Radeon and Powerplay support

[sastraxi]

Is there any testing an end-user like myself could do to help push this along? Once I got it all working, the powerplay support did not have any ill effects on the rest of my setup.

[agd5f]

What's the status of powerplay as of today? Are the patches still to be applied manually or have they been integrated in newer versions of the radeon driver? Is an x600 mobility supported?

Many thanks for any clarifications, signor_Rossi.

The patches for R5xx will get merged eventually. R6xx is still being investigated. Pre-ATOM asics (like the x600) will get support if I have time down the road.

Alex

[signor_rossi]

What's the status of powerplay as of today? Are the patches still to be applied manually or have they been integrated in newer versions of the radeon driver? Is an x600 mobility supported?

Many thanks for any clarifications, signor_Rossi.

[bridgman]

Yeah, the titles seem to be misleading but the numbers seem to support what Alex said.

[agd5f]

They certainly did apply cleanly. However, I'm not sure what exactly the line is to put in the xorg.conf file.

Powerplay V2 table found - 5 entries
No thermal controller
Powerplay 0: Optimal Performance 
	engine:32400 memory:13500
	Voltage GPIO entry:20 Panel Refresh:60 temp range:0-0 pcie lanes:1
	PM mode attributes:
	ATOM_PM_MISCINFO_VOLTAGE_DROP_SUPPORT
	ATOM_PM_MISCINFO_VOLTAGE_DROP_ACTIVE_HIGH
	ATOM_PM_MISCINFO_ENGINE_CLOCK_CONTRL_EN
	ATOM_PM_MISCINFO_MEMORY_CLOCK_CONTRL_EN
	ATOM_PM_MISCINFO_ASIC_SLEEP_MODE_EN
	ATOM_PM_MISCINFO_THERMAL_DIODE_MODE
	FM Disable
Powerplay 1: Optimal Battery Life 
	engine:21000 memory:13500
	Voltage GPIO entry:20 Panel Refresh:60 temp range:0-0 pcie lanes:1
	PM mode attributes:
	ATOM_PM_MISCINFO_VOLTAGE_DROP_SUPPORT
	ATOM_PM_MISCINFO_VOLTAGE_DROP_ACTIVE_HIGH
	ATOM_PM_MISCINFO_ENGINE_CLOCK_CONTRL_EN
	ATOM_PM_MISCINFO_MEMORY_CLOCK_CONTRL_EN
	ATOM_PM_MISCINFO_ASIC_SLEEP_MODE_EN
	ATOM_PM_MISCINFO_DEFAULT_DC_STATE_ENTRY_TRUE
	ATOM_PM_MISCINFO_DEFAULT_LOW_DC_STATE_ENTRY_TRUE
	ATOM_PM_MISCINFO_POWER_SAVING_MODE
	FM Disable
Powerplay 2: Optimal Performance 
	engine:39200 memory:25200
	Voltage GPIO entry:20 Panel Refresh:60 temp range:0-0 pcie lanes:16
	PM mode attributes:
	ATOM_PM_MISCINFO_VOLTAGE_DROP_ACTIVE_HIGH
	ATOM_PM_MISCINFO_ENGINE_CLOCK_CONTRL_EN
	ATOM_PM_MISCINFO_MEMORY_CLOCK_CONTRL_EN
	ATOM_PM_MISCINFO_ASIC_SLEEP_MODE_EN
	ATOM_PM_MISCINFO_THERMAL_DIODE_MODE
	FM Disable
Powerplay 3: High Battery 
	engine:39200 memory:25200
	Voltage GPIO entry:20 Panel Refresh:60 temp range:0-0 pcie lanes:16
	PM mode attributes:
	ATOM_PM_MISCINFO_VOLTAGE_DROP_ACTIVE_HIGH
	ATOM_PM_MISCINFO_ENGINE_CLOCK_CONTRL_EN
	ATOM_PM_MISCINFO_MEMORY_CLOCK_CONTRL_EN
	ATOM_PM_MISCINFO_ASIC_SLEEP_MODE_EN
	ATOM_PM_MISCINFO_POWER_SAVING_MODE
	FM Disable
Powerplay 4: Balanced 
	engine:43200 memory:39600
	Voltage GPIO entry:20 Panel Refresh:60 temp range:0-0 pcie lanes:16
	PM mode attributes:
	ATOM_PM_MISCINFO_VOLTAGE_DROP_ACTIVE_HIGH
	ATOM_PM_MISCINFO_ASIC_REDUCED_SPEED_SCLK_EN
	ATOM_PM_MISCINFO_LOAD_BALANCE_EN
	ATOM_PM_MISCINFO_DRIVER_DEFAULT_MODE
	FM Disable
	ATOM_PM_MISCINFO_DYNAMIC_CLOCK_DIVIDER_BY_2
	ATOM_PM_MISCINFO_DYNAMIC_HDP_BLOCK_EN
	ATOM_PM_MISCINFO_DYNAMIC_MC_HOST_BLOCK_EN

That's what I'm getting; so would I put something like

Option "PowerPlayMode" "1"

... for best battery life and

Option "PowerPlayMode" "2"

... for optimal performance? I suppose I'll just try it and take a look at powertop, then I'll deal with the DRM not loading issue I'm having.

1 would give you the best battery life, and 4 would be the best performance.

[sastraxi]

They certainly did apply cleanly. However, I'm not sure what exactly the line is to put in the xorg.conf file.

