13:05 johnny0: agd5f: sorry, I gave some bogus test results about PmStatusLog last week; after a PmStatusLogStart message, the log actually stays active for ~5s after the last sent PmStatusLogSample message (or just ~5s if no such message is sent)
13:07 johnny0: the PmStatusLogSample results themselves indeed look to be averages, even when called as fast as possible
13:08 johnny0: (the plots from abrupt power changes look like that of a capacitor charging / discharging)
13:12 johnny0: caching the last result is important though -- PmStatusLogSample behaves the same way when called too fast as it does when the PmStatusLog stops running from inactivity (nothing written to the ixSMU_PM_STATUS_95 status register)
13:59 johnny0: hmm... *maybe* a zero is written in the too-fast case, but nothing is written when the log is inactive
15:47 Mangix: looks like kernel 6.17.9 will fix amdgpu for me. hopefully iwlwifi too
15:56 Venemo: awesome
15:58 Mangix: I was ripping my hair out (not literally) trying to figure out why wifi was not working. I was blaming Impala.
16:15 johnny0: okay, confirmed zero is *not* being written in the too-fast case
17:22 johnny0: anyone know of a faster way to spike GPU power draw than showing/hiding the donut in furmark (or suspending/resuming)?
18:10 mareko: changing the power profile
18:10 mareko: via umr --gui
18:44 johnny0: mareko: thanks! I see the same smooth response
18:48 johnny0: and the response matches a navi23 which also has a power1_average