Skip to content

arm64: dts: apple: t6030-j514s: Add airflow, battery gauge and GPU temperatures - #488

Open
ilyasakin wants to merge 1 commit into
omacom:integration/aurorafrom
ilyasakin:j514s-sensors
Open

ilyasakin wants to merge 1 commit into
omacom:integration/aurorafrom
ilyasakin:j514s-sensors

Conversation

@ilyasakin

Copy link
Copy Markdown

What this changes and why

The SMC of the J514S (MacBook Pro 14", M3 Pro) reports more temperatures than its hwmon node lists. This adds seven of them to t6030-j514s.dts:

  • Airflow: TaLP "Airflow Left", TaRF "Airflow Right"
  • Battery: TB2T "Battery Gas Gauge", beside the existing hotspot TB0T
  • GPU: Tf24, Tf28, Tf29 and Tf2E as "GPU Temperature 1–4"

How each was checked against macOS on another J514S (macOS 27.2, SMC keys read through AppleSMC, HID sensors through IOHIDEventSystem):

  • Battery: TB2T reads exactly the HID sensor macOS names "gas gauge battery" (35.0 °C both). TB1T repeats TB0T on both systems, so it is left out.
  • GPU: macOS reads the GPU's own Tg keys. The Tf2x keys track them about 8 °C lower. In 30 s of Metal load, Tg05 went 62.6 → 70.4 °C and Tf24 53.0 → 62.8 °C, while the CPU's Tp0U cooled 77.2 → 66.3 °C.
  • Airflow: macOS publishes no name for TaLP/TaRF. The names come from the Stats app's key list, and the readings rise slowly with system heat. This is the least certain of the three.

Why Tf2x for the GPU: the GPU's own Tg keys read 0 °C or -3.15 °C under Linux, even with the GPU rendering. macOS on the same model reads 49–55 °C from them, so something in the Linux GPU/PMP path does not feed them yet. The Tf2x keys rose 11–14 °C with only the GPU busy and about 1 °C with all P cores busy, so they stand in for the GPU.

The new nodes are appended after the existing ones, so channel indices that thermal zones refer to (#481) do not change.

What was tested

What was NOT tested

  • J516S: it probably has the same keys, but I have no 16" machine to check.
  • The Tg keys: why they read 0 under Linux is not solved here.

Risk

  • Scope: device tree only, J514S only. A key the SMC lacks is skipped by the driver.

AI assistance

  • I used an AI tool or agent to write or draft part of this change: Claude (claude-opus-5-5), via Claude Code. It measured the keys, wrote the change and this text, with me running the machine and reviewing.
  • A human has read the whole diff and takes responsibility for it.

🤖 Generated with Claude Code

…mperatures

The SMC of a J514S (MacBook Pro 14", M3 Pro) reports more temperatures
than its hwmon node lists. Add the two airflow sensors (TaLP, TaRF), the
battery gas gauge's temperature (TB2T) and four GPU temperatures (Tf24,
Tf28, Tf29, Tf2E).

TB2T reads exactly what macOS publishes as "gas gauge battery" on the same
model; TB1T repeats the existing TB0T hotspot and is left out. The airflow
names are those macOS tools give these keys on Apple silicon.

The GPU's own Tg keys read 0 C or -3.15 C under Linux, even with the GPU
rendering, while macOS on the same model reads them normally. The Tf2x
keys follow the GPU: on macOS they track its Tg keys about 8 C below them
(both rose 8-10 C in 30 s of GPU load while the CPU cooled), and here they
rose 11-14 C with only the GPU busy and about 1 C with all P cores busy.

Assisted-by: Claude:claude-opus-5-5
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Ilyas Akin <ashnwor@gmail.com>

@scottjones scottjones left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks, this looks good. One question, not a blocker: does Tf2E read differently from Tf24 under GPU load? If they track each other, one of them may be redundant.

@ilyasakin

Copy link
Copy Markdown
Author

Measured on the J514S: 15 s idle, 90 s of glmark2 terrain fullscreen (all GPU cores busy), then 30 s cooling, one sample per second of all four keys.

Tf24 Tf28 Tf29 Tf2E
idle 35.9 35.9 35.5 35.9
end of load 49.8 49.7 49.5 49.8
max 50.0 49.9 49.5 50.0

Tf2E stayed within 0.2 C of Tf24 for the whole run (correlation 1.000; Tf28 1.000, Tf29 0.999). So under a full GPU load they track each other, and Tf2E adds nothing to the zone's maximum there. It may still differ under a partial load (one GPU region busy), which I haven't isolated.

It read on every one of the 135 samples here and in the earlier runs, so it doesn't add a read-failure risk I've seen. I'd keep it: it's cheap, and dropping it would lose the one key that might separate under a partial load. If you'd rather keep the die list minimal for the thermal cap, I'm happy to drop Tf2E from die_keys in #481 and leave it as a hwmon channel in #488.

@c4software

Copy link
Copy Markdown

Tested on a second J514S, at idle on AC: Airflow Left 36 °C, Airflow Right 37 °C, Battery Gas Gauge 28 °C (Battery Hotspot 29 °C), GPU Temperature 1–4 at 44 °C. All seven channels are present and readable.

Not tested: GPU load.

Test setup

Testing and write-up assisted by Claude Code.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants