Phone Farm

Loading compute health board

Scoring retained benchmark history, confidence thresholds, and device compute regressions.

Resolving route12%

Acurast Pulse

Loading live view

Syncing chain state, reward ledgers, and fleet telemetry.

Resolving route12%

Manager Health #2273 | Acurast Pulse

Manager #2273 · compute baseline

Latest benchmark vs last 200 retained metric samples · ≥30 samples required to score

36.6% healthy·30d↓44.7pp

LiveUpdated 2 hrs ago·Refresh now ↻

134 devices

49Healthy

67Watch

18Investigate

0Low confidence

16Regressions

26Improving

2Stale heartbeat

18Failures

2.2ACU/day at risk

Non-Green 85 Investigate 18 Watch 67 Healthy 49 Improving 26▥ Down in Compare ▥ Stale in Compare ▥ CPU variance in Compare

How are these conditions scored?Quick FAQ

● Healthy: All monitored pools (CPU 1c, CPU Mc, RAM) score within 5 % of this phone model's median. Treated as a good baseline — nothing to act on.
● Watch: At least one pool is 5–10 % below the model median on the latest run. A nudge, not yet an actionable problem.
● Investigate: At least one pool is 10 % or more below the model median. Persistent gaps that warrant a closer look — usually thermal throttling, background load, or a bad firmware release.
● Low confidence: The device hasn't reported enough benchmark runs to score it reliably. One of: brand-new device (first seen < 72 h ago), too few samples (< 30 baseline observations), or no benchmark history at all.
● Regressions: Devices with 3 or more consecutive non-healthy benchmark runs. A single bad run is noise; three in a row is a pattern. The streak resets the moment a healthy run lands.
● Improving: Devices with 3 or more trailing healthy runs immediately after at least one non-healthy run in recent history. The mirror of Regressions — recoveries worth celebrating, and a signal that whatever you changed is working.
● ACU/day at risk: Estimated daily ACU lost vs a healthy baseline, not the total earnings of flagged devices. Each non-healthy device's ACU/day is weighted by how far below its model median it scored — a device 12% below contributes ~12% of its earnings, not 100%. Watch and Investigate items without a measured delta fall back to status floors (5–10%).
● Stale heartbeat: Last heartbeat was over 50 minutes ago (stale) or over 6 hours ago (down). Healthy heartbeats land every ~50 minutes; anything past that suggests the device is offline or unreachable.
● Failures: Devices with a composite risk score ≥ 75. The score blends benchmark drop, heartbeat lag, and on-chain execution failures; the threshold marks devices that need attention soon.

Per-model rollup12 modelsClick to expand

Model	Devices	Healthy	Watch	Investigate	% Healthy	Improving
Untagged	54	0	42	12	0.0%	10
Google Pixel 6 Pro	5	0	4	1	0.0%	2
Google Pixel 7a

ModelBuildSortShowing 2 of 85

Alias▽Processor▽Device▽CPU 1c▽CPU Mc▽RAM▽Worst▽Trend

—

5DerfA5pqvgeNz...Aez88TYU5SCORE

3h lag · 2 failures · stale

Google Pixel 7 Pro

+0.1%

+0.5%

-0.0%

▸

—

5GxEoQ4o3bQEoX...rfYm1PA2XzCORE

3h lag · 2 failures · stale

Google Pixel 7 Pro

-1.4%

-2.9%

-0.0%

-2.9%

▸

5DerfA5pqvgeNz...Aez88TYU5SCORE3h lag · 2 failures · staleGoogle Pixel 7 Pro

-0.0%▸

CPU 1c+0.1%

CPU Mc+0.5%

RAM-0.0%

Worst-0.0%

5GxEoQ4o3bQEoX...rfYm1PA2XzCORE3h lag · 2 failures · staleGoogle Pixel 7 Pro

-2.9%▸

CPU 1c-1.4%

CPU Mc-2.9%

RAM-0.0%

Worst-2.9%