{"tracks": [{"id": "t1", "label": "Track 1", "title": "FPGA drafter + GPU target", "color": "#2a78d6"}, {"id": "t2", "label": "Track 2", "title": "FPGA target + drafter, HBM streaming", "color": "#0f9d8a"}, {"id": "t3", "label": "Track 3", "title": "Constant-folded compute planes", "color": "#8a4fd6"}], "blocks": {"t1": [{"id": "gpu", "label": "H200 target verify (SGLang)", "card": "HOST", "group": "host", "x": 20, "y": 40, "w": 200, "h": 70, "widths_label": "PCIe Gen4 x16"}, {"id": "host", "label": "Host relay CONS_HOST", "card": "HOST", "group": "host", "x": 20, "y": 140, "w": 200, "h": 70, "widths_label": "Python/C relay"}, {"id": "kvhost", "label": "Host KV writer (KVI)", "card": "HOST", "group": "host", "x": 20, "y": 240, "w": 200, "h": 50, "widths_label": "KV rows -> HBM"}, {"id": "pcie_h", "label": "PCIe Gen4 x8 QDMA (HEAD)", "card": "LINK", "group": "pcie", "x": 230, "y": 40, "w": 20, "h": 250, "widths_label": ""}, {"id": "pcie_t", "label": "PCIe Gen4 x8 QDMA (TAIL)", "card": "LINK", "group": "pcie", "x": 680, "y": 40, "w": 20, "h": 250, "widths_label": ""}, {"id": "h_csr", "label": "AXI-Lite CSR", "card": "HEAD", "group": "ctl", "x": 260, "y": 40, "w": 120, "h": 40, "widths_label": "32b data, 9b word addr"}, {"id": "h_fsm", "label": "Top FSM", "card": "HEAD", "group": "ctl", "x": 390, "y": 40, "w": 120, "h": 40, "widths_label": "S_G1..S_LP_BAR"}, {"id": "h_shape", "label": "SHAPE_RT table", "card": "HEAD", "group": "ctl", "x": 520, "y": 40, "w": 140, "h": 40, "widths_label": "0x200/0x204"}, {"id": "h_emb", "label": "G1 embedding gather", "card": "HEAD", "group": "dp", "x": 260, "y": 90, "w": 120, "h": 40, "widths_label": "m_axi_emb 256b"}, {"id": "h_fold", "label": "G4 EMBFOLD x9 masters", "card": "HEAD", "group": "dp", "x": 390, "y": 90, "w": 120, "h": 40, "widths_label": "9x256b HBM"}, {"id": "h_rms", "label": "RMSNorm", "card": "HEAD", "group": "dp", "x": 520, "y": 90, "w": 140, "h": 40, "widths_label": "256b chunk, H=3072"}, {"id": "h_array", "label": "GEMV array (G4/G6/G7g/G7u)", "card": "HEAD", "group": "dp", "x": 260, "y": 140, "w": 260, "h": 70, "widths_label": "LANES 256 x K_PAR16 x int16; acc 48b"}, {"id": "h_e3", "label": "DOT_E3 packed DSP58", "card": "HEAD", "group": "dp", "x": 530, "y": 140, "w": 130, "h": 70, "widths_label": "17 DSP/32 prod, LAT6"}, {"id": "h_wbank", "label": "URAM weight banks", "card": "HEAD", "group": "mem", "x": 260, "y": 220, "w": 130, "h": 40, "widths_label": "32Kx32, 4 URAM/lane (1024)"}, {"id": "h_scale", "label": "Scale BRAM Q3.13", "card": "HEAD", "group": "mem", "x": 400, "y": 220, "w": 120, "h": 40, "widths_label": "16b, group 128"}, {"id": "h_fastclk", "label": "2:1 array clock 132 MHz", "card": "HEAD", "group": "clk", "x": 530, "y": 220, "w": 130, "h": 40, "widths_label": "usr_1 132 MHz"}, {"id": "h_attn", "label": "G5 attention + RoPE + softmax", "card": "HEAD", "group": "dp", "x": 260, "y": 270, "w": 200, "h": 50, "widths_label": "32x96, HEAD_PAR 8"}, {"id": "h_kv", "label": "KV cache", "card": "HEAD", "group": "mem", "x": 470, "y": 270, "w": 190, "h": 50, "widths_label": "K URAM (176) / V BRAM, CTX 768"}, {"id": "h_kvi", "label": "KVI KV import", "card": "HEAD", "group": "dp", "x": 260, "y": 330, "w": 130, "h": 40, "widths_label": "KV_LOAD from HBM"}, {"id": "h_kvifan", "label": "KVI_FAN readers 312 MHz", "card": "HEAD", "group": "dp", "x": 400, "y": 330, "w": 130, "h": 40, "widths_label": "usr_2, 256b, burst 2KB"}, {"id": "h_lvlpipe", "label": "LEVEL_PIPE", "card": "HEAD", "group": "ctl", "x": 540, "y": 330, "w": 120, "h": 40, "widths_label": "export || next child"}, {"id": "h_seam", "label": "Seam export (SEAM_WIDE)", "card": "HEAD", "group": "dp", "x": 260, "y": 380, "w": 200, "h": 40, "widths_label": "m_axi_out 256b"}, {"id": "h_cl4", "label": "CL4 HEAD fwd/merge", "card": "HEAD", "group": "link", "x": 470, "y": 380, "w": 190, "h": 40, "widths_label": "PART 16B+8B/entry"}, {"id": "hbm_h", "label": "HBM (HEAD)", "card": "HEAD", "group": "mem", "x": 260, "y": 430, "w": 400, "h": 30, "widths_label": "16 ch, NoC 256b SIs"}, {"id": "t_ingest", "label": "Ingest bridge", "card": "TAIL", "group": "dp", "x": 710, "y": 40, "w": 130, "h": 40, "widths_label": "2240 x 256b beats"}, {"id": "t_seams", "label": "Seams r4/r5/r1/r8", "card": "TAIL", "group": "mem", "x": 850, "y": 40, "w": 130, "h": 40, "widths_label": "FF -> LUTRAM (R4)"}, {"id": "t_silu", "label": "SiLU x up (G7p)", "card": "TAIL", "group": "dp", "x": 990, "y": 40, "w": 150, "h": 40, "widths_label": "8192, Q6.12"}, {"id": "t_g9", "label": "G9 down + R7 (G9X/G9Y)", "card": "TAIL", "group": "dp", "x": 710, "y": 90, "w": 210, "h": 50, "widths_label": "LANES 128, 2-bit"}, {"id": "t_g11", "label": "G11 fc_out + RMSNorm", "card": "TAIL", "group": "dp", "x": 930, "y": 90, "w": 210, "h": 50, "widths_label": "a: 192x256b Q6.9"}, {"id": "t_earlya", "label": "EARLY_A a-export", "card": "TAIL", "group": "dp", "x": 710, "y": 150, "w": 130, "h": 40, "widths_label": "6 KB under GLM..G8"}, {"id": "t_glm", "label": "GLM low-rank LM-down", "card": "TAIL", "group": "dp", "x": 850, "y": 150, "w": 140, "h": 40, "widths_label": "3072->256, 64 lanes"}, {"id": "t_tree", "label": "G12 vocab sweep + TREE", "card": "TAIL", "group": "dp", "x": 1000, "y": 150, "w": 140, "h": 40, "widths_label": "1002 tiles, TREE_P 32"}, {"id": "t_g8", "label": "G8 exact fp16 rescore", "card": "TAIL", "group": "dp", "x": 710, "y": 200, "w": 210, "h": 50, "widths_label": "G8_NCH 4 x 256b, K_C 64"}, {"id": "t_fg8", "label": "FG8: G8 under TREE", "card": "TAIL", "group": "dp", "x": 930, "y": 200, "w": 210, "h": 50, "widths_label": "overlap"}, {"id": "t_lvl", "label": "Level FSM top-W", "card": "TAIL", "group": "ctl", "x": 710, "y": 260, "w": 140, "h": 40, "widths_label": "W x 8 children"}, {"id": "t_stop", "label": "TREE_STOP / EAGER", "card": "TAIL", "group": "ctl", "x": 860, "y": 260, "w": 140, "h": 40, "widths_label": "early level exit"}, {"id": "t_clk75", "label": "TAIL clock 75 MHz", "card": "TAIL", "group": "clk", "x": 1010, "y": 260, "w": 130, "h": 40, "widths_label": "usr_0 66->75"}, {"id": "t_sibx", "label": "SIBX-E sibling lanes", "card": "TAIL", "group": "dp", "x": 710, "y": 310, "w": 210, "h": 40, "widths_label": "eager across 2 TAILs"}, {"id": "t_cl4", "label": "CL4 PART/XITEM", "card": "TAIL", "group": "link", "x": 930, "y": 310, "w": 210, "h": 40, "widths_label": "msg 7/8"}, {"id": "t_wbank", "label": "URAM weight banks", "card": "TAIL", "group": "mem", "x": 710, "y": 360, "w": 210, "h": 40, "widths_label": "640 + 256 g12x9m"}, {"id": "t_shape", "label": "SHAPE_RT (TAIL)", "card": "TAIL", "group": "ctl", "x": 930, "y": 360, "w": 210, "h": 40, "widths_label": "0x200/0x204"}, {"id": "hbm_t", "label": "HBM (TAIL)", "card": "TAIL", "group": "mem", "x": 710, "y": 430, "w": 430, "h": 30, "widths_label": "LM-head fp16 rows, seams"}, {"id": "qsfp", "label": "QSFP28 Aurora 64B66B", "card": "LINK", "group": "link", "x": 260, "y": 520, "w": 880, "h": 40, "widths_label": "4 x 25.78 Gb/s, AXIS 256b @402.8 MHz"}, {"id": "linkdma", "label": "link_dma_eng v2", "card": "LINK", "group": "link", "x": 260, "y": 570, "w": 280, "h": 40, "widths_label": "4 TX bufs, MAXB 64, 39.9 us/70 KB"}, {"id": "d2loop", "label": "D2 card-side level loop", "card": "LINK", "group": "link", "x": 550, "y": 570, "w": 280, "h": 40, "widths_label": "loop_seq_d2"}, {"id": "mp2link1", "label": "2nd Aurora link (MP2)", "card": "LINK", "group": "link", "x": 840, "y": 570, "w": 300, "h": 40, "widths_label": "link_dma1, P2"}, {"id": "card2", "label": "2nd chain HEAD+TAIL (4-card)", "card": "LINK", "group": "card4", "x": 260, "y": 630, "w": 440, "h": 50, "widths_label": "2H2T / CL4"}, {"id": "p2merge", "label": "HEAD<->HEAD P2 merge (CL4)", "card": "LINK", "group": "card4", "x": 710, "y": 630, "w": 430, "h": 50, "widths_label": "6 KB per level"}, {"id": "c4_split", "label": "4C1 split: ATT pair + MLPV pair", "card": "LINK", "group": "card4", "x": 260, "y": 730, "w": 280, "h": 50, "widths_label": "ATT TP2 heads/rank; MLPV vocab shard; QSFP hop"}, {"id": "c4_kvonc", "label": "KV_ONCARD (KVONC) on ATT", "card": "HEAD", "group": "mem", "x": 550, "y": 730, "w": 190, "h": 50, "widths_label": "fused KV-only items, CTX_WIN 768"}, {"id": "c4_kvbanks", "label": "KV_BANKS 2x512 + G6 hold", "card": "HEAD", "group": "mem", "x": 750, "y": 730, "w": 190, "h": 50, "widths_label": "2 resident streams, 0x154 scoped clear"}, {"id": "c4_retser", "label": "RET_SER result return", "card": "HEAD", "group": "dp", "x": 950, "y": 730, "w": 190, "h": 50, "widths_label": "serialized array return, R=2"}, {"id": "c4_r3early", "label": "R3_EARLY + registered 0x90 fold", "card": "HEAD", "group": "ctl", "x": 260, "y": 790, "w": 200, "h": 50, "widths_label": "R3 launch at G6 start; TIMING_HARD A"}, {"id": "c4_q5120", "label": "Q5120 ATT geometry (LUTG A8)", "card": "HEAD", "group": "dp", "x": 470, "y": 790, "w": 200, "h": 50, "widths_label": "H 5120, 20x128/rank, K_MAX 10240"}, {"id": "c4_tp4", "label": "TP4 ATT rank (7-card)", "card": "LINK", "group": "card4", "x": 680, "y": 790, "w": 220, "h": 50, "widths_label": "4 ATT + 3 MLPV; 10x128/card, hypercube G6 sum"}, {"id": "c4_w2g8", "label": "G8 rescore on 2-bit lm_head rows", "card": "TAIL", "group": "dp", "x": 910, "y": 790, "w": 230, "h": 50, "widths_label": "HBM 1,408 B/row, TREE_V_MAX 248320"}, {"id": "c4_batch", "label": "BATCH_MODE 0x240 KV bank map", "card": "HEAD", "group": "ctl", "x": 260, "y": 850, "w": 280, "h": 50, "widths_label": "modes 0/1/2; 1x1024 .. 