I await my Mobility MXM 3870 to arrive in a week from Texas : ) I paid ~100€ for it.
It would surely not hurt if AMD allows dynamic powermanagement to materialize, I don't want to burn another mobile card.
Just to clarify that, if I understand bridgman correctly the code for power-management is currently under legal review. So it may be that it will be released tomorrow, but it also may be that it will be released never.
Well UVD support got released and I also thought that would be never, so who knows?
I don't have any inside info, just what bridgman has posted on this forum under different topics, that they are working on it but can't say how much it will take or if they will ever be able to release it.
It's so primitive thing. When GPU in use, switch higher power setting. But AMD OSS still flicker when changing GPU clocks and doesn't reduce GPU clock/voltage/memory clks as it needed...
I think it need to be in high priority list. Power is not FREE here. (~$0.15 Kw/h).
Default GPU power wastes +100w -> 2.4kw per day and 876 kwH per year. It cost $0.36 per day -> $131 per year. And it's just "wasted" power due AMD drivers!
Hey, I can buy a new GPU for this price that could adjust it's power!
Yes, video acceleration was not as critical as power management. If AMD don't do something about broken SI and PM, they will stagnate back into "worthless" abyss pretty fast. Even given the situation they have today. Common, even fglrx fails more piglit tests than radeon, do something.... My mobility HD3870 arrived from grand prerrie, TX; please, my notebook card needs some PM love!
For me, say, 5770 could scale from 150MHz to 850MHz. But opensource driver STILL suxx at automatic power management.
Full speed on that card (echo high > /sys/class/drm/card0/device/power_profile) is 850 mhz and is power hungry for sure.
Mid-range clock (medium profile) is 600MHz. And only barely colder than full speed.
Low clock (echo low > /sys/class/drm/card0/device/power_profile) is 150 MHz. This one provides decent powersave.
However, dynpm method only cycles between mid and high clocks. Should I admit this works very poorly? Cycling between 600 and 850MHz makes little sense. OTOH, forcing low profile (means 150MHz GPU clock) works like a charm, temp goes down by 15C.
Isn't it really silly that driver technically could do the trick but should be manually pointed to do so like some blind kitten? Btw, driver can go to low freq when display is powered down. But would immediately upclock GPU to 600 MHz if I'm about to use desktop, even if there are no any loads demanding it at all. That's not what I would call proper power management for sure. Proprietary driver downclocks that GPU to 150MHz correctly without manual intervention. I suspect that many users are just facing some similar things.