Overview panel
The Overview panel is a summary assembled from the same metrics shown in the detailed panels. It does not collect separate data.
Headline charts
| Chart | Metric/calculation | Meaning |
|---|---|---|
| Thermal Max | Maximum finite temperature across all current thermal metrics | Hottest reported on-die or board sensor at each cache sample. The sensor producing the maximum can change over time. |
| Current Power | power_current_watts | Latest valid total of PMBus rails successfully read in the corresponding power pass. See Power for partial-read behavior. |
| CPU | cpu_usage_pct | Aggregate Linux CPU busy percentage over the latest system sample window. |
| Memory | linux_mem_used_pct | (MemTotal - MemAvailable) / MemTotal × 100. |
| MLA Memory | mla_mem_allocated_mb | Allocated bytes reported by /dev/simaai-mem, converted to MiB. This is allocator usage, not MLA utilization. |
| Network | net_rx_mbps + net_tx_mbps | Combined aggregate receive and transmit rate for all non-loopback interfaces. Despite the key name, the displayed unit is MiB/s-equivalent, not megabits/s. |
The Thermal Max chart is useful for spotting the hottest location but should not replace the Thermal panel: it hides which sensor is responsible and can switch between on-die and board sensors.
Current status table
The table reports:
- normalized one-minute CPU load;
- Linux memory used in MiB;
- MLA allocator memory in MiB;
- current board-rail power;
- eMMC and optional NVMe filesystem usage;
- aggregate receive and transmit rates.
The columns mean:
- Metric: short display name from the cache metric definition.
- Group: subsystem category.
- Value: latest finite cached value.
- Status:
normal,warning,critical, orunknown, based on the metric's configured thresholds. - Trend: normalized sparkline across retained cache history.
Notes panel
The Notes panel displays top-level daemon errors when present. With no daemon errors it displays usage guidance. An empty Notes panel is not proof that every sensor is healthy; Power has its own degraded status and individual metric values can be unavailable without becoming top-level daemon errors.