You seem to be confusing "Windows development" with "Catalyst development" and "Linux development" with "Open source development". Remember that much of the code for the Catalyst driver is shared between operating systems.
Yes it looks like. But think on just Catalyst driver for Windows and for Linux... How much effort spent on "Windows only" part and how much for "Linux only" parts.
Using Radeon HD 6850, Catalyst for windows is usable and looks like error free from first day... But after a year, Catalyst for Linux still gives errors like reload gnome shell, force you log out etc. Yes it's faster than open source alternative but with full of bugs.. Means not quite usable!
Not Catalyst bug: http://ati.cchtml.com/show_bug.cgi?id=432#c14
Originally Posted by Death Knight
For me dynpm works and reclocks my GPU very well, causing a EPIC power usage drop. However,
- It is not enabled by default anywhere and it's completely wrong to force users to deal with things like this on their own.
- Sometimes it fails to switch clocks in a timely manner so screen could blink once or twice when some 3D game starts. Not a huge problem but could be better.
And profile based management is simply odd. Why can't guys just blacklist low mode for defective cards, etc? For me on medium-end GPU mid power profile is still too much. You see, it makes no sense to display desktop at increased GPU speeds. And low power profile sets lowest clocks and hence least heat and least cooler's noise.
Strange AMD guys can't sort all this weirdness out. I suspect it shouldn't be too much work, no?
For me dynpm doesn't put the clocks as deep as the low/mid profile. But let's see if it is better with 3.5 as soon as it works again.
PM works on my HD4850, but itīs not perfect. Some flickering now and then, dynpm is not as responsive as I think it should be, and I have experienced a few lock ups.
It also works on my laptopīs X200, but here dynpm just flat out locks up the system forcing me to rely on profiles.
I am casting healing rays into this thread and awaiking it from necro state.
My X1900 did flicker a LOT in Ubuntu Unity on dynpm. A lot is - 4-6 flickers per second when I open dash.
Similarly HD5850 flickers on dynpm.
Right now I am levelraging purchase of either Mobility 4650 or 3870, to replace my burned X1900, but I sure way will have to deal with the same damn issue....
I can't purchase nvidia now, I was amazed by radeon. :/ What should I do? My notebook accepts only MXM 2.1 III slot.
My question is - can we somehow solve the flickering issue? What state does dynpm enter is not so problematic, compared to this flickering effect.
This is sure the number one issue with opensource radeon now.
Last edited by brosis; 05-02-2013 at 02:19 PM.
bridgman was saying that there is a possibility to improve dynpm by doing the core and memory reclocks in different vertical blanking intervals. This would increase the chances that they complete during the VBI, but there is no guaranty. And since the proper power management exists and is under review at AMD noone is willing to work on improving dynpm. So until the proper power management code is released by AMD it would be better to stick to using profiles.
Profiles are vastly superior to dynpm at this point and everyone should use them.
Low profile is enough for pretty much everything (desktop effects, etc.) and you only need to manually put it in a high state when watching HD video or playing games.
Like Ansla says, proper, working, correct dynamic PM is pretty much written and is just awaiting release. Until then, just stick to the profiles.
05-03-2013, 03:09 AM
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.