AMD/ATI PowerPlay-Support in radeon(ati) or radeonHD?

[Vedran Rodic]

On Tue, Apr 8, 2008 at 12:19 AM, Alex Deucher <alexdeucher at gmail.com> wrote:
>
> On Mon, Apr 7, 2008 at 2:17 PM, Vedran Rodic <vrodic at gmail.com> wrote:
>  > On Sat, Apr 5, 2008 at 2:49 PM, Vedran Rodic <vrodic at gmail.com> wrote:
>  >  > >  I've just added atombios support for the "DynamicClocks" option on
>  >  >  >  r5xx/r6xx cards to the radeon driver.  It should enable dynamic clock
>  >  >  >  gating and static power management if the chip supports it.  This is
>  >  >  >  not full powerplay support, just one aspect.  Also, the bios may set
>  >  >  >  some or all of this up at post so, YMMV.
>  >  >  >
>  >  >
>  >  >  Great, thanks.
>  >  >
>  >
>  > >  fglrx driver is definitely doing something on rv670.  I still haven't
>  >  >  tested the power usage with "ati" driver though, but will later today.
>  >  >
>  >
>  >  Unfortunately, this option doesn't work for HD3850, it doesn't save
>  >  power. Also VT switch tends to lockup the machine, try to do it a
>  >  couple of times and it surely will. It also doesn't detect all the
>  >  video modes (higher resolutions and refresh rates). GIT version from
>  >  today, Xserver 1.4.0.90 xserver-xorg-core
>  >  (1.4.1~git20080131-1ubuntu6).
>  >
>
>  I put a rough cut of the initial powerplay support in my personal tree here:
>  http://cgit.freedesktop.org/~agd5f/xf86-video-ati/log/?h=agd-powerplay
>  it's builds on Dave's previous tree.  from the commit log:
>
>  Support for adjusting sclk, mclk, and pcie lanes. pcie lane support is
>  commented out at the moment as the results are a bit tweaky on my box.
>  If you try it let me know how it goes. voltage support and version 4.x
>  powerplay tables are not yet supported

Unfortunately I only have a 3850 available for testing right now and
that chip has version 4 powerplay tables. I guess that this means that
for the 3850 just using the DynamicClocks option doesn't do anything
since the driver doesn't know which frequency it can set? The
structure of the data ATOM_POWERPLAY_INFO_Vx and the structure
ATOM_POWERMODE_INFO_Vx look pretty similar to one another, so I guess
that v4 tables are also not that different? Anyway, here is the
-verbose 9 log for your agd-powerplay branch on the HD3850, just for
the reference: http://nskunca.pbf.hr/~vedran/xorg2.log VT switch
problem and the video mode detection problems remain on this version
of course.

I hope I'll have a X1950 available for the weekend. Although power
saving for that card already works well even in the default BIOS and
vesa driver mode, so that is not so interesting to test. But who
knows, maybe I can get it to use even less power than in Windows, and
that would be great.

Thanks
Vedran