You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Two verdicts the app was stating without the evidence behind them.
Apple M5 Max is described correctly. That chip has no Efficiency cores — its perf levels are Super (6) and Performance (12) — and SiliconScope got it wrong four ways at once: a "6E+12P" header, "E-cores @ 0 MHz", a usage split taken from the wrong cores, and no CPU power at all. The header now reads "6 Super + 12 Performance", both clusters report their real DVFS ceilings, and the watts land on the right rows. Every other Mac is unchanged. (#30 — measured on real hardware by @ben0112)
Which CPU belongs to which cluster is now read, not assumed. macOS publishes core counts per perf level but no ordering; that comes from the device tree. On M5 Max the top tier is the last six cores — the old assumption was wrong in direction, not just in label.
A machine cooling down is no longer reported as throttled. After a heavy run, both cards outlined red while nothing was running: thermal pressure stays elevated while the fans spin, and every clock had dropped because nothing was asking for performance. Throttling now requires a die that is hot right now, and a GPU actually drawing power — 0.4 W on the minimum clock is idle, not held back.
Also
sscope-cli --cpu-debug dumps the perf levels, the device-tree cluster map and every DVFS table present.
The 80 °C / 95 °C thresholds are shared by the throttle detector and the temperature colours, so the app cannot disagree with itself about what "hot" means.