Mailinglist Archive: radeonhd (311 mails)

["Alex Deucher" <alexdeucher@xxxxxxxxx>]

On Fri, Feb 22, 2008 at 5:45 AM, Jerome Glisse <glisse@xxxxxxxxxxxxxxx> wrote:
> On Tue, 19 Feb 2008 12:38:20 -0500
>  "Bridgman, John" <John.Bridgman@xxxxxxx> wrote:
>
>  > Yep, no argument there. We are just in the process of pushing PowerPlay
>  > into the closed source drivers so we should have some experience to
>  > leverage soon.
>  >
>  > -----Original Message-----
>  > From: Matthias Hopf [mailto:mhopf@xxxxxxx]
>  > Sent: Tuesday, February 19, 2008 11:55 AM
>  > To: radeonhd@xxxxxxxxxxxx
>  > Cc: Bridgman, John; Deucher, Alexander
>  > Subject: Re: [radeonhd] any power saving modes possible?
>  >
>  > On Feb 19, 08 11:35:08 -0500, Bridgman, John wrote:
>  > > Aren't there AtomBIOS calls for setting engine clock ?
>  >
>  > You're correct. We just haven't looked at that, probably because we
>  > didn't have any use case for it so far.
>  >
>  > OTOH we don't know how this call is actually used, and which limits
>  > apply - there are entries in the FirmwareInfo, but they proved to be
>  > somewhat unreliable for pixel clocks. And we all know how picky PLLs can
>  > be when being programmed wrongly ;-) In contrast to pixel clock you also
>  > don't see a wobbly clock immediately, it just might corrupt some
>  > calculations.
>  > Also there might be side effects, like you might have to reprogram
>  > memory refresh etc.
>  >
>  > All in all this is something that we have deliberately left as something
>  > for future work. We could probably work something that works with the
>  > current information available, but there are more pressing issues ATM.
>  >
>  > CU
>  >
>  > Matthias
>  >
>
>  There is power register informations already :) see
>  http://www.g2inf.one.pl/~anszom/MBP-ATI/radeontool-qq.patch
>  it's a patch against radeontool
>  http://fdd.com/software/radeon/
>  I can save few watts with ./radeontool power low
>  I guess we can save more watts but haven't bothered to look
>  deeper (i don't use my laptop that much :))

This patch enables dynamic clock gating similar to that the
dynamicclocks options does on older radeons.  However, we'd need to be
careful of any errata as the old ones were full of it (you of anyone
should know that ;)

I'll take a look this weekend and see if I can't get some patches out.

Alex
-- 
To unsubscribe, e-mail: radeonhd+unsubscribe@xxxxxxxxxxxx
For additional commands, e-mail: radeonhd+help@xxxxxxxxxxxx