Skip to content

app: smc: pytest: add e2e tests for power#936

Draft
jgrowdenTT wants to merge 1 commit intotenstorrent:mainfrom
jgrowdenTT:power_test
Draft

app: smc: pytest: add e2e tests for power#936
jgrowdenTT wants to merge 1 commit intotenstorrent:mainfrom
jgrowdenTT:power_test

Conversation

@jgrowdenTT
Copy link
Contributor

@jgrowdenTT jgrowdenTT commented Jan 6, 2026

Add e2e stress test to test power ioctl. Ensure a suitable power delta is achieved and that the device remains active.

Needs KMD 2.6.0, otherwise the test is skipped. Locally, this passes on my p150a with KMD 2.6.0.

I would not expect this to pass on galaxy.

@jgrowdenTT jgrowdenTT marked this pull request as draft January 6, 2026 16:54
@jgrowdenTT jgrowdenTT force-pushed the power_test branch 2 times, most recently from 8d67d0c to 4d0e637 Compare January 6, 2026 17:36
@jgrowdenTT jgrowdenTT force-pushed the power_test branch 3 times, most recently from 7e6dc5c to cb1a9ae Compare January 19, 2026 15:09
@jgrowdenTT jgrowdenTT marked this pull request as ready for review January 19, 2026 16:41
@jgrowdenTT jgrowdenTT requested review from ShLiTT and afongTT January 19, 2026 16:45
Test toggling between high and low power states and verify TDP delta.

Toggles between high and low power states and verifies that the TDP
difference between the two states is greater than 60W.
Copy link
Contributor

Choose a reason for hiding this comment

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

Comments say 60 W, but code says 30 W.


def test_power_state_toggle(arc_chip_dut, asic_id):
"""
Validates that toggling between high and low power states results in a TDP delta > 60W
Copy link
Contributor

Choose a reason for hiding this comment

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

Same as other comment, 60 or 30 W?

Copy link
Contributor Author

@jgrowdenTT jgrowdenTT Jan 20, 2026

Choose a reason for hiding this comment

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

The p100 isn't saving as much power as the others.

To pass CI, I'm decreasing the limit.

I also see locally my new board (p150a) is saving less than my old board. So I've been investigating what's going on.

The test should go in; so lets start with 30, see if we can do better, then adjust limits later.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

As discussed offline -> We shouldn't be using TDP for this but should be using the total input power; this allows us to capture stuff like the MRISC power going down and up with the power command.

ShLiTT
ShLiTT previously approved these changes Jan 20, 2026
Add e2e stress test to test power ioctl. Ensure a suitable power delta
is achieved and that the device remains active.

Needs KMD 2.6.0, otherwise the test is skipped.

Signed-off-by: James Growden <jgrowden@tenstorrent.com>
@alexapostolu
Copy link
Member

@afongTT @ShLiTT pinging for another review

@jgrowdenTT jgrowdenTT marked this pull request as draft February 9, 2026 17:39
@jgrowdenTT
Copy link
Contributor Author

Just a reminder, no merge until KMD 2.7.0 is on our runners.

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.

4 participants