On Wed, Oct 8, 2008 at 2:46 PM, Yang Zhao <yang@yangman.ca> wrote:
2008/10/8 Rafał Miłecki <zajec5@gmail.com>:
Did you compare power consumption with and without patch?
I gathered some numbers through ACPI battery state, but I frankly don't trust it.
On my T60, no wireless, lowest brightness, and CPU at lowest stepping mode yielded ~15W with stock radeonhd. Switching on PowerPlay dropped this number slightly, but I personally don't consider it statistically significant. Manually switching to mode 0, which down-clocks GPU and memory as low as it will go, gives 0.5 to 0.8W of saving.
It seems that the hardware already initializes to a fairly efficient state, at least for me, regardless of PowerPlay.
Many laptop bioses enable basic power saving stuff by default. There are atom command tables to do this: DynamicClockGating EnableASIC_StaticPwrMgt Take a look at the radeon driver to see how to use them.
As I understand it, dynamic clock gating of the memory controller block and dynamic voltage drop can yield more, but these areas are proving problematic.
Much of the clock gating stuff is handled by the tables I mentioned above. The voltage stuff involves gpio twiddling and can be problematic, I haven't really had a chance to play with that in depth yet.
How hard will it be to introduce R600/R700 support in your patch?
I've been told PowerPlay was changed entirely for R600, so it's hard to say until they give us those headers. In any case, the AtomBIOS portion will most likely just work, or do nothing; the existing bit-banging parts won't affect and can't be re-sued for R600+ PowerPlay at all.
The r6xx and newer powerplay bits got a lot more complex and the necessary bits are no longer int atombios.h. Alex N§˛ćěr¸yéZÚuę' ߎ˹Ęâmę)z{.ąę+Z+i×bś*'jW(f§vÇŚj)kiר~ éi˘§˛ë˘¸