Radeon Driver Power Management Has Room For Improvement

[d4ddi0]

Not long ago there was no power management at all


It is easy to see these specs and look at it as a dig against the work tht has been done.
I tend to see it the opposite way. With power benchmarking we can see, well, where are we now, where have we come from, and are we improving?

Hopefully, at some point the open source drivers will get to the point where catalyst can be abandoned on Linux, even for commercial/industrial support. But we ain't there yet.

[chronniff]

Great article.

I'm curious now what effect the different power levels have on the desktop? (e.g. Gnome shell). I rarely ever run 3d games so I don't need that much horsepower, just enough to composite my desktop smoothly. Curious now whether running low power mode is enough.

It's not really enough on my 5770, on 2.6.38 and stock drivers in natty it would freeze up for a second or two, and I'm now using 2.6.39 with git snapshots from this ppa:

http://phoronix.com/forums/showthread.php?50038-Updated-and-Optimized-Ubuntu-Free-Graphics-Drivers

and there are no freeze ups, but desktop effects are extremely jerky, and unpleasent (at least with unity and the compiz rewrite). Actually things like scrolling a maximized browser are ok, and youtube 720p videos works fine, but 1080p flash vids get jerky. haven't tried out regular video files. But if u want a nice fluid desktop its not really enough

[ahlaht]

Powersaving

Curious now whether running low power mode is enough.

That of course depends on hardware and software used.

For me, "low" is usually enough. My preferred solution even in the long term is to have a keyboard shortcut for performance mode. Normally I keep everything forced to powersaving, but if there is a sudden need for more speed I press a key that activates performance mode.

[HokTar]

It is easy to see these specs and look at it as a dig against the work tht has been done.
I tend to see it the opposite way. With power benchmarking we can see, well, where are we now, where have we come from, and are we improving?

Hopefully, at some point the open source drivers will get to the point where catalyst can be abandoned on Linux, even for commercial/industrial support. But we ain't there yet.

While being positive is nice but the problem is the following:
Alex did some serious work about a year ago to get to the stage with pm where we are now. That code has been practically not touched in about 9 months (if memory serves). So the progress was promising but there are basically no advancements for a long time now.

Another interesting thing is that there is a separate logic (as noted earlier in this thread) for pm which is not used in -radeon at all. I assume that programming that is more difficult than what has been done so far, but still. As far as I know documentation is not available for those parts.

I appreciate the work Alex has done, because at least I can manually set to low my gpu but dynpm never really worked for me. I can't wait for the new oss guys at AMD to start working... I hope that will bring some improvements.

[Drago]

That of course depends on hardware and software used.

For me, "low" is usually enough. My preferred solution even in the long term is to have a keyboard shortcut for performance mode. Normally I keep everything forced to powersaving, but if there is a sudden need for more speed I press a key that activates performance mode.

Name that button "TURBO"

[ahlaht]

Name that button "TURBO"

No wonder it didn't feel like a fresh idea.

Well the difference here is that this button is more easily accessible. And ancient 386 computers were always noisy with and without coprocessor.

[bridgman]

I appreciate the work Alex has done, because at least I can manually set to low my gpu but dynpm never really worked for me. I can't wait for the new oss guys at AMD to start working... I hope that will bring some improvements.

I feel compelled to mention that we are talking about generic open source code, annotated with a summary of required changes, and not waiting for any additional info from AMD. Anyone can work on that code, not just AMD developers.

[Adarion]

The test is sadly but definitely true. If I might share my own numbers on a Sapphire HD 3870 (passive cooling, 512M).
The overall system usage at idle is 5W lower with fglrx on recent kernel (no pcie_aspm=force yet) than Catalyst on WinXP, and this is still about 5W lower than xf86-video-ati/Kernelside driver part. which makes it a 10W difference, idle. CPU etc. should be about the same so I think these 10W are purely GPU side.
(For absolute numbers I had to plug the power meter inbetween. Didn't notice that much difference (as expected) with just the onboard HD 3300 (64M sideport).)
So even on the somewhat older RV670 there is room for improvement. Still it is much better than my test some months before where I had the GPU running at full power/speed all the time with the free drivers and that was like 30W idle vs. 70W idle.

[HokTar]

I feel compelled to mention that we are talking about generic open source code, annotated with a summary of required changes, and not waiting for any additional info from AMD. Anyone can work on that code, not just AMD developers.

While this is true - assuming that the mentioned "pm logic" is documented (which I admittedly don't know) - it still feels that not working on this is pretty much defeating the whole effort.

I mean, all right task #1 is enabling chips. I think no one ever questioned that. But what is #2?
For many it might be performance - well, they won't use the oss driver in the foreseeable future if they are serious and not just occasional gamers.
Maybe features? Sure, that's nice but too many things are lacking, Alex alone clearly does not have enough time to do all that so why pick 1 and forget about the rest? I think it's reasonable that you are not doing this.

So I guess after enabling basic functionality - which seemed more or less straightforward for the past few generations due to small/moderate changes - in my opinion the next vital thing would be pm. Especially in the era of mobile devices and the Fusion APU which is best known for its low consumption.

So burning power unnecessarily with the APUs seem to defeat the whole purpose and should be a key focus for AMD. Hence Average Linux Joe buys a new machine, installs $HISFAVOURITEDISTRO and he doesn't care anymore 'cause desktop effects work and there's plenty of battery time remaining.

Well, I'm just an outsider so what do I know about AMD's aims and goals but I think I heard somewhere that it was to give good out-of-the-box experience for Linux users. However, halving the battery life out-of-the-box will not please many users... so what I described still makes sense to me.

Please point out my mistakes and misunderstandings if you have time. Thanks!

[Death Knight]

I mention from that in AMD OpenSource forum couple of time.
I also set low to my GPU every time I boot.

And its really weird that why GPU low is consumes higher than fglrx in idle state.
I think it's needed to be top priority thing since all open source stack users just wasting their electricity and I guess it's simple to fix than trying to put better 3D stuff.