8x256 slots"}, {"id": "c4_prefill", "label": "PF_V on-card KV prefill", "card": "HEAD", "group": "dp", "x": 550, "y": 850, "w": 280, "h": 50, "widths_label": "4th arm of the one KV write port"}, {"id": "c4_mtag", "label": "MLPV stream tag + S1 overlap", "card": "TAIL", "group": "ctl", "x": 840, "y": 850, "w": 300, "h": 50, "widths_label": "per-stream OUT base, runtime-gated"}, {"id": "c4_lutg", "label": "D9 LUT-GEMM A8 array (ATT)", "card": "HEAD", "group": "dp", "x": 260, "y": 910, "w": 280, "h": 50, "widths_label": "multiply-free 2-bit lanes, LANES 512 or 256, K_PAR 16, A8Q"}, {"id": "c4_g5ovl", "label": "G5_OVL head-group overlap + G6_REPLAY_DB", "card": "HEAD", "group": "dp", "x": 550, "y": 910, "w": 280, "h": 50, "widths_label": "16 heads / HEAD_PAR 8 = 2 groups, passes overlapped"}], "t2": [{"id": "host", "label": "Host serve_card.py / serve_pp.py", "card": "HOST", "x": 20, "y": 40, "w": 210, "h": 70, "widths_label": "token id in, argmax out"}, {"id": "qdma", "label": "QDMA H2C / C2H", "card": "HOST", "x": 20, "y": 130, "w": 210, "h": 62, "widths_label": "weights load: 1.48 GiB in 1.8 s [M]"}, {"id": "mmio", "label": "MMIO BAR0: CSR, START, stats", "card": "HOST", "x": 20, "y": 210, "w": 210, "h": 62, "widths_label": "1.37 us per read [M]"}, {"id": "pcie", "label": "PCIe Gen5 x8", "card": "LINK", "x": 250, "y": 40, "w": 36, "h": 224, "widths_label": ""}, {"id": "seq", "label": "Descriptor sequencer", "card": "HBMS", "x": 310, "y": 40, "w": 250, "h": 70, "widths_label": "256-op table, BR + LREL layer loop"}, {"id": "spec", "label": "Spec-decode ISA (r7-r9)", "card": "HBMS", "x": 310, "y": 130, "w": 250, "h": 62, "widths_label": "VROWS ATTN.TREE TOPK TACCEPT GATHER PSTEP"}, {"id": "cab", "label": "CAB activation + KV window", "card": "HBMS", "x": 310, "y": 210, "w": 250, "h": 62, "widths_label": "int8 activations; KV 27-32 pos (KV_MODE 0)"}, {"id": "stats", "label": "Stat BRAM", "card": "HBMS", "x": 310, "y": 290, "w": 250, "h": 52, "widths_label": "per-op / per-port cycles (stall_tot 3.9% [M])"}, {"id": "gemv", "label": "GEMV engine", "card": "HBMS", "x": 590, "y": 40, "w": 270, "h": 70, "widths_label": "MD=4 activation columns; W8 / W3 / W2 per op"}, {"id": "vpu", "label": "VPU", "card": "HBMS", "x": 590, "y": 130, "w": 270, "h": 62, "widths_label": "RMSNORM ROPE SILUMUL RESADD ARGMAX; VPU4 (-3% [M])"}, {"id": "attn", "label": "Attention", "card": "HBMS", "x": 590, "y": 210, "w": 270, "h": 62, "widths_label": "MQ, ATTN_MODE 0/1, KV_MODE 0/1"}, {"id": "ports", "label": "HBM port engines x32", "card": "HBMS", "x": 590, "y": 290, "w": 270, "h": 52, "widths_label": "weight FIFO TH 32 beats; 225 MHz"}, {"id": "ports64", "label": "HBM port engines x64", "card": "HBMS", "x": 590, "y": 290, "w": 270, "h": 52, "widths_label": "NP64 W8, ~2x beat rate [P]"}, {"id": "hbm", "label": "HBM2e 32 GB", "card": "MEM", "x": 880, "y": 40, "w": 250, "h": 294, "widths_label": "NP32: ~230 GB/s streamed [M]  ·  spec 819 GB/s"}, {"id": "clk", "label": "Clocks", "card": "CLK", "x": 880, "y": 360, "w": 250, "h": 52, "widths_label": "hbms_clk 225 MHz (0x1233: 200) · usr 50 MHz"}, {"id": "clk_dr", "label": "Clocks (done-right)", "card": "CLK", "x": 880, "y": 360, "w": 250, "h": 52, "widths_label": "hbms_clk 150 / hbms_mclk 200 MHz"}, {"id": "ports_dr", "label": "HBM ports x32 + HBM-FOLD", "card": "HBMS", "x": 590, "y": 290, "w": 270, "h": 52, "widths_label": "64 NMUs, 2 per engine port"}, {"id": "gemv_mr3", "label": "GEMV engine + MR3", "card": "HBMS", "x": 590, "y": 40, "w": 270, "h": 70, "widths_label": "multi-row verify: DSP58 INT8 3-lane MAC, fold F3, M<=4"}, {"id": "mx", "label": "MX engine: on-card Qwen mixer (MXE0 / MXE1)", "card": "HBMS", "x": 310, "y": 350, "w": 550, "h": 56, "widths_label": "rtl_sd6: 288 DeltaNet slots (2 x 144), PM 524,288 words, ROM 512"}, {"id": "mx6", "label": "MX engine TP6 (MXE0 / MXE1, rtl_t6)", "card": "HBMS", "x": 310, "y": 350, "w": 550, "h": 56, "widths_label": "384 DeltaNet slots (2 x 192), PM 1,048,576 words, ROM 1,024"}, {"id": "ure_head", "label": "URE drafter HEAD (card e1)", "card": "HEAD", "x": 20, "y": 288, "w": 270, "h": 56, "widths_label": "0x1306 (Llama eps) / 0x1319"}, {"id": "ure_tail", "label": "URE drafter TAIL (card cf)", "card": "TAIL", "x": 20, "y": 350, "w": 270, "h": 56, "widths_label": "0x1307; host relay link"}, {"id": "ring_arq", "label": "QSFP28 Aurora ring + ARQ link fix", "card": "LINK", "x": 310, "y": 430, "w": 550, "h": 56, "widths_label": "FIX63 / ARQ; BIST 16/16 0 errors, 0 AR retries [M F1407]"}, {"id": "ar_dr", "label": "Ring all-reduce on card (done-right lineage)", "card": "LINK", "x": 310, "y": 500, "w": 550, "h": 56, "widths_label": "HBMS-AR <-> ring at 150 / 200 MHz; TP8 exact [M F1465]"}, {"id": "chain6", "label": "Open chain: 6 target cards", "card": "CARD8", "x": 20, "y": 570, "w": 270, "h": 56, "widths_label": "63-50-3d-17-bd-99, RING_CFG[6]"}, {"id": "chain4", "label": "4+4 split: two TP4 chains", "card": "CARD8", "x": 20, "y": 570, "w": 270, "h": 56, "widths_label": "RING_CFG[6] on one 8-ring"}, {"id": "pfx", "label": "PFX chain all-reduce (RING_CFG[7])", "card": "LINK", "x": 310, "y": 570, "w": 550, "h": 56, "widths_label": "prefix/suffix per link; LEN 6144 6,314 -> 2,682 cyc [M xsim F1505]"}, {"id": "dsv4", "label": "DeepSeek host-hybrid", "card": "HOST", "x": 880, "y": 430, "w": 250, "h": 56, "widths_label": "6 cards, W8 experts from NVMe"}, {"id": "warn_lfwa", "label": "Qwen spec decode invalid", "card": "WARN", "x": 880, "y": 500, "w": 250, "h": 56, "widths_label": "M>1 verify: DSP58 lf_wa bug"}, {"id": "lanefault", "label": "Lane fault at mclk 300", "card": "WARN", "x": 880, "y": 570, "w": 250, "h": 56, "widths_label": "output wrong from token 0"}, {"id": "notexact", "label": "Not bit-exact on silicon", "card": "WARN", "x": 880, "y": 570, "w": 250, "h": 56, "widths_label": "wrong from token 0; closed"}, {"id": "ring", "label": "QSFP28 Aurora ring core (cages 0 / 3)", "card": "LINK", "x": 310, "y": 430, "w": 550, "h": 56, "widths_label": "4 x 25.78 Gb/s per direction, pace 8 + heal, DROP_RST; 99.84 Gb/s raw [M]"}, {"id": "bridge", "label": "HBMS-AR <-> ring bridge", "card": "LINK", "x": 310, "y": 500, "w": 550, "h": 56, "widths_label": "async FIFO 225 -> 150 MHz, <=256-beat frames, ALLRED P=8"}, {"id": "nopath", "label": "no data path between HBMS and the ring in this image", "card": "WARN", "x": 310, "y": 500, "w": 550, "h": 56, "widths_label": "pathfinder: floorplan + timing + ladder + BIST only"}, {"id": "pp8", "label": "8 cards: pipeline-parallel via host", "card": "CARD8", "x": 20, "y": 430, "w": 270, "h": 56, "widths_label": "70B: 10 layers / card, residual over MMIO"}, {"id": "ring8", "label": "8-card ring 17>3d>50>63>cf>e1>99>bd", "card": "CARD8", "x": 20, "y": 500, "w": 270, "h": 56, "widths_label": "TP8 all-reduce on card"}, {"id": "loop", "label": "internal loopback", "card": "WARN", "x": 310, "y": 430, "w": 550, "h": 56, "widths_label": "0x1250: ring ports end in a loopback FIFO; no transceivers"}], "t3": [{"id": "host", "label": "Host harness", "card": "HOST", "x": 20, "y": 40, "w": 210, "h": 70, "widths_label": "embedding, norms, attention, LM head"}, {"id": "mmio", "label": "AXI-Lite MMIO", "card": "HOST", "x": 20, "y": 130, "w": 210, "h": 62, "widths_label": "weight reload per layer (1.84 s/token) [P]"}, {"id": "pcie", "label": "PCIe", "card": "LINK", "x": 250, "y": 40, "w": 36, "h": 224, "widths_label": ""}, {"id": "act", "label": "int8 activation bus", "card": "FOLD", "x": 310, "y": 40, "w": 260, "h": 62, "widths_label": "x in, 8-bit"}, {"id": "fold", "label": "Folded ternary planes (LUTRAM)", "card": "FOLD", "x": 310, "y": 120, "w": 540, "h": 110, "widths_label": "7 projections (q k v o gate up down), 3.01 M weights {-1,0,+1} as LUT constants"}, {"id": "adders", "label": "Neuron adder trees", "card": "FOLD", "x": 590, "y": 40, "w": 260, "h": 62, "widths_label": "266 cycles / projection @ 50 MHz [M]"}, {"id": "readback", "label": "Neuron read-back", "card": "FOLD", "x": 310, "y": 256, "w": 260, "h": 56, "widths_label": "3.7 us each over MMIO"}, {"id": "resident", "label": "All layers weight-resident", "card": "FOLD", "x": 590, "y": 256, "w": 260, "h": 56, "widths_label": "no reload, no weight memory traffic"}, {"id": "uram", "label": "URAM K/V banks", "card": "MEM", "x": 880, "y": 40, "w": 250, "h": 110, "widths_label": "explicit URAM banks (s7); s3/s5 fell back to ~50 Mbit LUTRAM"}, {"id": "clk", "label": "Clocks", "card": "CLK", "x": 880, "y": 176, "w": 250, "h": 52, "widths_label": "usr 50 MHz (TUNIT) · 150 MHz (FOLDRING8)"}, {"id": "ring", "label": "QSFP28 ring (talon b8 PHY)", "card": "LINK", "x": 310, "y": 340, "w": 540, "h": 56, "widths_label": "8 cards, per-card clocks +-1000 ppm, 0 link errors in xsim [M xsim]"}, {"id": "card8", "label": "8-card layer pipeline", "card": "CARD8", "x": 20, "y": 340, "w": 270, "h": 56, "widths_label": "TriLM-99M across 8 V80s"}]}, "images": {"t1": [{"name": "C8", "bid": "0xE4B711C8", "date": "2026-09-24", "ver": "M01 · Baseline HEAD/TAIL split pair (C8+CF)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t"], "hl": ["h_array", "h_attn", "t_tree", "host"], "val": [153.4, "M", "pair batch-1"], "cap": ["Bottleneck: the full EAGLE drafter layer did not fit or close timing on one V80.", "Added: a stage split. HEAD runs embedding, QKV, attention, O and MLP gate/up on a 128-lane 2-bit GEMV array; TAIL runs down-proj, fc_out, low-rank LM head, vocab tree and G8 rescore. The host relays the seam each level over PCIe Gen4 x8.", "Result: 153.4 tok/s [M F0001]."], "facts": [["role", "HEAD"], ["timing", "WNS +0.052 / WHS +0.011 ns"], ["state", "tested"], ["result", "112.73 @ 2.22 w8; texts 50/50; G5 5.85x faster"]], "flows": ["level"]}, {"name": "CF", "bid": "0xE4B711CF", "date": "2026-09-24", "ver": "M01 · Baseline HEAD/TAIL split pair (C8+CF)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t"], "hl": ["h_array", "h_attn", "t_tree", "host"], "val": [153.4, "M", "pair batch-1"], "cap": ["Bottleneck: the full EAGLE drafter layer did not fit or close timing on one V80.", "Added: