igneum-network/igneum
1
0
Fork 0
Code Issues Pull requests Releases Activity

Counter ASIC 3.0 status: PC 2 clear, the detector and vendor-share dry rows on the live tables, the restart constraint

Browse source
This commit is contained in:
igneum-josh 2026-10-06 09:26:52 +01:00
parent 477729e9dc
commit 7bd1943231
1 changed files with 10 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

11
docs/plans/counter-asic-3-status.md
View file

@ -9,7 +9,7 @@ The test every result is judged against (Josh, 6 October): a chip maker must hav
| Machine | State | Rule |
|---|---|---|
| Mac M5 Max | mining paused; Metal worker free | measurements under `with-lock.sh measure` only |
| PC 2 (1ccfe586, RTX 5090) | the proving agent's jobs `segments-pc2-pv1` runs b and c (run b claimed nothing on a PowerShell key bug; run c from about 07:53Z); "PC 2 clear" expected about 08:35Z; then three 3.0 jobs one at a time (items 2, 6, 8); the prover-floor agent queues after "PC 2 released" | nothing published until "PC 2 clear"; one job at a time (`/tmp/igneum-devnet/pc2-ca3.lock`); released by message when done |
| PC 2 (1ccfe586, RTX 5090) | CLEAR at 08:24:27Z (the clear file carries the proving agent's constraints: prover left ON, no quit or restart, /opt/igneum, /opt/igneum-segal and settings.json untouched); the three 3.0 jobs run one at a time through the mkdir lock (items 8, 2, 6); "PC 2 released" to the proving agent after the last; the prover-floor agent next. Before that: the proving agent's jobs `segments-pc2-pv1` runs b and c (run b claimed nothing on a PowerShell key bug; run c from about 07:53Z); "PC 2 clear" expected about 08:35Z; then three 3.0 jobs one at a time (items 2, 6, 8); the prover-floor agent queues after "PC 2 released" | nothing published until "PC 2 clear"; one job at a time (`/tmp/igneum-devnet/pc2-ca3.lock`); released by message when done |
| PC 1 (ae432dc7, RTX 5090 + RX 9070 XT) | Josh's desk | not used today; every AMD row OWED |
## 2. The items
@ -71,6 +71,15 @@ Consequence per tier (interim): the M5 Max at 21 W is the per-joule best honest
Step cost per family on the M5 Max as a ratio to the add-xor-rotate chain (881 G steps/s; best of 3; load average 7.64; all bit-exact): shl 0.85, shr 0.86, bfe 0.77, andn 0.75, byte permute 1.13 (emulated on Apple), popcount 0.87, clz 1.01, select 0.76, indexed shuffle 1.91 (2.2x the xor shuffle), live rotr 1.13, dot4 unsigned 1.60, dot4 signed 4.73. Every 32-bit datapath family costs an Apple lane under 1.2x a step, inside the 8x emulation bound of 1.13.2 with room; the matrix family is the only one past 1.6x. The rows feed item 8's ALU pricing. The full table with consequences lands at item 6's close.
### Items 4 and 7 on the live tables (dry mode, read-only, 09:2x UTC; the live observer still runs master's code)
| Module | Reading | Consequence |
|---|---|---|
| Detector (window epochs 41 to 46, tip DAA 171,165, 20 ids) | network 125 to 152 MH/s per epoch, settled; bands from the fleet log 5090 99.4 / 115.6 / 123 MH/s (n 849), M5 Max 26.5 / 27.4 / 28.5 (n 586), Intel UHD 1.7 / 1.8 / 2.2 (n 758); correlation pairs 6, max r 0.94, edges 1 (the two 5090 ids, PC 1 back since 07:2x), groups none; alert inactive, held 0 of 6; events none | quiet on the devnet; two findings sent to the detector worker: two honest same-model cards cross the 0.8 edge (the group rule of 3 or more ids is what keeps the alert off; the edge may be a network-estimate artefact), and each 5090 id reads about 50 MH/s on chain against the 115 MH/s band (a factor of two to explain before an unknown miner is banded). Until both are answered the detector's alert is advisory, not a trigger |
| Vendor share (10-minute window, network 253 MH/s) | fleet-reported: NVIDIA 241.7 MH/s (2 workers, 0.920), AMD 18.8 (1, 0.072), Intel 2.2 (1, 0.008), Apple 0 (the Mac paused); chain-attributed: NVIDIA 0.914 (529 of 579 blue blocks, 10 ids), AMD 0.078 (45 blocks, 8 ids), Intel 0.009, unknown 0 (every id maps to a fleet key, 30 mapped) | the two readings agree within 1 percent; the devnet is a one-vendor fleet at 91 percent NVIDIA, which is what the metric is for: an AMD owner's share of income is 0.078 for 1 of 4 cards, an Apple owner's 0 while paused; the public testnet's number is the one that matters |
The live observer runs the shared checkout, which autosync fast-forwards from origin/master; new code reaches it only through a push to master, which is Josh's call (CLAUDE.md: push only when Josh asks). So the one restart main asked for waits on that push; the dry rows above are the evidence until then.
### Consequences per tier (item 1)
| Tier | Meaning | Being done |