AMD Has Massive Radeon Patch Set - Power Management!

[jrch2k8]

Yeah, with the oss. haven't tried KDE in a long while, but gnome-shell is just bad. I used to keep it at medium power settings, but not much improvement with max either. I tried everything that archlinux had to offer in terms of docs, no performance improvements.
Have you tried gnome-shell recently? how did it work for you with the oss, and which distro, please

with the new kernel release i will definitely make a new test, meanwhile my laptop is bricked because i tried to activate the IGP in the bios to compare performance, when the laptop does not support dual graphics, haha. If I had good drivers, I would have never done that

i chop one of my testicles that either you have kwin using xrender+native[worst possible scenario] or something is messed up in your drivers and is using llvm/pipe as fallback.

post glxinfo and an screen of your systemsettings/Desktop effects/advanced

[atmartens]

I bet a lot of these improvements are being driven by AMD's desire to land the Steambox. Then they'd be in the WiiU, PS4, XboxOne and Steambox, wrapping up all of the big names in video games.

[jrch2k8]

how can i monitor now the freq since /sys/kernel/debug/dri/0/radeon_pm_info is not there anymore?

[barti_ddu]

First of all, my sincere thanks to all devs/staff that made it happen!
It is cool, ubercool indeed.


As for testing, It seems I encounter the same thing as Vim_User did: the GPU clock never goes up, even under load.

Tested with wip-2/wip-4 kernel (64-bit), the kernel driver is compiled as a module and loaded with dpm=1 parameter.

power_method and power_dpm_state say dpm and performance respectively.

I've also failed to find anything human-readable in debugfs regarding clocks/voltages.

The card is 6670 with 1GB of GDDR5 memory (sapphire ultimate series).

Just a stupid thought, can it be related with this card being fanless (like some thermal control detecting fan speed is not going up and preventing clocks going up)?

On the positive side, the desktop feels a bit snappier than with low profile (especially seen when dragging windows) and ~1.5-2 degrees cooler.

[Scias]

No luck here on RV635/HD3650 laptop. While the boot occurs normally, the screen gets black and the system hangs when the login prompt appears.
However if I remove the radeon.dpm parameter, it works fine (KWin/3D/DRI...) minus the power management.

[GreatEmerald]

i chop one of my testicles that either you have kwin using xrender+native[worst possible scenario] or something is messed up in your drivers and is using llvm/pipe as fallback.

post glxinfo and an screen of your systemsettings/Desktop effects/advanced

KWin and system settings on GNOME 3?..

I bet a lot of these improvements are being driven by AMD's desire to land the Steambox. Then they'd be in the WiiU, PS4, XboxOne and Steambox, wrapping up all of the big names in video games.

I bet they aren't. I bet Valve doesn't care whether the drivers are open or not in production devices. I also bet that the whole legal review process of this all started before the SteamBox was even revealed.

[droste]

As for testing, It seems I encounter the same thing as Vim_User did: the GPU clock never goes up, even under load.

Tested with wip-2/wip-4 kernel (64-bit), the kernel driver is compiled as a module and loaded with dpm=1 parameter.

power_method and power_dpm_state say dpm and performance respectively.

I've also failed to find anything human-readable in debugfs regarding clocks/voltages.

Same for me (wip-4). "power_method" says "dpm", "power_dpm_state" says "performance" and never changes and I have no way to actually tell if it's changing clocks because of the missing "radeon_pm_info" in debugfs. I tested with and without compositing.

[gigaplex]

Thanks for nothing, since I've had to use llvm for months on my HD7870 ... And so within, say, another 6 months, I might have reasonable power usage. Cool ... Next time, I'm still planning on buying Intel instead, like one of the R series CPUs.

Your comment makes absolutely no sense. Even if the driver supported the power management correctly, the HD7870 would still consume way more power than an integrated Intel chip. The Intel chip would also be severely underperforming in comparison, even with the R series. You do also realise the R series chips are soldered onto their motherboards, they don't come in a socketed version, right?

[SafetyNinja]

Window User Told Me

Window User Told Me:

fglrx, mesa, xorg, MIR, xMIR, wayland, LLVM, Gallium, Gallium3D, Gallium2D, Clang, UVD, NVDPAU, BumbleBee, Prime, r600g, WTF?! Are You guys NUTS??? I just use windows and my graphic works. I tried many Linux types from distrowatch and none of the Linuxes work. Either they don't boot or crashe. Either loud fans and hot or no picture at all. And video is choppy and when I drag windows they break weirdely.

[timothyja]

Window User Told Me:

fglrx, mesa, xorg, MIR, xMIR, wayland, LLVM, Gallium, Gallium3D, Gallium2D, Clang, UVD, NVDPAU, BumbleBee, Prime, r600g, WTF?! Are You guys NUTS???

Windows has underlying technologies too you just generally dont know about them or what their name is. This is what makes Linux interesting, if you want to you can learn exactly how your system works. If you dont you can just ignore all this just like you would on windows as many user already do.

Either loud fans and hot or no picture at all. And video is choppy and when I drag windows they break weirdely.

Well you will be able to tell them that these patches, the previous video acceleration patches and Wayland fix all these issues. All should land in the next releases of most major distributions.