a stage split. HEAD runs embedding, QKV, attention, O and MLP gate/up on a 128-lane 2-bit GEMV array; TAIL runs down-proj, fc_out, low-rank LM head, vocab tree and G8 rescore. The host relays the seam each level over PCIe Gen4 x8.", "Result: 153.4 tok/s [M F0001]."], "facts": [["role", "TAIL"], ["timing", "WNS +0.015 / WHS +0.011 ns"], ["state", "on card"], ["result", "w2 153.4 @1.83 (best); w3 146.8 @1.97; w8 111 @2.21"]], "flows": ["level"]}, {"name": "E1", "bid": "0xE4B711E1", "date": "2026-09-24", "ver": "M02 · SEAM_WIDE burst export + KV-only attention skip (E1)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t"], "hl": ["h_seam", "h_attn"], "val": [178.16, "M", "pair batch-1"], "cap": ["Bottleneck: a 32-b seam export and attention spent on KV-only rows.", "Added: SEAM_WIDE = 256-b AXI burst export (m_axi_out to HBM), and KVONLY_SKIP_ATTN.", "Result: 178.16 tok/s [M F0002]; export ~41 µs/item [D F0057]."], "facts": [["role", "HEAD"], ["state", "on card"], ["result", "w2 178.2 (+17.6% vs C8), w3 175.9 @1.97 unprimed; primed w3 d2 118.96 @2.61 (priming −32%); unprimed w3 d3/5/8/15 = 2.08/2.14/2.15/2.15 flat; primed w3 d3/5/8 = 3.15/3.79/4.36 grow"]], "flows": ["level"]}, {"name": "CD", "bid": "0xE4B711CD", "date": "2026-09-24", "ver": "M03 · 2:1 fast array clock (CD) + tile overlap (I12)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk"], "hl": ["h_fastclk", "h_array"], "val": null, "cap": ["Bottleneck: the GEMV phases dominated the HEAD item at 66 MHz.", "Added: a synchronous 2:1 array clock, clk_usr_1 = 132 MHz, with tile overlap (ARRAY_FAST_TP).", "Result: 1.77× per item in xsim [M F0013], later corrected for double-counted tile overhead; CD routes +0.005 ns [M F0014]."], "facts": [["role", "HEAD"], ["timing", "WNS +0.005 / WHS +0.010 ns"], ["state", "superseded (untested)"], ["result", "superseded by E1 before flash"]], "flows": ["level"]}, {"name": "A9b", "bid": "0xE4B711A9", "date": "2026-09-24", "ver": "M03 · 2:1 fast array clock (CD) + tile overlap (I12)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk"], "hl": ["h_fastclk", "h_array"], "val": null, "cap": ["Bottleneck: the GEMV phases dominated the HEAD item at 66 MHz.", "Added: a synchronous 2:1 array clock, clk_usr_1 = 132 MHz, with tile overlap (ARRAY_FAST_TP).", "Result: 1.77× per item in xsim [M F0013], later corrected for double-counted tile overhead; CD routes +0.005 ns [M F0014]."], "facts": [["role", "HEAD"], ["timing", "WNS 0.000 / WHS +0.002 ns"], ["state", "tested"], ["result", "[M] +3-6% vs A8; -8.6..-22% vs A5 (BR063) | same build id as its twin row but a separate build copy (own slot/md5) - not a duplicate row"]], "flows": ["level"]}, {"name": "E7", "bid": "0xE4B711E7", "date": "2026-09-26", "ver": "M04 · 256-lane HEAD GEMV array (E7 / L256)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk"], "hl": ["h_array", "h_wbank"], "val": [191.52, "M", "pair batch-1"], "cap": ["Bottleneck: array throughput.", "Added: 256 lanes (from 128). Weights in URAM288 configured 32K×32, 4 URAM per lane = 1,024 banks; 256-b activation broadcast (16 × int16); 48-b scaled accumulators, 32-b lane outputs.", "Result: HEAD item 0.734× [D F0042]; pair 191.52 tok/s [M F0003]."], "facts": [["role", "HEAD"], ["timing", "WNS 0.000 / WHS +0.008 ns"], ["state", "tested"], ["result", "silicon 2026-09-26 with F3: texts mt w2 d2 50/50 = E1+CF, accept identical; HEAD G4/G6/G7 0.506x [M]; WNS 0.000 no hang (~40 runs); weight/scale drops 0"]], "flows": ["level"]}, {"name": "F3", "bid": "0xE4B711F3", "date": "2026-09-26", "ver": "M04 · 256-lane HEAD GEMV array (E7 / L256)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk"], "hl": ["h_array", "h_wbank"], "val": [191.52, "M", "pair batch-1"], "cap": ["Bottleneck: array throughput.", "Added: 256 lanes (from 128). Weights in URAM288 configured 32K×32, 4 URAM per lane = 1,024 banks; 256-b activation broadcast (16 × int16); 48-b scaled accumulators, 32-b lane outputs.", "Result: HEAD item 0.734× [D F0042]; pair 191.52 tok/s [M F0003]."], "facts": [["role", "TAIL"], ["timing", "WNS +0.280 / WHS +0.010 ns"], ["state", "on card (tested)"], ["result", "silicon 2026-09-26 with E7: GLM 546,000 -> 136,752 cyc (0.25x, ~814/child) and TREE 2,190,384 -> 1,350,384 (-38%) [M]; primed d3 TAIL busy 60.6 -> 41.7 ms"]], "flows": ["level"]}, {"name": "A8", "bid": "0xE4B711A8", "date": "2026-09-26", "ver": "M05 · KV import (KVI) + host FASTGLUE", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi"], "hl": ["h_kvi", "kvhost"], "val": [234.62, "M", "pair batch-1"], "cap": ["Bottleneck: unprimed levels; the card re-primed KV on every request.", "Added: KV import (KVI). The host writes verified KV rows to HBM and the HEAD loads them (S_KVL), plus host FASTGLUE.", "Result: 234.62 tok/s [M F0005]."], "facts": [["role", "HEAD"], ["timing", "WNS +0.002 / WHS +0.010 ns"], ["state", "tested"], ["result", "silicon: LANES 256, 0 drops, texts 50/50 = E1+CF; mt d2 w2 192.51 / w3 189.01 @1.98 (first arms); KVI L0 PASS | silicon best 234.62 tok/s with KVI + FASTGLUE=4 (main 2026-09-26) | "]], "flows": ["level"]}, {"name": "F4", "bid": "0xE4B711F4", "date": "2026-09-26", "ver": "M05 · KV import (KVI) + host FASTGLUE", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi"], "hl": ["h_kvi", "kvhost"], "val": [234.62, "M", "pair batch-1"], "cap": ["Bottleneck: unprimed levels; the card re-primed KV on every request.", "Added: KV import (KVI). The host writes verified KV rows to HBM and the HEAD loads them (S_KVL), plus host FASTGLUE.", "Result: 234.62 tok/s [M F0005]."], "facts": [["role", "TAIL"], ["timing", "WNS +0.054 / WHS +0.010 ns"], ["state", "on card (tested with A5)"], ["result", "silicon: TREE -33.2% vs F3 (primed and unprimed), matches unit xsim; GLM = F3"]], "flows": ["level"]}, {"name": "A5", "bid": "0xE4B711A5", "date": "2026-09-26", "ver": "M06 · Pipelined attention HEAD (A5 / ATTN_PIPE)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi"], "hl": ["h_attn"], "val": [268.01, "M", "pair batch-1"], "cap": ["Bottleneck: serial attention on primed and KVI levels.", "Added: pipelined G5 attention (ATTN_PIPE, CSA par 6) on the 256-lane HEAD; 32 heads × 96, HEAD_PAR 8.", "Result: 268.01 tok/s [M F0006]."], "facts": [["role", "HEAD"], ["timing", "WNS 0.000 / WHS +0.001 ns"], ["state", "tested"], ["result", "[M] KVI+FG4 mt w2 d2 268.01 (1r), 264.8 (3r); fold v2.1 neutral d2, +2.5% d5, +1.4% d7 (BR057); protocol v1 37 cells"]], "flows": ["level"]}, {"name": "FA", "bid": "0xE4B711FA", "date": "2026-09-27", "ver": "M07 · EARLY_A TAIL + EAGER/SPEC0 + KVI_FAN 312 MHz", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop"], "hl": ["t_earlya", "h_kvifan", "t_stop"], "val": [307.67, "M", "pair batch-1"], "cap": ["Bottleneck: the chain latency TAIL → next HEAD item.", "Added: EARLY_A (the TAIL exports the next `a` vector under GLM..G8), eager survivor push, SPEC0, and KVI_FAN readers on a 312 MHz HBM clock (256-b AXI, 2 KB bursts, 8 outstanding).", "Result: 304.39 tok/s [M F0335]; 307.67 on a fast boot [M F0507]."], "facts": [["role", "TAIL"], ["timing", "WNS +0.069 / WHS +0.011 ns"], ["state", "LANDED -> trantor pdi_inbox"], ["result", "handoff /shared/scratch/chain_latency/handoff/; BR060; 18:48Z STEP5 PDI oneflip_p2s_tail_asgard_0xE4B711FA_66m_12f04f1b.pdi relayed, md5+magic verified on box1, NOTE_FA_20260927.md"]], "flows": ["level"]}, {"name": "AF", "bid": "0xE4B711AF", "date": "2026-09-27", "ver": "M07 · EARLY_A TAIL + EAGER/SPEC0 + KVI_FAN 312 MHz", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop"], "hl": ["t_earlya", "h_kvifan", "t_stop"], "val": [307.67, "M", "pair batch-1"], "cap": ["Bottleneck: the chain latency TAIL → next HEAD item.", "Added: EARLY_A (the TAIL exports the next `a` vector under GLM..G8), eager survivor push, SPEC0, and KVI_FAN readers on a 312 MHz HBM clock (256-b AXI, 2 KB bursts, 8 outstanding).", "Result: 304.39 tok/s [M F0335]; 307.67 on a fast boot [M F0507]."], "facts": [["role", "HEAD"], ["timing", "WNS -0.761 / WHS 0.000 ns"], ["state", "LANDED olegvm3 11:19Z - TIMING FAIL (not relayed)"], ["result", "staged md5-OK 19:1xZ asgard plain peak 72; expected FAIL enforce-build-manifest | moved asgard -> olegvm3 (vmpar2) 19:14Z for AG2 priority; withdrawn pre-admission 09:00Z: KVI_FAN "]], "flows": ["level"]}, {"name": "A3", "bid": "0xE4B711A3", "date": "2026-09-27", "ver": "M08 · G4 embedding fold (EMBFOLD, 9 HBM masters)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop", "h_fold"], "hl": ["h_fold"], "val": null, "cap": ["Bottleneck: G4 QKV, 3,570 cycles per HEAD item [M F0724].", "Added: embedding constant-fold (EMBFOLD) fetched through 9 parallel 256-b HBM masters.", "Result: fold A/B 264.3–264.8 [M]; dropped from the HM line, kept in the MP2/CL4 HEADs (G4_EMBFOLD = 1 [M strap])."], "facts": [["role", "HEAD"], ["timing", "WNS 0.000 / WHS +0.010 ns"], ["state", "built"], ["result", "PDI 05:35Z olegvm2 ~/eagle4_artifacts/uni_oneflip/oneflip_p2s_head_asgard_0xE4B711A3_66m_1fd8740a.pdi; FAIL classes vs E1: +strap-watcher tooth (DEPTHN_GATE_WATCH.log) | 2026-09-26"]], "flows": ["level"]}, {"name": "HM", "bid": "0xE4B711BB", "date": "2026-09-27", "ver": "M08 · G4 embedding fold (EMBFOLD, 9 HBM masters)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop", "h_fold"], "hl": ["h_fold"], "val": null, "cap": ["Bottleneck: G4 QKV, 3,570 cycles per HEAD item [M F0724].", "Added: embedding constant-fold (EMBFOLD) fetched through 9 parallel 256-b HBM masters.", "Result: fold A/B 264.3–264.8 [M]; dropped from the HM line, kept in the MP2/CL4 HEADs (G4_EMBFOLD = 1 [M strap])."], "facts": [["role", "HEAD"], ["timing", "WNS -0.013 / WHS +0.002 ns"], ["state", "BUILT - HOLD (timing FAIL) -> trantor pdi_inbox/HOLD_HM (DO NOT FLASH); HM-r2 re-route FAILED to close (WNS -0.038, 268 fail, no PDI, 