Bug 106428 - reported VDDGFX GPU voltage jumps above set limits in pp_od_clk_voltage when UVD is used for VAAPI/VDPAU
Summary: reported VDDGFX GPU voltage jumps above set limits in pp_od_clk_voltage when ...
Status: RESOLVED NOTABUG
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/AMDgpu (show other bugs)
Version: unspecified
Hardware: x86-64 (AMD64) Linux (All)
: medium normal
Assignee: Default DRI bug account
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2018-05-07 13:12 UTC by tempel.julian
Modified: 2018-11-29 10:58 UTC (History)
0 users

See Also:
i915 platform:
i915 features:

Attachments

Description tempel.julian 2018-05-07 13:12:27 UTC 
I adjusted my pstates in "/sys/class/drm/card0/device/pp_od_clk_voltage" that the GPU voltage should never be above 900 mV:

(...)
5:       1126Mhz        880 mV
6:       1180Mhz        890 mV
7:       1209Mhz        900 mV

And in normal 3D applications "watch -n 0.5  cat /sys/kernel/debug/dri/0/amdgpu_pm_info" accordingly never reports more than 900 mV for VDDGFX.

But when I play a video in mpv with VAAPI/VDPAU, reported VDDGFX often jumps to 1100 mV, which clearly is undesired behavior. It sticks again to normal 900 mV as soon as I disable hardware decoding.

Now the question is, if the reported voltage is correct (then adjusting pp_od_clk_voltage has a bug) or if the real voltage is just reported wrong (than information in amdgpu_pm_info would be wrong).

Happens with both 4.17rc2 and drm-next-4.18-wip.
GPU is RX 560.
Comment 1 tempel.julian 2018-05-10 13:40:29 UTC 
I did some more tests and now I'm very sure that the voltage is just reported wrong when using UVD, and not being actually that high.
I also jumps to reported 1.1V without any pstate adjustments or overclock, while the ASIC's default maximum voltage is actually 0.986V.
Maybe the UVD got its own voltage of 1.1V, and that is sporadically reported instead of the core GPU voltage when UVD is used?
Anyway, not really important since it's just a little glitch and not a real problem. :)
Comment 2 tempel.julian 2018-11-29 10:58:44 UTC 
Closing, since it's harmless and the Windows driver shows the same behavior.

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.