Summary: | power_profile not working for HD5650 | ||||||
---|---|---|---|---|---|---|---|
Product: | DRI | Reporter: | Kamil Bar <nevehanter> | ||||
Component: | DRM/Radeon | Assignee: | Default DRI bug account <dri-devel> | ||||
Status: | RESOLVED MOVED | QA Contact: | |||||
Severity: | major | ||||||
Priority: | medium | ||||||
Version: | DRI git | ||||||
Hardware: | x86-64 (AMD64) | ||||||
OS: | Linux (All) | ||||||
Whiteboard: | |||||||
i915 platform: | i915 features: | ||||||
Attachments: |
|
Description
Kamil Bar
2013-05-28 16:54:57 UTC
Created attachment 80495 [details]
dmesg | grep radeon after drm.debug=2 and some profile switching
And I need to correct my first post, on auto/low/mid/default profiles voltage is 900 mV, but if I switch to high my screen flick, and it goes to 1000 mv. I cannot attach video bios, because my notebook has vbios built-in into notebook's bios, but if U propose how to get it, I could try ( note: on windows I cannot dump it too) but the notebook bios itself is here: http://h10025.www1.hp.com/ewfrf/wc/softwareDownloadIndex?softwareitem=ob-100446-1&cc=us&destPage=siteHome&dlc=en&lc=en&os=4063&product=4217144&query=dv6-3030ew&tool= I have found an issue, my vbios default clocks are set to 150000 and 157000, and my switching profiles aren't working because of this lines in radeon_pm.c: if (sclk > rdev->pm.default_sclk) { sclk = rdev->pm.default_sclk; } and if (mclk > rdev->pm.default_mclk) { sclk = rdev->pm.default_mclk; } when I try to switch to 550000Mhz ( pass sclk == 55000 ) for clock this compares it with default ( rdev->pm.default_sclk == 150000 which atombios provides from vbl ), and then kernel switch this request to default. So, anyone know solution to this, with possibility to push the fix to linux main branch, so I and possibly many other people can use profiles in future kernels without recompiling the kernel and apply patch? It would require modifying radeon_atombios.c and it's methods, I could try to fix it someway, but it will require testing on many devices. You can try my patch *if* you have good cooling as Alex indicated (on IRC) he would not be applying the patch due to potential overheating. We're all still awaiting the lawyers "review" of improved power manangement code. http://lists.freedesktop.org/archives/dri-devel/2013-May/039247.html My patch didn't include memory clocks as I'm running an APU rather than a discrete card so could not test. So if you add a max_mclk setting, note to guard it with an "if !(rdev->flags & RADEON_IS_IGP)" in radeon_pm_get_max_clocks(). Unsurprisingly my frame rates improved over 100% from the default 200MHz but idles back down on DPMS off with the auto power profile. Thanks, it works fine now, I added memory clocking, and I will try to create OC possibilities using debugfs. I have two questions, will new system have overclocking capabilities, in ex. passing values to debugfs by radeon_oc_sclk/radeon_oc_mclk/radeon_oc_volts, I have tried hardcoding overclocked values for my gpu, and I can overclock my HD5650 from 550/800 with 1.00V to 750/950 with 1.05V without problems and it is stable ( I can give more, but I don't want to risk burning it with too much voltage, but we should have this features ), radeon_pm shows right clocks with this values. And why REDWOODs are that slow, especially with 1080p resolutions? -- 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/332. |
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.