06:25Z"], ["result", "handoff /shared/scratch/max_head/handoff_hm (fork f320609c, MD5LIST 729d92c0); static: pyslang 0 err x6, pkg/BD dry-run OK; xsim HG0/HG2/HRED queued olegvm3; fit = AG2's (SLR1 slic"]], "flows": ["level"]}, {"name": "HX1", "bid": "0xE4B711BD", "date": "2026-09-29", "ver": "M09 · Packed-DSP58 GEMV (HX1 DOT_E3) + FS/FSL TAIL", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop", "h_fold", "h_e3"], "hl": ["h_e3", "t_seams"], "val": [323.3, "M", "pair batch-1"], "cap": ["Bottleneck: LUT congestion in the HEAD's 2-bit CSA trees, and TAIL SLR1 flip-flop saturation.", "Added: DOT_E3. One DSP58 computes a·(w_L + 2²³·w_L+1) for a lane pair; 17 DSP58 per 32 products, latency 6, bit-exact. TAIL: serializer skip (FS) and R4 seams in LUTRAM (FSL).", "Result: 323.30 tok/s [M F0511], then 331.17 with TREE_STOP [M F0567]."], "facts": [["role", "HEAD"], ["timing", "WNS +0.004 / WHS +0.009 ns"], ["state", "LANDED -> trantor pdi_inbox (QM relay 2026-09-28T09:35:11Z, md5+magic verified on box1) - FLASH CANDIDATE (user OK; pair FB2)"], ["result", "handoff /shared/scratch/dsp_gemv/handoff_hx1 (HANDOFF_MD5 + fork MD5LIST OK asgard+olegvm3); queued 09:1xZ peak 72; race_class.sh arbitrates; STEP4c-e3 per-SLR block -> DSP-GEMV ag"]], "flows": ["level"]}, {"name": "FSL", "bid": "0xE4B711CC", "date": "2026-09-29", "ver": "M09 · Packed-DSP58 GEMV (HX1 DOT_E3) + FS/FSL TAIL", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop", "h_fold", "h_e3"], "hl": ["h_e3", "t_seams"], "val": [323.3, "M", "pair batch-1"], "cap": ["Bottleneck: LUT congestion in the HEAD's 2-bit CSA trees, and TAIL SLR1 flip-flop saturation.", "Added: DOT_E3. One DSP58 computes a·(w_L + 2²³·w_L+1) for a lane pair; 17 DSP58 per 32 products, latency 6, bit-exact. TAIL: serializer skip (FS) and R4 seams in LUTRAM (FSL).", "Result: 323.30 tok/s [M F0511], then 331.17 with TREE_STOP [M F0567]."], "facts": [["role", "TAIL"], ["timing", "WNS +0.210 / WHS +0.010 ns"], ["state", "RELAYED box1 pdi_inbox 23:34Z (auto v3)"], ["result", "xsim [M]: G0 strap-0 vs FS 18/18 0 mism cycles identical; G1 strap-1 vs strap-0 18/18 0 mism cycles identical (SUM 477,390); RED idx^1 FIRED 2.33M mism first G7P; S0 OFF diag 1012 "]], "flows": ["level"]}, {"name": "HX1 BD", "bid": "0xE4B711BD", "date": "2026-10-01", "ver": "M10 · LEVEL_PIPE HEAD (HX1 BD) + FG8/FGI TAIL + RELAY_LEAN", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop", "h_fold", "h_e3", "h_lvlpipe", "t_fg8"], "hl": ["h_lvlpipe", "t_fg8"], "val": [337.22, "M", "pair batch-1"], "cap": ["Bottleneck: serial export and the G8 tail on the per-level critical path.", "Added: LEVEL_PIPE (child k exports while child k+1 runs its front), FG8 (G8 rescore under the TREE sweep), ING_SILU_OVL, and host RELAY_LEAN.", "Result: HEAD item 184.0 µs b2b [M F0679]; FGI TAIL 11,211 cyc [M F0558]; pair 337.22 tok/s [M F0669]."], "facts": [["role", "HEAD"], ["timing", "WNS +0.004 / WHS +0.009 ns"], ["state", "LANDED -> trantor pdi_inbox (QM relay 2026-09-28T09:35:11Z, md5+magic verified on box1) - FLASH CANDIDATE (user OK; pair FB2)"], ["result", "handoff /shared/scratch/dsp_gemv/handoff_hx1 (HANDOFF_MD5 + fork MD5LIST OK asgard+olegvm3); queued 09:1xZ peak 72; race_class.sh arbitrates; STEP4c-e3 per-SLR block -> DSP-GEMV ag"]], "flows": ["level"]}, {"name": "FGI", "bid": "0xE4B711AE", "date": "2026-10-01", "ver": "M10 · LEVEL_PIPE HEAD (HX1 BD) + FG8/FGI TAIL + RELAY_LEAN", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop", "h_fold", "h_e3", "h_lvlpipe", "t_fg8"], "hl": ["h_lvlpipe", "t_fg8"], "val": [337.22, "M", "pair batch-1"], "cap": ["Bottleneck: serial export and the G8 tail on the per-level critical path.", "Added: LEVEL_PIPE (child k exports while child k+1 runs its front), FG8 (G8 rescore under the TREE sweep), ING_SILU_OVL, and host RELAY_LEAN.", "Result: HEAD item 184.0 µs b2b [M F0679]; FGI TAIL 11,211 cyc [M F0558]; pair 337.22 tok/s [M F0669]."], "facts": [["role", "TAIL"], ["timing", "WNS 0.014 / WHS 0.010 ns"], ["state", "RELAYED 2026-10-01 02:24Z (box1 verified); SUPERSEDED by FGIpx 746d2f94 (spare)"], ["result", "[M xsim] -517 cyc/TAIL item (11728->11211), 0 mismatches; [P] -8.5 us/level, +0.3..0.4% batch 1; F0558-F0562 | [M] PDI: NEW 0 vs FG8Lax 26a5f241, FSLCHECK PASS, ISOCHECK PASS (iso_"]], "flows": ["level"]}, {"name": "TCLK75", "bid": "0xE4B71199", "date": "2026-10-01", "ver": "M11 · TAIL clock 66 -> 75 MHz (TCLK75) - RECORD", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop", "h_fold", "h_e3", "h_lvlpipe", "t_fg8", "t_clk75"], "hl": ["t_clk75"], "val": [339.12, "M", "pair batch-1"], "cap": ["Bottleneck: the single TAIL kernel clock.", "Added: the same FGI RTL with the TAIL clock moved 66 → 75 MHz; routes WNS +0.016 / WHS +0.012 ns [M].", "Result: the RECORD, 339.12 tok/s [M F0686] vs GPU-only EAGLE 446–486 [M F0007/F0286]."], "facts": [["role", "TAIL"], ["timing", "WNS 0.016 / WHS 0.012 ns"], ["state", "LANDED 2026-10-01 midyard (TCLK75md); PASS; relayed pdi_inbox"], ["result", "[M] routed @75 MHz WNS +0.016/WHS +0.012; TCLKCHECK/FSLCHECK/ISOCHECK PASS; only new class OOC-WNS -0.031 (pre-route) vs FGI d197bac9"]], "flows": ["level"]}, {"name": "HQ_LV2", "bid": "0xE4B711D4", "date": "2026-10-01", "ver": "M12 · QSFP card link (Design D / CP1Q / LV2, link_dma v2)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop", "h_fold", "h_e3", "h_lvlpipe", "t_fg8", "t_clk75", "qsfp", "linkdma"], "hl": ["qsfp", "linkdma"], "val": null, "cap": ["Bottleneck: host relay at 81–88 µs per hop [M].", "Added: a QSFP28 card link. Aurora 64B66B 4 × 25.78 Gb/s on GTM, AXIS 256 b at 402.8 MHz, and link_dma_eng v2 (4 BRAM TX buffers).", "Result: 39.9 µs per 70 KB hop [M xsim F0608]. Cables were not yet installed."], "facts": [["role", "HEAD"], ["timing", "WNS 0.000 / WHS 0.000 ns"], ["state", "LANDED 2026-10-01 20:37 EDT olegvm2; RELEASED 2026-10-01 ~22:00 EDT -> trantor FOUR_CARD/C_QSFP/LV2 (second QSFP option; host CP1Q_LINK_HOST"], ["result", "[M] 0 real new FAIL vs HQ_DC_v4 (strap-watcher early-sample artifact: re-score finds g_depth_n/emb_gather present)"]], "flows": ["level", "qsfp"]}, {"name": "TQ_LV2md", "bid": "0xE4B711D5", "date": "2026-10-01", "ver": "M12 · QSFP card link (Design D / CP1Q / LV2, link_dma v2)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop", "h_fold", "h_e3", "h_lvlpipe", "t_fg8", "t_clk75", "qsfp", "linkdma"], "hl": ["qsfp", "linkdma"], "val": null, "cap": ["Bottleneck: host relay at 81–88 µs per hop [M].", "Added: a QSFP28 card link. Aurora 64B66B 4 × 25.78 Gb/s on GTM, AXIS 256 b at 402.8 MHz, and link_dma_eng v2 (4 BRAM TX buffers).", "Result: 39.9 µs per 70 KB hop [M xsim F0608]. Cables were not yet installed."], "facts": [["role", "TAIL"], ["timing", "WNS 0.004 / WHS 0.008 ns"], ["state", "LANDED 2026-10-01 20:20 EDT midyard; RELEASED 2026-10-01 ~22:00 EDT -> trantor FOUR_CARD/C_QSFP/LV2 (second QSFP option; host CP1Q_LINK_HOST"], ["result", "[M] 0 NEW / 0 DROPPED FAIL classes vs TQ_EB_v4 492a12fe; in-pipeline link G1..G7 INCONCLUSIVE as parent"]], "flows": ["level", "qsfp"]}, {"name": "HQ_D2c", "bid": "", "date": "2026-10-03", "ver": "M13 · D2 card-side level loop", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop", "h_fold", "h_e3", "h_lvlpipe", "t_fg8", "t_clk75", "qsfp", "linkdma", "d2loop"], "hl": ["d2loop"], "val": [348.9, "P", "pair batch-1"], "cap": ["Bottleneck: host glue of ~226 µs per level.", "Added: the D2 card-side level loop. TAIL survivors and the parent `a` return over QSFP and the HEAD issues the next level without the host.", "Result: projected 348.9 tok/s [P F0644]; pair images and kit staged."], "facts": [], "flows": ["level", "qsfp", "d2"]}, {"name": "TQ_D2", "bid": "0xE4B7119F", "date": "2026-10-03", "ver": "M13 · D2 card-side level loop", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop", "h_fold", "h_e3", "h_lvlpipe", "t_fg8", "t_clk75", "qsfp", "linkdma", "d2loop"], "hl": ["d2loop"], "val": [348.9, "P", "pair batch-1"], "cap": ["Bottleneck: host glue of ~226 µs per level.", "Added: the D2 card-side level loop. TAIL survivors and the parent `a` return over QSFP and the HEAD issues the next level without the host.", "Result: projected 348.9 tok/s [P F0644]; pair images and kit staged."], "facts": [["role", "TAIL"], ["timing", "WNS 0.006 / WHS 0.003 ns"], ["state", "LANDED 2026-10-02 02:16; PASS (gates+census) but NOT relayed: pair HEAD HQ_D2 FAIL"], ["result", "[M] d2c strap1 G3/G4/G6/G6b PASS; D2/G7 zeros = regex artifact: census loop_seq_d2 + d2_eng/link_dma_eng_lp_v2/v2c + hop_card_rx present"]], "flows": ["level", "qsfp", "d2"]}, {"name": "HQ_MP2s2", "bid": "0xE4B711EA", "date": "2026-10-04", "ver": "M14 · Two-link HEAD (MP2)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop", "h_fold", "h_e3", "h_lvlpipe", "t_fg8", "t_clk75", "qsfp", "linkdma", "d2loop", "mp2link1"], "hl": ["mp2link1"], "val": null, "cap": ["Bottleneck: 4-card topologies need a second port on the HEAD.", "Added: a two-link HEAD (aur_link1 + link_dma1), with NoC SIs S20/S21 at 256 b.", "Result: HQ_MP2s2 routes 0.000 / +0.010 ns [M]; the pair is projected at 305–320 [P F0789]."], "facts": [["role", "HEAD"], ["timing", "WNS 0.000 / WHS 0.010 ns"], ["state", "EARLY-LANDED 2026-10-04 03:05 from build dir; timing PASS (cascade hold did not recur); link gates pending official landing"], ["result", "[M] 0 failing EP; 0 new FAIL vs HQ_MP2 (partial); BCAST body 0.000; qm18/early/gMP2s2"]], "flows": ["level", "qsfp", "d2"]}, {"name": "TQ_LV2ae", "bid": "0xE4B711D5", "date": "2026-10-04", "ver": "M14 · Two-link HEAD (MP2)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop", "h_fold", "h_e3", "h_lvlpipe", "t_fg8", "t_clk75", "qsfp", "linkdma", "d2loop", "mp2link1"], "hl": ["mp2link1"], "val": null, "cap": ["Bottleneck: 4-card topologies need a second port on the