Bug 69395

Summary: [DPM] Forcing lowest power state when no display is attached is bad because of OpenCL
Product: DRI Reporter: darkbasic <darkbasic>
Component: DRM/RadeonAssignee: Default DRI bug account <dri-devel>
Status: RESOLVED MOVED QA Contact:
Severity: normal    
Priority: medium CC: mike
Version: XOrg git   
Hardware: Other   
OS: All   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
dmesg while bitmining
none
possible fix none

Description darkbasic 2013-09-15 23:54:34 UTC 
Actually I can't use my HD 7950 with KDE becuase of

https://bugs.freedesktop.org/show_bug.cgi?id=69341
https://bugs.freedesktop.org/show_bug.cgi?id=64297

so I attached my display the Intel HD 4000.

I still do bitcoin mining (OpenCL) on the 7950 and I notice it doesn't switch to higher power states anymore (everything works flawlessly if I attach the display to the 7950). Forcing the card to the lowest power state if no display is attached is a bad behavious because of GPGPU computing.

I'm using Linux 3.12-pre (drm-next).
Comment 1 Alex Deucher 2013-09-16 00:13:35 UTC 
Please attach your dmesg output when the radeon card is doing bitmining.
Comment 2 darkbasic 2013-09-16 00:34:18 UTC 
Created attachment 85884 [details]
dmesg while bitmining

Here it is.
Comment 3 Alex Deucher 2013-09-16 00:42:40 UTC 
What does the dpm debugfs output show when you are bitmining?  e.g.,
cat /sys/kernel/debug/dri/64/radeon_pm_info

Also does manually forcing the performance level to high help?  E.g.,
echo high > /sys/class/drm/card0/device/power_dpm_force_performance_level

Note, you may need to adjust the paths above depending on your system.
Comment 4 Alex Deucher 2013-09-16 00:50:23 UTC 
Created attachment 85885 [details] [review]
possible fix

Does this patch help?
Comment 5 darkbasic 2013-09-16 01:26:37 UTC 
When mining I get lowest power level (power level 0).
Of course if I echo "high" into power_dpm_force_performance_level I get power level 2 (the highest).
I will try attached patch tomorrow, thanks.
Comment 6 darkbasic 2013-09-16 11:54:08 UTC 
The fix worked, thanks.
I hope you can help me with my KDE issue too: it's a pity having to use the Intel HD 4000 which doesn't even support my monitor's resolution (2560x1600).
Comment 7 Mike Lothian 2013-10-10 10:50:01 UTC 
Has this been pushed upstream?

Will this also fix the muxless render provider scenario too?
Comment 8 Mike Lothian 2013-10-10 11:00:38 UTC 
Or maybe I should ask does this big affect more than just SI?
Comment 9 Alex Deucher 2013-10-10 22:29:48 UTC 
It's in my 3.13-next branch:
http://cgit.freedesktop.org/~agd5f/linux/log/?h=drm-next-3.13-wip
Comment 10 Vedran Miletić 2017-03-22 15:16:20 UTC 
Was this fixed?
Comment 11 Martin Peres 2019-11-19 08:38:10 UTC 
-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/drm/amd/issues/381.

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.