 "prior_work": [
  {
   "title": "RFC 5041: Direct Data Placement over Reliable Transports",
   "by": "H. Shah, J. Pinkerton, R. Recio, P. Culley, IETF RFC, 2007",
   "url": "https://www.rfc-editor.org/rfc/rfc5041"
  },
  {
   "title": "Multi-Path Transport for RDMA in Datacenters (MP-RDMA)",
   "by": "Yuanwei Lu, Guo Chen, Bojie Li, Kun Tan, Yongqiang Xiong, Peng Cheng, Jiansong Zhang, Enhong Chen, Thomas Moscibroda, USENIX NSDI, 2018",
   "url": "https://www.usenix.org/system/files/conference/nsdi18/nsdi18-lu.pdf"
  },
  {
   "title": "Revisiting Network Support for RDMA (IRN)",
   "by": "Radhika Mittal, Alexander Shpiner, Aurojit Panda, Eitan Zahavi, Arvind Krishnamurthy, Sylvia Ratnasamy, Scott Shenker, arXiv (its Comments field: \"Extended version of the paper appearing in ACM SIGCOMM 2018\"), 2018",
   "url": "https://arxiv.org/abs/1806.08159"
  },
  {
   "title": "STrack: A Reliable Multipath Transport for AI/ML Clusters",
   "by": "Yanfang Le, Rong Pan, Peter Newman, Jeremias Blendin, Abdul Kabbani, Vipin Jain, Raghava Sivaramu, Francis Matus, arXiv, 2024",
   "url": "https://arxiv.org/abs/2407.15266"
  }
 ],
 "source": {
  "claims_file": "registry/CLAIMS_CURRENT.jsonl",
  "claims_sha256": "81776df15a1498abec7228ea179c952d01c54e031ffd4144498a30660963074b",
  "ranking_file": "views/VALUE_RANKING.md",
  "ranking_sha256": "2e980e26178b34dc49ec1d365f966be292f57d7c636143fc62c12feab72ae470",
  "dossier_sha256": "5dd9838d39b1c9196ac9e028248b6602ffda90279d018eb350daac510d000a5d",
  "commit": "25d755d2f06518b632c3658ae83f20312daac53e"
 }
}