HEAD.", "Added: a two-link HEAD (aur_link1 + link_dma1), with NoC SIs S20/S21 at 256 b.", "Result: HQ_MP2s2 routes 0.000 / +0.010 ns [M]; the pair is projected at 305–320 [P F0789]."], "facts": [["role", "TAIL"], ["timing", "WNS 0.019 / WHS 0.008 ns"], ["state", "LANDED 2026-10-04 15:39 midyard; PASS (QM-20) - better-margin TAIL for the MP2 pair; relay to trantor FOUR_CARD/QSFP_MP armed (box1 unreacha"], ["result", "[M] adjudicate_pdi.sh vs TQ_LV2md 96296e0d: 0 new / 0 dropped (18/18); g1fix3 link-gate replay (midyard mesh Midyard-844385, peak 33.6 GB): G1/G3/G4/G6/G6b/G7 PASS, G2/G5 waived ce"]], "flows": ["level", "qsfp", "d2"]}, {"name": "SIBXE x2", "bid": "0xE4B71197", "date": "2026-10-05", "ver": "M15 · 4-card SIBX-E 2H2T over PCIe", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop", "h_fold", "h_e3", "h_lvlpipe", "t_fg8", "t_clk75", "qsfp", "linkdma", "d2loop", "mp2link1", "t_sibx", "card2"], "hl": ["t_sibx", "card2"], "val": null, "cap": ["Bottleneck: batch-1 needs sibling lanes in parallel across cards; a deeper relay does not help.", "Added: SIBX-E 2H2T over PCIe, with two HEAD→TAIL chains and a host merge per level.", "SETBACK: on the SMI-slowed chassis, 4-card 50.00 vs 2-card 96.95 tok/s [M F0819]; KV load 9.08 ms. The record pair itself drops to 140.09 there [M F0764]."], "facts": [["role", "TAIL"], ["timing", "WNS 0.023 / WHS 0.007 ns"], ["state", "LANDED 2026-10-01 15:59 EDT asgard; PASS -> relayed FOUR_CARD/B_SIBX (4-card only)"], ["result", "[M] 0 real new FAIL vs SIBX ae08eb73: 1 new class 'STEP3: SIBX not wired into u_tree_level' = grep artifact (.SIBX_EN(SIBX), .SIBX_E_EN(SIBX_E)) u_tree_level at eagle4_drafter.sv:4"]], "flows": ["level", "qsfp", "d2", "card4"]}, {"name": "HQ_CL4", "bid": "0xE4B71200", "date": "2026-10-04", "ver": "M16 · CL4-L ring (HEAD<->HEAD QSFP merge)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop", "h_fold", "h_e3", "h_lvlpipe", "t_fg8", "t_clk75", "qsfp", "linkdma", "d2loop", "mp2link1", "t_sibx", "card2", "h_cl4", "t_cl4", "p2merge"], "hl": ["h_cl4", "t_cl4", "p2merge"], "val": [367, "P", "pair batch-1"], "cap": ["Bottleneck: the host merge in SIBX-E.", "Added: the CL4-L ring. Two D2 chains merge on card over HEAD↔HEAD QSFP P2 (PART = 16 B + 8 B per entry, XITEM `a` = 6 KB).", "Result: HQ_CL4 +0.004 / TQ_CL4 0.000 ns WNS [M]; projected 366–368 tok/s [P]. Needs 3 cables."], "facts": [["role", "HEAD"], ["timing", "WNS 0.004 / WHS 0.010 ns"], ["state", "PDI written 2026-10-04 06:49 (tooth still running); PASS (HQ early adj) -> RELAYED as THE CL4 HEAD (NOTE_BOX1_CL4_HEAD_TO_USE.md, md5 verifi"], ["result", "[M] F0796/F0797/F0802: 0 new FAIL vs HQ_MP2/HQ_D2_LV2/HQ_LV2 (gate table to PDI); census CL4/D2/G7 PASS; MP2 WHS-0.008 class did not recur"]], "flows": ["level", "qsfp", "d2", "card4", "cl4"]}, {"name": "TQ_CL4", "bid": "0xE4B71201", "date": "2026-10-04", "ver": "M16 · CL4-L ring (HEAD<->HEAD QSFP merge)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop", "h_fold", "h_e3", "h_lvlpipe", "t_fg8", "t_clk75", "qsfp", "linkdma", "d2loop", "mp2link1", "t_sibx", "card2", "h_cl4", "t_cl4", "p2merge"], "hl": ["h_cl4", "t_cl4", "p2merge"], "val": [367, "P", "pair batch-1"], "cap": ["Bottleneck: the host merge in SIBX-E.", "Added: the CL4-L ring. Two D2 chains merge on card over HEAD↔HEAD QSFP P2 (PART = 16 B + 8 B per entry, XITEM `a` = 6 KB).", "Result: HQ_CL4 +0.004 / TQ_CL4 0.000 ns WNS [M]; projected 366–368 tok/s [P]. Needs 3 cables."], "facts": [["role", "TAIL"], ["timing", "WNS 0.0 / WHS 0.005 ns"], ["state", "LANDED 2026-10-04 06:36:00 EDT olegvm2; HOLD (not relayed)"], ["result", "[M] vs TQ_D2_LV2 0xE4B7119F e759f0a6 (class new only if new vs ALL): NEW 5 (real 5); adjudication /shared/scratch/cl4/hq_cl4h/watch/TQ_CL4_oneflip_p2s_tail_asgard_0xE4B71201_66m_e4"]], "flows": ["level", "qsfp", "d2", "card4", "cl4"]}, {"name": "SG_HEAD", "bid": "0xE4B71206", "date": "2026-10-05", "ver": "M17 · Runtime shape table (SHAPE_RT)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop", "h_fold", "h_e3", "h_lvlpipe", "t_fg8", "t_clk75", "qsfp", "linkdma", "d2loop", "mp2link1", "t_sibx", "card2", "h_cl4", "t_cl4", "p2merge", "h_shape", "t_shape"], "hl": ["h_shape", "t_shape"], "val": null, "cap": ["Bottleneck: the drafter shape is fixed at compile time.", "Added: SHAPE_RT, a runtime GEMV shape table at CSR 0x200/0x204 (10-b AXI-Lite).", "Result: 3B is cycle-exact, 0 of 1,292,916 cycles differ [M]. Path to 8B (~303) and Qwen-27B (~172 tok/s) on 4-card TP [P]."], "facts": [["role", "HEAD"], ["state", "BUILDING 2026-10-05 (mesh pdi_build_head_sg peak 72; drain-wait behind HX3TF at submit) -- gate FH_SG4b vs FH_REF3 cycle-exact PASS [M]"]], "flows": ["level", "qsfp", "d2", "card4", "cl4"]}, {"name": "SG_TAIL", "bid": "0xE4B71207", "date": "2026-10-05", "ver": "M17 · Runtime shape table (SHAPE_RT)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop", "h_fold", "h_e3", "h_lvlpipe", "t_fg8", "t_clk75", "qsfp", "linkdma", "d2loop", "mp2link1", "t_sibx", "card2", "h_cl4", "t_cl4", "p2merge", "h_shape", "t_shape"], "hl": ["h_shape", "t_shape"], "val": null, "cap": ["Bottleneck: the drafter shape is fixed at compile time.", "Added: SHAPE_RT, a runtime GEMV shape table at CSR 0x200/0x204 (10-b AXI-Lite).", "Result: 3B is cycle-exact, 0 of 1,292,916 cycles differ [M]. Path to 8B (~303) and Qwen-27B (~172 tok/s) on 4-card TP [P]."], "facts": [["role", "TAIL"], ["timing", "WNS 0.015 / WHS 0.008 ns"], ["state", "LANDED 2026-10-05 18:15 EDT olegvm3; PASS (QM-22); pair relay to SHAPE_GENERIC/PAIR2 waits for SG_HEAD"], ["result", "[M] 0 real new FAIL vs TCLK75 df7af3df (strap-watcher race, enorm watcher INCONCLUSIVE->PASS); qm22/SG_PAIR2/SG_TAIL_VERDICT.txt"]], "flows": ["level", "qsfp", "d2", "card4", "cl4"]}, {"name": "SG_CL4H", "bid": "0xE4B71208", "date": "2026-10-05", "ver": "M17 · Runtime shape table (SHAPE_RT)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop", "h_fold", "h_e3", "h_lvlpipe", "t_fg8", "t_clk75", "qsfp", "linkdma", "d2loop", "mp2link1", "t_sibx", "card2", "h_cl4", "t_cl4", "p2merge", "h_shape", "t_shape"], "hl": ["h_shape", "t_shape"], "val": null, "cap": ["Bottleneck: the drafter shape is fixed at compile time.", "Added: SHAPE_RT, a runtime GEMV shape table at CSR 0x200/0x204 (10-b AXI-Lite).", "Result: 3B is cycle-exact, 0 of 1,292,916 cycles differ [M]. Path to 8B (~303) and Qwen-27B (~172 tok/s) on 4-card TP [P]."], "facts": [["role", "HEAD"], ["timing", "WNS 0.000 / WHS 0.000 ns"], ["state", "LANDED 2026-10-05 19:20 EDT (verdicts 20:50); PASS at zero margin post_route_phys_opt WNS 0.000 / WHS 0.000, 0 failing EP (route_design was "], ["result", "[M] 0 real new FAIL vs HQ_CL4 5a4f219d (strap-watcher race, async EVAC timeout); qm22/SG_CL4H/SG_CL4H_v1_VERDICT.txt"]], "flows": ["level", "qsfp", "d2", "card4", "cl4"]}, {"name": "SG_CL4T", "bid": "0xE4B71209", "date": "2026-10-05", "ver": "M17 · Runtime shape table (SHAPE_RT)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "h_fastclk", "h_kvi", "t_earlya", "h_kvifan", "t_stop", "h_fold", "h_e3", "h_lvlpipe", "t_fg8", "t_clk75", "qsfp", "linkdma", "d2loop", "mp2link1", "t_sibx", "card2", "h_cl4", "t_cl4", "p2merge", "h_shape", "t_shape"], "hl": ["h_shape", "t_shape"], "val": null, "cap": ["Bottleneck: the drafter shape is fixed at compile time.", "Added: SHAPE_RT, a runtime GEMV shape table at CSR 0x200/0x204 (10-b AXI-Lite).", "Result: 3B is cycle-exact, 0 of 1,292,916 cycles differ [M]. Path to 8B (~303) and Qwen-27B (~172 tok/s) on 4-card TP [P]."], "facts": [["role", "TAIL"], ["state", "BUILDING 2026-10-05 16:41 EDT mesh olegvm2-1424345 peak 100 (QM-22 submit_when_free after TQ_CL4s2 exit) -- gate COMPARE_TGL_CL4_levers/v2 P"]], "flows": ["level", "qsfp", "d2", "card4", "cl4"]}, {"name": "KVONC ATT (0x130C)", "bid": "0xE4B7130C", "date": "2026-11-08", "ver": "4C1 · KV_ONCARD ATT (with MLPV 0x128B r2)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_kvi", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "qsfp", "linkdma", "d2loop", "c4_split", "h_fold", "c4_kvonc"], "hl": ["c4_kvonc"], "val": [810.1, "M", "Llama-3.2-3B humaneval, 4 cards (talon2), KVONC off"], "cap": ["Bottleneck: host-side KV commit ops on every step.", "Added: KV_ONCARD strap 1 on the 0x129C dual-model ATT, so fused KV-only items commit KV on the card (CTX_WIN 768).", "Result: Llama 810.10 tok/s @ 5.88 and Qwen3.8-27B 248.19 tok/s @ 7.14 [M]. KVONC host fix A/B: 749.57 vs 738.32 tok/s @ 5.49 (+1.52 %) [M]; both below 810.10 because the KV-commit stack runs at accept 5.49.", "Status: tested on silicon (talon2)."], "facts": [["role", "ATT pair (talon2 17/3d) + MLPV 0x128B r2 (50/63)"], ["timing", "WNS +0.040 / WHS +0.012 ns"], ["state", "tested 2026-11-08; flashed 17:44Z"], ["result", "Llama 810.10 @ 5.88; Qwen 248.19 @ 7.14; KVONC K 749.57 vs B 738.32 @ 5.49; texts K=B 80/80"], ["pdi md5", "02b1efd5"], ["source", "FOURCARD_DESIGN_20261016/4C1_DELIVERY_TRACKER.md:1372-1373,1414; KVONC_HOSTFIX_20261108.md §5; ADJUDICATION_0x130C_20261108.md"]], "flows": ["level", "qsfp", "d2"]}, {"name": "KV_BANKS ATT (0x1334)", "bid": "0xE4B71334", "date": "2026-11-08", "ver": "4C1 · KV_BANKS two-stream ATT", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_kvi", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "qsfp", "linkdma", "d2loop", "c4_split", "h_fastclk", "c4_kvbanks"], "hl": ["c4_kvbanks"], "val": null, "cap": ["Bottleneck: one KV context per ATT card, so a second stream replays its KV at every switch (N2 measured 239.67 tok/s aggregate).", "Added: KV_BANKS 2 x 512 slots per head with a scoped 0x154 clear, G6_HOLD_GATE (holds raw-tile issue during a link stall) and the K8W1 floorplan. No KV_ONCARD.", "Result: xsim G6 hold loses 0/36 tiles at stalls of 100 and 512 cycles, +0 cycles; plain SV reference 0 differing words [M].", "Status: flash-ready (adjudicated, 0 flash-relevant FAILs); not yet run on silicon."], "facts": [["role", "ATT (dual model), pairs with MLPV 0x128B r2"], ["timing", "WNS +0.041 / WHS +0.011 ns"], ["state", "flash-ready 2026-11-08"], ["pdi md5", "dd975b24"], ["note", "run with EAGLE4_CTX_WIN=512"], ["source", "FOURCARD_DESIGN_20261016/ADJUDICATION_0x1334_20261108.md; BUILD_VERIFY_20261109/V_0x1334_0x1322.md"]], "flows": ["level", "qsfp", "d2"]}, {"name": "RET_SER R=2 ATT (0x1322)", "bid": "0xE4B71322", "date": "2026-11-08", "ver": "4C1 · RET_SER return serializer on KV_BANKS", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_kvi", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "qsfp", "linkdma", "d2loop", "c4_split", "h_fastclk", "c4_kvbanks", "c4_retser"], "hl": ["c4_retser"], "val": [759.02, "M", "Llama-3.2-3B humaneval 80x256, 4 cards (talon2), CTX_WIN 512, mean of 3"], "cap": ["Bottleneck: SLR1-SLR0 wires. The full-width result return of 0x1309 needed 18,767 of 18,870 SLLs.", "Added: RET_SER R=2, which serializes the array result return, on the K8W1 KV_BANKS ATT.", "Result: full-drafter replay bit-identical 71/71 on both streams; cost +10,736 busy cycles (+1.04 %), +2,684 on each of G4/G6/G7g/G7u [M xsim].", "Status: tested on silicon (talon2, 2026-11-09): Llama 759.02 @ 5.45, Qwen 233.73 @ 6.72 [M]."], "facts": [["role", "ATT (dual model) + MLPV 0x128B r2"], ["timing", "WNS +0.042 / WHS +0.012 ns, 0 route errors"], ["result", "Llama 759.02 @ 5.45; Qwen 233.73 @ 6.72"], ["correctness", "Llama 13/13, Qwen 9/9 trees = 0x1339; texts 80/80 = 0x1334"], ["pdi md5", "45897272"], ["source", "FOURCARD_DESIGN_20261016/SILICON_0x1322_20261109.md §0; ADJUDICATION_0x1322_20261108.md"]], "flows": ["level", "qsfp", "d2"]}, {"name": "R3_EARLY ATT (0x1339)", "bid": "0xE4B71339", "date": "2026-11-08", "ver": "4C1 · R3_EARLY + registered 0x90 fold", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_kvi", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "qsfp", "linkdma", "d2loop", "c4_split", "h_fastclk", "c4_kvbanks", "c4_r3early"], "hl": ["c4_r3early"], "val": [764.64, "M", "Llama-3.2-3B humaneval 80x256, 4 cards (talon2), CTX_WIN 512"], "cap": ["Bottleneck: R3 waited for G6 done; the 0x90 debug XOR fold was a 135-LUT6 chain (F0 post-place -17.376).", "Added: on 0x1334, R3_EARLY 1 (R3 launches at G6 start) and TIMING_HARD class A (registered 0x90 fold).", "Result: identity replay vs 0x1334 PASS 38/38 and 33/33, X count 0, R3 RED bites; plain SV reference 0 differing words [M]. Post-place WNS +0.002 (route not final). Silicon 2026-11-09: Llama 764.64 @ 5.45 and Qwen 234.94 @ 6.72; goldens and texts = 0x1334 [M].", "Status: tested on silicon (talon2, 2026-11-09); the 11-08 F0b failure did not reproduce on this image."], "facts": [["role", "ATT (dual model) + MLPV 0x128B r2"], ["result", "Llama 764.64 @ 5.45; Qwen 234.94 @ 6.72 (CTX_WIN 512)"], ["correctness", "goldens Llama 8/8, Qwen 9/9 = 0x1334; texts 80/80 = 0x1334"], ["state", "tested 2026-11-09"], ["source", "FOURCARD_DESIGN_20261016/SILICON_0x1339_20261109.md §0; R3EARLY_0x1339_20261108.md §5"]], "flows": ["level", "qsfp", "d2"]}, {"name": "Q5120 ATT L256 (0x1344)", "bid": "0xE4B71344", "date": "2026-11-08", "ver": "4C1 · Qwen3.8-27B drafter Q5120 ATT (TP2)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_kvi", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "qsfp", "linkdma", "d2loop", "c4_split", "c4_q5120", "c4_kvbanks", "c4_retser"], "hl": ["c4_q5120"], "val": null, "cap": ["Bottleneck: the Qwen3.8-27B drafter (H 5120, 40 heads, inter 16384) does not fit the v1 split: 1,536 array URAM against 1,248 in SLR1+SLR2.", "Added: v0 split (ATT = G4, G5, G6; G7 moves to MLPV), 20x128 heads per rank, LUT-GEMM A8 at L256, K_MAX 10240, SCALE_FRAC 16, RET_SER.", "Result: full-drafter xsim root + L1 bit-exact vs the bit-true golden; ATT item 17,773 cycles = 269 µs at 66 MHz [M xsim].", "Status: landed 2026-11-09, quarantined: post-route WNS -2.616 / WHS +0.000; constraint retime with F1+F2 -0.545 / +0.001 [M]. Closure image 0x1347."], "facts": [["role", "ATT TP2"], ["timing", "post-route WNS -2.616 / WHS +0.000; retime F1+F2 WNS -0.545 / WHS +0.001 ns"], ["state", "quarantined (negative WNS)"], ["xsim", "G4/KV/G5/G6sum/h1 0 differing, both ranks"], ["source", "QWEN27B_DRAFTER_20261020/TIMING_0x1344_20261109.md:107-109; 4C1_DELIVERY_TRACKER.md:1457"]], "flows": ["level", "qsfp", "d2"]}, {"name": "Q5120 ATT L512 (0x1345)", "bid": "0xE4B71345", "date": "2026-11-08", "ver": "4C1 · Qwen3.8-27B drafter Q5120 ATT (TP2)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_kvi", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "qsfp", "linkdma", "d2loop", "c4_split", "c4_q5120", "c4_kvbanks", "c4_retser"], "hl": ["c4_q5120"], "val": null, "cap": ["Bottleneck: the L256 ATT item is 17,773 cycles, of which G4 QKV is 10,712.", "Added: LANES 512 on the same Q5120 ATT source (att_g6_tp skew 6-bit).", "Result: no card-level xsim at L512; evidence is the L256 xsim and the stage gate.", "Status: staged, not launched."], "facts": [["role", "ATT TP2"], ["state", "staged (buildx04 / buildx02)"], ["source", "QWEN27B_DRAFTER_20261020/QWEN5120_P1_20261108.md §4; BUILD_VERIFY_20261109/V_Q5120_ATT.md finding 3"]], "flows": ["level", "qsfp", "d2"]}, {"name": "Q5120 TP4 ATT (0x1346)", "bid": "0xE4B71346", "date": "2026-11-09", "ver": "4C1 · Qwen3.8-27B drafter Q5120 ATT (TP4, 7-card)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_kvi", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "qsfp", "linkdma", "d2loop", "c4_split", "c4_q5120", "c4_tp4"], "hl": ["c4_tp4"], "val": null, "cap": ["Bottleneck: the TP2 ATT carries all of QKV at K 10240 per card.", "Added: a 7-card drafter rank: 4 ATT (TP4) + 3 MLPV, 10x128 heads per card, G4 N 3840, G6 K 1280, hypercube G6 sum, one-die array.", "Result: OOC array alone routed WNS +0.294 / WHS +0.010 [M]; 4-rank xsim h1 0/5,120 differing on all 4 ranks [M].", "Status: landed 2026-11-09, quarantined: post-route WNS -2.565 / WHS +0.006 [M]; superseded by 0x134A (TP4 closure) and 0x1347 (TP2 closure)."], "facts": [["role", "ATT TP4 rank"], ["timing", "post-route WNS -2.565 / WHS +0.006 ns"], ["state", "quarantined; superseded by 0x134A / 0x1347"], ["source", "4C1_DELIVERY_TRACKER.md:1466,1461; QWEN27B_DRAFTER_20261020/QWEN5120_TP4_ATT_20261108.md"]], "flows": ["level", "qsfp", "d2"]}, {"name": "Q5120 MLPV L512 (0x1348)", "bid": "0xE4B71348", "date": "2026-11-09", "ver": "4C1 · Qwen3.8-27B drafter Q5120 MLPV", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_kvi", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_lvl", "t_wbank", "hbm_t", "qsfp", "linkdma", "d2loop", "c4_split", "c4_q5120", "c4_w2g8"], "hl": ["c4_w2g8"], "val": null, "cap": ["Bottleneck: the MLPV item at Q5120 (G7g/G7u, G9, 248,320-id vocab) and the 788 MB-class fp16 lm-head gather.", "Added: on the 0x1342 MLPV source, LUT-GEMM A8 at L512, TREE_P 64, vocab shard per rank, and G8 exact rescore on 2-bit lm_head rows from HBM (1,408 B/row); fp16 G8 dropped.", "Result: unit gates 0 differing (46,080 / 46,080 / 90,320); full-MLPV pair xsim not yet green at verify time.", "Status: staged (buildx04); flash quarantined until the pair xsim and post-synth checks pass."], "facts": [["role", "MLPV"], ["state", "staged"], ["open", "full 349.6 MB lm-head table per card; BD-wrapper grep"], ["source", "QWEN27B_DRAFTER_20261020/QWEN5120_P2_MLPV_20261108.md; BUILD_VERIFY_20261109/V_Q5120_MLPV.md"]], "flows": ["level", "qsfp", "d2"]}, {"name": "Q5120 MLPV L256 (0x1349)", "bid": "0xE4B71349", "date": "2026-11-09", "ver": "4C1 · Qwen3.8-27B drafter Q5120 MLPV", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_kvi", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_lvl", "t_wbank", "hbm_t", "qsfp", "linkdma", "d2loop", "c4_split", "c4_q5120", "c4_w2g8"], "hl": ["c4_w2g8"], "val": null, "cap": ["Bottleneck: as 0x1348; this is the L256 / TREE_P 32 arm.", "Added: Q5120 MLPV at L256, TREE_P 32, same 2-bit-row G8 rescore and vocab shard.", "Result: full-MLPV pair xsim root green on both ranks, 381,760 words 0 differing; planted id 140001 green, rank-bit RED fails [M].", "Status: building (buildx00, launched 2026-11-09 00:49Z); quarantined until md>=2 + REDs + post-synth grep."], "facts": [["role", "MLPV"], ["state", "building"], ["source", "BUILD_VERIFY_20261109/V_Q5120_MLPV.md; QWEN5120_P2_MLPV_20261108.md; 4C1_DELIVERY_TRACKER.md:1425"]], "flows": ["level", "qsfp", "d2"]}, {"name": "BATCH + prefill ATT (merged)", "bid": "", "date": "2026-11-09", "ver": "4C1 · BATCH_MODE + prefill pair (design)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_kvi", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "qsfp", "linkdma", "d2loop", "c4_split", "c4_batch", "c4_prefill"], "hl": ["c4_batch", "c4_prefill"], "val": null, "cap": ["Bottleneck: one stream per card; switching streams re-stages KV (N2 239.67 tok/s aggregate measured, 78.9 % of card time re-stage).", "Added (design): runtime BATCH_MODE at 0x240 (0 single, 1 max-batch tree-grain, 2 balanced item-grain) over ONE KV bank/position map shared with PF_V prefill; base pf_kvprefill_20261109 (D9 LUTG + SIBW + PF_V).", "Result: none on silicon. MODE 0 must stay bit- and cycle-exact to the parent.", "Status: design; ATT built by box1-qwen-junior."], "facts": [["role", "ATT"], ["state", "design"], ["source", "BATCH_BUBBLES_20261109/IMAGE_A_MAXBATCH_NOLOSS.md §0; BATCH_BUBBLES_20261109/00_SYNTHESIS.md; 4C1_DELIVERY_TRACKER.md:1422"]], "flows": ["level", "qsfp", "d2"]}, {"name": "BATCH MLPV (merged)", "bid": "", "date": "2026-11-09", "ver": "4C1 · BATCH_MODE + prefill pair (design)", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_kvi", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "qsfp", "linkdma", "d2loop", "c4_split", "c4_batch", "c4_prefill", "c4_mtag"], "hl": ["c4_mtag"], "val": null, "cap": ["Bottleneck: the MLPV is tree-serial (one frontier per card) and its S1 overlap mode is locked off.", "Added (design): stream id in the hop frame header [29:28], per-stream OUT-DMA base + stride, tree-id tag and S1 overlap, all runtime-gated.", "Result: none on silicon.", "Status: design; MLPV built by box1-two-v2."], "facts": [["role", "MLPV"], ["state", "design"], ["source", "BATCH_BUBBLES_20261109/IMAGE_A_MAXBATCH_NOLOSS.md §0.3; 00_SYNTHESIS.md; 4C1_DELIVERY_TRACKER.md:1422"]], "flows": ["level", "qsfp", "d2"]}, {"name": "F3t G5_OVL ATT (0x13F9)", "bid": "0xE4B713F9", "date": "2026-11-09", "ver": "4C1 · E2b + G5_OVL 1 + G6_REPLAY_DB 1 + TIMING_HARD 2'b11", "present": ["gpu", "host", "kvhost", "pcie_h", "pcie_t", "h_csr", "h_fsm", "h_emb", "h_rms", "h_array", "h_wbank", "h_scale", "h_attn", "h_kv", "h_kvi", "h_kvifan", "h_seam", "hbm_h", "t_ingest", "t_seams", "t_silu", "t_g9", "t_g11", "t_glm", "t_tree", "t_g8", "t_lvl", "t_wbank", "hbm_t", "qsfp", "linkdma", "d2loop", "c4_split", "c4_lutg", "c4_retser", "c4_g5ovl"], "hl": ["c4_g5ovl"], "val": [831.14, "M", "Llama-3.2-3B humaneval 80x256, 4 cards (talon2), w2 d7, K_C 64; track-1 record"], "cap": ["Bottleneck: G5 runs its two head groups (16 heads / HEAD_PAR 8) one after the other.", "Added: G5_OVL (overlaps the two head-group passes), G6_REPLAY_DB and TIMING_HARD 2'b11 on E2b.", "Result: Llama 831.14 / 830.09 @ 5.89 vs same-day E2b mean 814.44 (+1.99 %); Qwen 249.62 / 249.04 @ 7.14 (inside E2b noise). G5 per tree -30.8 %, tree issue->done p50 -8.3 % [M]. Texts 80/80 and goldens identical to E2b.", "Status: tested on silicon (talon2); track-1 record."], "facts": [["role", "ATT pair (talon2 17/3d) + MLPV 0x1342"], ["timing", "FINAL_TRUE WNS +0.040 / WHS +0.011 ns"], ["result", "Llama 831.14 @ 5.89; Qwen 249.62 @ 7.14"], ["card census", "G5 751.55 -> 519.98 us/tree (-30.8 %); issue->done p50 2,422 -> 2,220 us (-8.3 %), Llama"], ["pdi md5", "97d8445c"], ["source", "TRACK1_RECORD_F3T_20261109/EVALUATION.md; DRAFTER_SPEED_RESEARCH_20261107/F3T_FLASH_ADJUDICATION_20261109.md"]], "flows": ["level", "qsfp", "d2"]}], "t2": [{"name": "SLR1233np32s6_200", "bid": "0xE4B71233", "date": "2026-10-09", "ver": "HBMS NP32 base", "present": ["host", "qdma", "mmio", "pcie", "seq", "spec", "cab", "stats", "gemv", "vpu", "attn", "ports", "hbm", "clk"], "hl": ["ports", "hbm"], "val": null, "cap": ["Three-SLR rtl_s6 floorplan: ports 0-15 far tile (SLR2, local BLI, KV ports), ports 16-31 near tile (SLR0).", "NP32 at 200 MHz; the first image of the HBM-streaming engine.", "Result: PASS QM-34, relayed do-not-flash [M F1174]."], "facts": [["state", "PASS QM-34, relayed do-not-flash"], ["clock", "hbms_clk 200 MHz"]]}, {"name": "SLR1235np32s6_225", "bid": "0xE4B71235", "date": "2026-10-10", "ver": "NP32 @ 225 MHz", "present": ["host", "qdma", "mmio", "pcie", "seq", "spec", "cab", "stats", "gemv", "vpu", "attn", "ports", "hbm", "clk"], "hl": ["clk"], "val": null, "cap": ["Same RTL as 0x1233 (13 of 14 sources md5-identical); clock raised to 225 MHz.", "Margin upgrade s6u: true hbms_clk +0.104 ns [D] vs +0.040.", "Result: PASS QM-35 / QM-36 [M F1194, F1200]."], "facts": [["timing", "WNS +0.040 / WHS +0.010 ns [M]"], ["clock", "hbms_clk 225 MHz"]]}, {"name": "R8C4 0x1250 b1", "bid": "0xE4B71250", "date": "2026-10-11", "ver": "RING8 C4 one-image", "present": ["host", "qdma", "mmio", "pcie", "seq", "spec", "cab", "stats", "gemv", "vpu", "attn", "ports", "hbm", "clk", "loop"], "hl": ["loop", "attn"], "val": null, "cap": ["0x1235 + crumbc C4, flashed on all 8 talon cards (partition 0).", "SETBACK: ATTN_MODE0 / layer gates fail on silicon (a0 lane-mapping hazard in hbml_attn) [M].", "No transceivers: ring ports are an internal loopback, so the ring cannot carry traffic."], "facts": [["state", "PASS QM-38; on all 8 talon cards, partition 0"], ["silicon", "a0/layer FAIL [M]"]]}, {"name": "R8_C4_attnfix s2 (0x1254)", "bid": "0xE4B71254", "date": "2026-10-15", "ver": "attnfix, serving image", "present": ["host", "qdma", "mmio", "pcie", "seq", "spec", "cab", "stats", "gemv", "vpu", "attn", "ports", "hbm", "clk", "pp8"], "hl": ["attn", "gemv", "pp8"], "val": [28.0, "M", "Llama-3.1-8B, one V80"], "cap": ["hbml_attn lane fix + Explore/AggressiveExplore route: WNS +0.002 / WHS +0.010 ns [M].", "Ladder 52/52 on all 8 talon cards [M F1257]. Serves real Llama models with no new PDI.", "Measured batch-1: 70B on 8 cards 2.34 · 8B 28.0 · 3B 54.8 · 1B 155.6 · TriLM-99M 1,018 tok/s [M]."], "facts": [["PDI", "259a2800"], ["silicon", "52/52 x 8 cards [M]"], ["70B, 8 cards PP", "2.34 tok/s [M]"], ["3B x 8 streams", "439 tok/s batch-8 aggregate [M]"], ["model switch", "3.1-3.9 ms [M]"]]}, {"name": "ring-fix link image (0x12B7)", "bid": "0xE4B712B7", "date": "2026-10-14", "ver": "ring only", "present": ["host", "mmio", "pcie", "ring", "ring8", "clk"], "hl": ["ring", "ring8"], "val": null, "cap": ["Link-only image: Aurora ring on cages 0/3 with RX pacing + heal (hop_aurora_cdc RX FIFO overflow fixed).", "BIST PASS 16/16 at len 64 / 256 / 960; raw 99.84 Gb/s; BER <= 5e-14 [M].", "RING_MAP OK; ring order 17 > 3d > 50 > 63 > cf > e1 > 99 > bd."], "facts": [["BIST", "16/16 PASS [M]"], ["raw", "99.84 Gb/s [M]"]]}, {"name": "R8_HBMSRING_merge (0x1255)", "bid": "0xE4B71255", "date": "2026-10-15", "ver": "pathfinder, building", "present": ["host", "qdma", "mmio", "pcie", "seq", "spec", "cab", "stats", "gemv", "vpu", "attn", "ports", "hbm", "clk", "ring", "nopath"], "hl": ["ring", "nopath"], "val": null, "cap": ["0x1254 s2 HBMS + the talon ring core in one image, side by side.", "RTL check: HBMS ring ports end in hbml_core's loopback FIFO, so there is no data path to the ring [M, RTL].", "Worth: floorplan/timing pathfinder + ladder + ring BIST in one image. 4 seeds building."], "facts": [["state", "BUILDING: 4 seeds (olegvm2, olegvm3, midyard)"], ["risk", "SLR1 slices ~98-99%"]]}, {"name": "R8_HBMSRING_AR8 (0x1256)", "bid": "0xE4B71256", "date": "2026-10-15", "ver": "design: true 8-card TP8", "present": ["host", "qdma", "mmio", "pcie", "seq", "spec", "cab", "stats", "gemv", "vpu", "attn", "ports", "hbm", "clk", "ring", "bridge", "ring8"], "hl": ["bridge", "ring8"], "val": [206, "P", "Llama-3.2-3B TP8, 8 cards"], "cap": ["Adds the HBMS-AR <-> ring bridge and widens ALLRED to P=8 (cfg_p 3 -> 4 bits).", "3B TP8 per-card programs + loader are ready and bit-exact in emulation; 56 all-reduces per token.", "Projected ~206 tok/s batch-1 at 12.6 us per all-reduce [P]."], "facts": [["state", "DESIGN (RTL in progress)"], ["runtime", "tp_serve.py ready, emulator bit-exact [M emu]"]]}, {"name": "R8_C4_np64w8 (0x1257)", "bid": "0xE4B71257", "date": "2026-10-15", "ver": "design: 64 ports", "present": ["host", "qdma", "mmio", "pcie", "seq", "spec", "cab", "stats", "gemv", "vpu", "attn", "hbm", "clk", "ports64"], "hl": ["ports64", "hbm"], "val": [97, "P", "Llama-3.2-3B, one V80"], "cap": ["Decode is >= 92% W8 weight streaming, capped by the 32-port beat rate (230 GB/s) [D on M profile].", "64 ports (W8-only if area requires), same ISA, margin RTL of 0x125C.", "Projected 3B on one card ~95-100 tok/s [P]."], "facts": [["state", "DESIGN: feasibility first"], ["bottleneck", "port beat rate, HBM stalls 2-4% [M]"]]}, {"name": "RING8 v8 (0x1262)", "bid": "0xE4B71262", "date": "2026-10-17", "ver": "v8: TP8 ring all-reduce, serving image", "present": ["host", "qdma", "mmio", "pcie", "seq", "spec", "cab", "stats", "gemv", "vpu", "attn", "ports", "hbm", "clk", "ring", "bridge", "ring8", "dsv4"], "hl": ["bridge", "ring8"], "val": [169.1, "M", "Llama-3.2-3B TP8, 8 cards (VPU4)"], "cap": ["0x1256-family recovery: HBMS-AR <-> ring bridge on the 8-ring, C_COLS 2 at 225 MHz, GT SerDes island; WNS +0.003 (true +0.093) [M F1299].", "First real models on the 8-card ring, bit-exact vs int sim: Llama-1B 398.83 [M F1341], 3B 169.1 with VPU4 [M F1325], 8B 97.44 tok/s [M F1343].", "Qwen3.6-27B real W8 TP8 9.85 tok/s, 36/36 greedy vs GPU golden [M F1318]. Open bug: ~1 AR in 300-1500 aborts, host-recovered [M F1310]."], "facts": [["PDI", "a41a8e90 (talon p0 x8)"], ["silicon", "ladder 52/52 x 8, BIST 16/16 dirs 0 errors [M F1308]"], ["Qwen3.8-27B", "19.37 tok/s, host levers [M F1395]"], ["DeepSeek", "V4-Flash W8, 6 cards: 0.33 tok/s [M F1348]"]]}, {"name": "DSpark v5 A6 (0x1267)", "bid": "0xE4B71267", "date": "2026-10-21", "ver": "A-prime: 64 NMUs, mclk 300", "present": ["host", "qdma", "mmio", "pcie", "seq", "spec", "cab", "stats", "gemv", "vpu", "attn", "hbm", "clk", "ports_dr", "ring", "bridge", "ring8", "lanefault"], "hl": ["ports_dr", "lanefault"], "val": null, "cap": ["Target 2a line: 64 HBM NMUs + HBM-FOLD (2 per engine port), engine 225 / mclk 300; margin re-route passed QM-38: WNS +0.040 / WHS +0.013 [M F1383].", "SETBACK on talon: G0 8/8 PASS, but Llama-3B/8B and Qwen3.6 output wrong from token 0; fixed-lane data errors on engine ports 2/9/11 [M F1392].", "Same lane-fault class as 0x126D [M F1394]; the line became exact only at hbms_clk 150 / mclk 200 (0x1317) [M F1465]."], "facts": [["PDI", "252dd6b5 (mg_ds5r_V3p80)"], ["state", "QM-38 PASS, silicon FAIL [M F1392]"], ["clocks", "engine 225 / mclk 300 MHz"]]}, {"name": "R2 done-right @225 (0x126D)", "bid": "0xE4B7126D", "date": "2026-10-21", "ver": "done-right + on-card MX, mclk 300", "present": ["host", "qdma", "mmio", "pcie", "seq", "spec", "cab", "stats", "gemv", "vpu", "attn", "hbm", "clk", "ports_dr", "mx", "ring", "bridge", "ring8", "lanefault"], "hl": ["mx", "lanefault"], "val": null, "cap": ["Done-right arm B (QWMX): adds the on-card Qwen MX engine (MXE0 / MXE1) to the 64-NMU HBM-FOLD target; QM-38 cleared with the reset-release MCP [M F1391].", "SETBACK: port/lane data errors on ports 7/12; corner-prohibit