Powerplay V2 table found - 5 entries
No thermal controller
Powerplay 0: Optimal Performance 
	engine:32400 memory:13500
	Voltage GPIO entry:20 Panel Refresh:60 temp range:0-0 pcie lanes:1
	PM mode attributes:
	ATOM_PM_MISCINFO_VOLTAGE_DROP_SUPPORT
	ATOM_PM_MISCINFO_VOLTAGE_DROP_ACTIVE_HIGH
	ATOM_PM_MISCINFO_ENGINE_CLOCK_CONTRL_EN
	ATOM_PM_MISCINFO_MEMORY_CLOCK_CONTRL_EN
	ATOM_PM_MISCINFO_ASIC_SLEEP_MODE_EN
	ATOM_PM_MISCINFO_THERMAL_DIODE_MODE
	FM Disable
Powerplay 1: Optimal Battery Life 
	engine:21000 memory:13500
	Voltage GPIO entry:20 Panel Refresh:60 temp range:0-0 pcie lanes:1
	PM mode attributes:
	ATOM_PM_MISCINFO_VOLTAGE_DROP_SUPPORT
	ATOM_PM_MISCINFO_VOLTAGE_DROP_ACTIVE_HIGH
	ATOM_PM_MISCINFO_ENGINE_CLOCK_CONTRL_EN
	ATOM_PM_MISCINFO_MEMORY_CLOCK_CONTRL_EN
	ATOM_PM_MISCINFO_ASIC_SLEEP_MODE_EN
	ATOM_PM_MISCINFO_DEFAULT_DC_STATE_ENTRY_TRUE
	ATOM_PM_MISCINFO_DEFAULT_LOW_DC_STATE_ENTRY_TRUE
	ATOM_PM_MISCINFO_POWER_SAVING_MODE
	FM Disable
Powerplay 2: Optimal Performance 
	engine:39200 memory:25200
	Voltage GPIO entry:20 Panel Refresh:60 temp range:0-0 pcie lanes:16
	PM mode attributes:
	ATOM_PM_MISCINFO_VOLTAGE_DROP_ACTIVE_HIGH
	ATOM_PM_MISCINFO_ENGINE_CLOCK_CONTRL_EN
	ATOM_PM_MISCINFO_MEMORY_CLOCK_CONTRL_EN
	ATOM_PM_MISCINFO_ASIC_SLEEP_MODE_EN
	ATOM_PM_MISCINFO_THERMAL_DIODE_MODE
	FM Disable
Powerplay 3: High Battery 
	engine:39200 memory:25200
	Voltage GPIO entry:20 Panel Refresh:60 temp range:0-0 pcie lanes:16
	PM mode attributes:
	ATOM_PM_MISCINFO_VOLTAGE_DROP_ACTIVE_HIGH
	ATOM_PM_MISCINFO_ENGINE_CLOCK_CONTRL_EN
	ATOM_PM_MISCINFO_MEMORY_CLOCK_CONTRL_EN
	ATOM_PM_MISCINFO_ASIC_SLEEP_MODE_EN
	ATOM_PM_MISCINFO_POWER_SAVING_MODE
	FM Disable
Powerplay 4: Balanced 
	engine:43200 memory:39600
	Voltage GPIO entry:20 Panel Refresh:60 temp range:0-0 pcie lanes:16
	PM mode attributes:
	ATOM_PM_MISCINFO_VOLTAGE_DROP_ACTIVE_HIGH
	ATOM_PM_MISCINFO_ASIC_REDUCED_SPEED_SCLK_EN
	ATOM_PM_MISCINFO_LOAD_BALANCE_EN
	ATOM_PM_MISCINFO_DRIVER_DEFAULT_MODE
	FM Disable
	ATOM_PM_MISCINFO_DYNAMIC_CLOCK_DIVIDER_BY_2
	ATOM_PM_MISCINFO_DYNAMIC_HDP_BLOCK_EN
	ATOM_PM_MISCINFO_DYNAMIC_MC_HOST_BLOCK_EN

That's what I'm getting; so would I put something like

Option "PowerPlayMode" "1"

... for best battery life and

Option "PowerPlayMode" "2"

... for optimal performance? I suppose I'll just try it and take a look at powertop, then I'll deal with the DRM not loading issue I'm having.

[agd5f]

Just to be clear: does your powerplay repository contain the recent stuff that makes everything work (like compiz if one also acquired the bleeding-edge X server) on R500 chips?

nope. It's pretty old, but the patches should apply pretty cleanly to the more recent stuff.

[sastraxi]

My hero! Thank you, Alex. I will be switching over tonight as that was the last feature I was missing! Say hello, decent-looking video...

Just to be clear: does your powerplay repository contain the recent stuff that makes everything work (like compiz if one also acquired the bleeding-edge X server) on R500 chips?

[sastraxi]

My hero! Thank you, Alex. I will be switching over tonight as that was the last feature I was missing! Say hello, decent-looking video...

[agd5f]

Would anyone be able to inform me of the status of Powerplay (if it matters, for an R500-generation X1400) in the radeon or radeonhd drivers? Have documents been released? Is there any support?

I've got is mostly sorted out in my personal repo for r5xx:

http://cgit.freedesktop.org/~agd5f/xf86-video-ati/log/?h=agd-powerplay

Only thing missing is some of the voltage stuff and drivers for the actual thermal chips if you want to be able to read the temperature (not necessary for operation). Unfortunately, at the moment you have to pick a powerplay mode statically in your config. This will be alleviated and you will be able to change things dynamically when we get get gpu objects in xrandr 1.3 or via sysfs in the drm once we have kernel modesetting.

Dynamic clock gating is already available via the DynamicClocks option in the radeon driver in git.

R6xx and newer are still being looked at as things have changed a bit there.

Alex