Docker: sustained container load is thermally untested #68

New issue

Open

opened 2026-08-30 11:38:48 +00:00 by tiagoagueda · 0 comments

tiagoagueda commented

2026-08-30 11:38:48 +00:00

Owner

Split out of the Docker enablement work.

Every load this board has carried so far has been an interactive session or a bounded
soak that someone was watching. Containers are the first thing that will run it hard,
unattended, indefinitely.

What is known

58 C idle across four thermal zones, no throttling observed
no cpufreq - fixed clock, so there is no DVFS headroom to give back under thermal
pressure, and no governor to observe
AR100 DVFS is blocked (#30), so that is not a near-term fix
4 A7 cores; the A15 cluster stays down (#53)

What is not known

What four cores at 100 % for hours does to die temperature, and whether anything
throttles or resets before that.

The 4 GiB work (57-4gb-reclaimed.md) established a decent methodology for this shape
of question - sustained load, fixed duration, check dmesg for hardware errors
afterwards - and it applies directly.

Done when

A multi-hour CPU soak has been run with thermal-zone logging, and peak temperature plus
any throttling events are recorded. If it turns out fine, that is a one-line answer
worth having before something important is hosted here.

Split out of the Docker enablement work. Every load this board has carried so far has been an interactive session or a bounded soak that someone was watching. Containers are the first thing that will run it hard, unattended, indefinitely. ## What is known - 58 C idle across four thermal zones, no throttling observed - **no cpufreq** - fixed clock, so there is no DVFS headroom to give back under thermal pressure, and no governor to observe - AR100 DVFS is blocked (#30), so that is not a near-term fix - 4 A7 cores; the A15 cluster stays down (#53) ## What is not known What four cores at 100 % for hours does to die temperature, and whether anything throttles or resets before that. The 4 GiB work (`57-4gb-reclaimed.md`) established a decent methodology for this shape of question - sustained load, fixed duration, check `dmesg` for hardware errors afterwards - and it applies directly. ## Done when A multi-hour CPU soak has been run with thermal-zone logging, and peak temperature plus any throttling events are recorded. If it turns out fine, that is a one-line answer worth having before something important is hosted here.

tiagoagueda added the

hardware

P3-normal

labels

2026-08-30 11:38:48 +00:00