re-places cleared the lanes, but output stayed wrong from token 0 [M F1394, F1548].", "RCA: the exact build RTL == ISA emulator in xsim, so the fault is below the RTL [M xsim F1414]; fix = the 150 / 200 MHz derate (0x1317)."], "facts": [["PDI", "42226d20; re-places CPW 9a258951, CP2BS2 4e017b3f"], ["tok/s", "41-106 tok/s runs had wrong output: refuted [F1546-F1548]"], ["clocks", "hbms_clk 225 / mclk 300 MHz"]]}, {"name": "v8 + ARQ ring fix (0x130E)", "bid": "0xE4B7130E", "date": "2026-10-22", "ver": "B2r: ring link fix, DeepSeek host-hybrid", "present": ["host", "qdma", "mmio", "pcie", "seq", "spec", "cab", "stats", "gemv", "vpu", "attn", "ports", "hbm", "clk", "ring_arq", "bridge", "ring8", "dsv4"], "hl": ["ring_arq", "dsv4"], "val": [169.19, "M", "Llama-3.2-3B TP8, 8 cards"], "cap": ["v8 + ring link fix (FIX63 / ARQ) for the err 0x6 AR stalls: BIST 16/16 0 errors, ARQ counters 0, AR retries 0 [M F1407].", "Bit-exact batch-1: Llama-1B 413.66-422.48, 3B 169.19, 8B 101.23 tok/s [M F1407, F1408]; Qwen3.8-27B 16.62-19.08 tok/s, exact 4/4 [M F1409].", "DeepSeek-V4-Flash W8 host-hybrid on 6 cards, experts streamed from NVMe: 0.35 tok/s, tokens == the v8 run [M F1452]."], "facts": [["PDI", "433cf603 (talon p0 x8)"], ["timing", "G-MARGIN PASS, true hbms_clk +0.090 [M F1402]"], ["demo", "1B 365 / 3B 166 / 8B 99.9 tok/s, live switching [M F1456]"]]}, {"name": "URE drafter pair (0x1306 / 0x1307)", "bid": "0xE4B71306", "date": "2026-10-23", "ver": "2-card URE drafter, Llama + Qwen", "present": ["host", "qdma", "mmio", "pcie", "ure_head", "ure_tail"], "hl": ["ure_head", "ure_tail"], "val": null, "cap": ["Unified URE drafter on 2 V80s: HEAD 0x1306 HA_V2 on e1, TAIL 0x1307 T75_V3 on cf; the host relays drafts and verify results to the target cards.", "Silicon PASS on talon: HEAD 29 files byte-for-byte + NONZERO 32/32, TAIL 18/18 [M F1458].", "Draft step 3.89 ms overlapped / 6.27 ms serial vs GPU EAGLE3 3.555 ms [D F1473]; one URE session per boot [M F1473]."], "facts": [["PDIs", "HEAD 2d85a008, TAIL fca9622b (talon e1 / cf p0)"], ["tok/s", "drafter only, no end-to-end tok/s yet"]]}, {"name": "CHAIN4 4+4 split (0x1314)", "bid": "0xE4B71314", "date": "2026-10-23", "ver": "4 target + 4 draft split, closed", "present": ["host", "qdma", "mmio", "pcie", "seq", "spec", "cab", "stats", "gemv", "vpu", "attn", "ports", "hbm", "clk", "ring_arq", "bridge", "ring8", "chain4", "notexact"], "hl": ["chain4", "notexact"], "val": null, "cap": ["0x130E + RING_CFG[6] open-chain all-reduce: two TP4 groups at once on one 8-ring; xsim 16/16 and non-BEH netlist gate PASS [M F1406, F1413].", "SETBACK: on talon every Llama 1B/3B/8B prompt is wrong from token 0 and Qwen exactness FAILs [M F1444]; the corner-DSP re-close 0x1314-CP is also not exact [M F1536].", "Speeds of wrong outputs are refuted [F1536]. Line closed; the chain moved to the exact done-right lineage as 0x1318 [M F1479]."], "facts": [["PDIs", "b1 34c8ed05, CP 908ee653"], ["state", "NOT exact on silicon, line closed [M F1479]"], ["RCA", "2026-10-22/SPLIT_4P4/RCA_1314_SILICON.md"]]}, {"name": "DRM done-right (0x1317)", "bid": "0xE4B71317", "date": "2026-10-23", "ver": "done-right at 150 / 200 MHz, talon demo image", "present": ["host", "qdma", "mmio", "pcie", "seq", "spec", "cab", "stats", "gemv", "vpu", "attn", "ports_dr", "hbm", "clk_dr", "mx", "ring_arq", "ar_dr", "ring8", "warn_lfwa"], "hl": ["clk_dr", "ar_dr", "mx"], "val": [135, "M", "Llama-3.2-3B TP8, 8 cards"], "cap": ["The 0x126D done-right RTL derated to hbms_clk 150 / mclk 200 (NoC read bw 6400): the first done-right image exact end to end on silicon [M F1465].", "Bit-exact batch-1 on 8 cards: Llama-1B 345, 3B 135, 8B 77.8 tok/s [M F1465-F1467]; Qwen3.8-27B via the on-card MX path 33.02 tok/s [M F1470].", "The talon demo image: PDI d76ab1b1 on talon p0 10-23 to 10-26; QM-38 passed as diagnostic-only, with a waiver on the async mclk recovery path (+0.006 ns) [M F1459]."], "facts": [["PDI", "d76ab1b1 (buildz05 arm z05)"], ["timing", "true hbms_clk +0.093, mclk +0.086; async mclk +0.006 waived [M F1459]"], ["demo v4", "Qwen 32.48 / 33.81, Llama-8B 76.04 tok/s, 0 AR retries [M F1471]"], ["Llama runs", "short 15-24 token selftests [M F1465-F1467]"], ["limit", "Qwen spec decode (M>1) NOT valid: DSP58 lf_wa [M xsim F1532]"]]}, {"name": "DR_CHAIN6 6+2 (0x1318)", "bid": "0xE4B71318", "date": "2026-10-24", "ver": "done-right + 6-card open chain", "present": ["host", "qdma", "mmio", "pcie", "seq", "spec", "cab", "stats", "gemv", "vpu", "attn", "ports_dr", "hbm", "clk_dr", "mx", "ring_arq", "ar_dr", "ring8", "chain6", "ure_head", "ure_tail", "warn_lfwa"], "hl": ["chain6"], "val": null, "cap": ["0x1317 + RING_CFG[6] open chain for 6 target + 2 URE: chain 63-50-3d-17-bd-99 (ranks 0-5), e1 / cf free for the drafter pair; 8-card mode == 0x1317.", "Gates: chain netlist 19/19 [M xsim F1480]; TP6 programs == 1-card int sim for Llama 1B/3B/8B [M F1482]; G-MARGIN PASS, async mclk +0.031, no waiver [M F1486].", "Not yet on silicon (talon down). Projected 6+2 Llama-3B as built: 106 tok/s, below TP8 target-only [P F1499]."], "facts": [["PDI", "b1 4c5d32e3, QM PASS; b2 5cf01a17 spare [M F1514]"], ["Qwen TP6", "does not fit: 288 of 384 DeltaNet slots [D F1492]"], ["limit", "Qwen spec decode (M>1) NOT valid [M xsim F1532]"]]}, {"name": "URE HEAD eps fix (0x1319)", "bid": "0xE4B71319", "date": "2026-10-24", "ver": "URE HEAD for Qwen", "present": ["host", "qdma", "mmio", "pcie", "ure_head", "ure_tail"], "hl": ["ure_head"], "val": null, "cap": ["URE HEAD with the Qwen eps 1e-6 + e_norm wrap / saturation fixes; pairs with TAIL 0x1307 on cf.", "HG_V2 PDI cc4d3c09: true WNS +0.045 / WHS +0.030 [M F1503]; non-BEH netlist gate 49/49 bit-exact to RTL [M xsim F1506].", "QM PASS, staged on talon; not yet on silicon (talon down)."], "facts": [["PDI", "HG_V2 cc4d3c09 (preferred)"], ["backup", "HH_V3 ab237555, WNS +0.041 [M F1503]"]]}, {"name": "DRC6X + PFX + MR3 (0x131A)", "bid": "0xE4B7131A", "date": "2026-10-25", "ver": "6+2 with PFX all-reduce + MR3", "present": ["host", "qdma", "mmio", "pcie", "seq", "spec", "cab", "stats", "vpu", "attn", "ports_dr", "hbm", "clk_dr", "mx", "ring_arq", "ar_dr", "ring8", "gemv_mr3", "pfx", "chain6", "ure_head", "ure_tail", "warn_lfwa"], "hl": ["pfx", "gemv_mr3"], "val": null, "cap": ["0x1318 + PFX chain all-reduce (RING_CFG[7]) + MR3 multi-row GEMV (DSP58 INT8 3-lane, fold F3): the two RTL levers of the 6+2 projection.", "Gates: ring PFX on / off 57/57 [M xsim F1509], full core 19/19 [M xsim F1510], MR3 non-BEH netlist 4/4 [M F1513]; arm C G-MARGIN PASS, async mclk +0.044 [M F1526].", "Flashed on talon 10-26 (T2 relay), untested: talon down. Projected 6+2 Llama-3B at M=4: 165 tok/s = 1.27x TP8 0x1317 [P F1512]."], "facts": [["PDI", "arm C accbfdd5, QM PASS (arm A 18a4a902 backup)"], ["MR3", "K-padded M=4 1,841 vs M=1 1,777 cycles [M xsim F1498]"], ["limit", "Qwen spec decode (M>1) NOT valid [M xsim F1532]"]]}, {"name": "UNIFIED6 (0x131B)", "bid": "0xE4B7131B", "date": "2026-10-26", "ver": "Llama + Qwen TP6, DSP58 fix", "present": ["host", "qdma", "mmio", "pcie", "seq", "spec", "cab", "stats", "vpu", "attn", "ports_dr", "hbm", "clk_dr", "ring_arq", "ar_dr", "ring8", "gemv_mr3", "mx6", "pfx", "chain6", "ure_head", "ure_tail", "dsv4"], "hl": ["mx6"], "val": null, "cap": ["One 6-card target image for Llama and Qwen3.8-27B TP6: 0x131A + the TP6 MX engine, 2 x 192 = 384 DeltaNet slots per card (needed 384) [D F1517].", "DSP58 lf_wa fix (log address as shift-add): MXE netlist gate 16/16 + REDs 3/3 [M xsim F1565]; arm A2 PDI 152f73e0 G-MARGIN PASS after the reset-net re-route, async mclk +0.056 [M F1569].", "Not yet on silicon. Projected 6+2 Llama-3B at M=4: 165 tok/s [P F1512, proj_out.md row D]; DeepSeek Plan A on 6 cards 0.6 tok/s [P F1523]."], "facts": [["PDI", "A2 152f73e0, QM PASS (no waiver)"], ["Qwen", "TP6 only on this image; Qwen TP8 stays on 0x1317"], ["DeepSeek W8", "needs 8.54 cards, does not fit 6 [D F1518]"]]}], "t3": [{"name": "TUNIT_0x1258 s2", "bid": "0xE4B71258", "date": "2026-10-14", "ver": "one-card correctness unit", "present": ["host", "mmio", "pcie", "act", "fold", "adders", "readback", "clk"], "hl": ["fold", "adders"], "val": null, "cap": ["TriLM-99M layer 0: all 7 ternary projections folded into LUTRAM on one V80.", "Silicon: 0 mismatches / 10,240, 266 cycles per projection [M] (box1 card 42).", "Full model on one card reloads weights per layer over MMIO: ~0.47 tok/s [P]; the fold itself is ~1,677 tok/s [D]."], "facts": [["timing", "WNS 0.587 / WHS 0.010 ns [M]"], ["silicon", "0 / 10,240 mismatches [M]"]]}, {"name": "FOLDRING8 s3/s5 (0x1259 / 0x125A)", "bid": "0xE4B71259", "date": "2026-10-15", "ver": "stopped: URAM infeasible", "present": ["host", "pcie", "act", "fold", "adders", "resident", "uram", "clk", "ring", "card8"], "hl": ["uram"], "val": [4550, "D", "TriLM-99M, 8 cards, xsim"], "cap": ["8-card weight-resident TriLM ring; xsim G3-LINK: 6/6 tokens bit-exact, 32,985 cycles/token at 150 MHz [M xsim].", "SETBACK: synthesis found ram_style=ultra infeasible for K/V; ~50 Mbit fell back to LUTRAM. Stopped on the user's order.", "~4,550 tok/s batch-1 [D from xsim cycles]."], "facts": [["state", "STOPPED (invalid)"], ["xsim", "6/6 tokens bit-exact [M xsim]"]]}, {"name": "FOLDRING8_fast s7 (0x125A)", "bid": "0xE4B7125A", "date": "2026-10-15", "ver": "rebuild, explicit URAM", "present": ["host", "pcie", "act", "fold", "adders", "resident", "uram", "clk", "ring", "card8"], "hl": ["uram", "ring"], "val": [4550, "D", "TriLM-99M, 8 cards, xsim"], "cap": ["K/V rewritten as explicit URAM banks, with a synth-feasibility gate before the PDI.", "Building on olegvm3 (fr8_build_fast.sh s7fast, 150 MHz).", "First end-to-end constant-fold tok/s on silicon when it lands."], "facts": [["state", "BUILDING (olegvm3)"]]}]}}