Announcement

Collapse
No announcement yet.

PowerXpress Runtime Power Management Added

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • phoronix
    started a topic PowerXpress Runtime Power Management Added

    PowerXpress Runtime Power Management Added

    Phoronix: PowerXpress Runtime Power Management Added

    Open-source developers are adding run-time power management support to the open-source Radeon DRM graphics driver for supporting ATI/AMD PowerXpress laptops...

    http://www.phoronix.com/vr.php?view=MTQ2NjI

  • dffx
    replied
    Originally posted by Andrecorreia View Post
    this works like primusrun for nvidia? with bumblebee?
    No, this is only power management (ie, turns off the discrete GPU when it's not in use). This does not actually perform the act of GPU offloading. For that you want one of the following:

    1. When using fglrx (catalyst) you can use the common-amd branch of bumblebee. there are some tweaks to primus that will make it work (check the primus github)

    2. When using radeon (xf86-video-ati) you can use PRIME offloading with Xorg 1.14

    As I understand it the PXP power management as mentioned here is *only* for the open source drivers -- the official catalyst drivers have their own built-in power management. I am not sure when/where these options will be implemented, maybe someone else in this thread who knows a bit more can enlighten us.

    Leave a comment:


  • Andrecorreia
    replied
    how its work?

    this works like primusrun for nvidia? with bumblebee?

    Leave a comment:


  • Rakot
    replied
    Originally posted by agd5f View Post
    AFAIK, it works similarly on windows (i.e., user can select which GPU to use for a specific app, although there may be some default profiles for more common apps). X handles multiple GPUs in PX-like configurations just fine starting with xserver 1.14 or newer. See this page for how to use the GPUs in PX/Optimus laptops with X:
    http://nouveau.freedesktop.org/wiki/Optimus/
    Hello, Alex,

    I tried your patches on top of kernel 3.11.1+drm-next.
    My hardware is intel hd 4000 + mobility radeon 7750M.
    Here is my dmesg . So I didn't do anything on my dGPU after I booted my laptop. However it looks like dGPU was turned off and then turned on again. Could you explain please what exactly can trigger dGPU to turn off and turn on (DRI_PRIME only?) ?
    Also cat /sys/kernel/debug/vgaswitcheroo/switch shows following:
    Code:
    0:DIS: :DynPwr:0000:01:00.0
    1:IGD:+:Pwr:0000:00:02.0
    How can I check whether my dGPU is active?

    Thank you in advance for your answers! And many thanks for a great work!

    Leave a comment:


  • dffx
    replied
    Originally posted by thirtythreeforty View Post
    Does this mean good things for those of us with PXP 4.0, "Enduro"? Even AMD's proprietary drivers don't work most of the time on my laptop.
    Yes, this is precisely what it's for, but for use with the open source radeon module, not proprietary Catalyst.

    By the by, if you're interested in getting Catalyst to work (decently) with a muxless PXP setup, consider looking into the AMD branch of Bumblebee.

    Leave a comment:


  • thirtythreeforty
    replied
    Enduro

    Does this mean good things for those of us with PXP 4.0, "Enduro"? Even AMD's proprietary drivers don't work most of the time on my laptop.

    Leave a comment:


  • Adarion
    replied
    *does the Ballmer-dance*
    I have four words for ya:
    "Aaiiiih laaaaaaaaaaahhvv diss commpaniiiiiieeeeeeee... yeaaaaaaaaahhh!"
    Actually it's 5 words but I wanted to stay true to Steve.

    Now we just need things like this on classic mainboards in real computers. Y'know, having a nice APU with small graphics for KDE and console work and "on demand" a big dedicated GPU on a card in a PCIe slot for gaming or other hungry applications.
    Probably needs cooperation from mainboard vendors, also to get some kind of VGA out multiplexing done so you can use 2 chips on one interface of HDMI/DVI/VGA/DP/LVDS.

    Leave a comment:


  • agd5f
    replied
    Originally posted by Ericg View Post
    Hey Alex 2 questions for ya.

    How is PowerXPress / Optimus handled in Windows? Is it a window hint or does the driver just decide based on load of the integrated which to use? Or something else? (And can we use the same idea on Linux?)

    I know you're more kernel side.. but do you have any idea how letting X load multiple GPU drivers is coming along?
    AFAIK, it works similarly on windows (i.e., user can select which GPU to use for a specific app, although there may be some default profiles for more common apps). X handles multiple GPUs in PX-like configurations just fine starting with xserver 1.14 or newer. See this page for how to use the GPUs in PX/Optimus laptops with X:
    http://nouveau.freedesktop.org/wiki/Optimus/

    Leave a comment:


  • Ericg
    replied
    Originally posted by agd5f View Post
    The dGPU will get powered down automatically when it's not in use. So it should save some power if you weren't manually powering up/down the dGPU before. Perfornance shouldn't be affected.
    Hey Alex 2 questions for ya.

    How is PowerXPress / Optimus handled in Windows? Is it a window hint or does the driver just decide based on load of the integrated which to use? Or something else? (And can we use the same idea on Linux?)

    I know you're more kernel side.. but do you have any idea how letting X load multiple GPU drivers is coming along?

    Leave a comment:


  • agd5f
    replied
    The dGPU will get powered down automatically when it's not in use. So it should save some power if you weren't manually powering up/down the dGPU before. Perfornance shouldn't be affected.

    Leave a comment:

Working...
X