Giving radeonhd a try...

[borkdox]

I decided to dump catalyst and dip into the opensource waters. I run archlinux. I took some PKGBUILDs from the AUR modified them to point to the latest MASTER code, and build DRM and RADEONHD from git. After couple of insmods, modifying xorg.conf, and restarting X, it is looking pretty good.

Video playback is on par with Catalyst, maybe even with less flickering. 2D feels much faster also.

On the downside, my card's fan is running louder than with catalyst for some reason... Maybe I warmed up the room with excitement? or maybe radeonhd code overwrites BIOS fan adjustment?

There is also no 3D accel and no Compiz yet. I still encourage anyone with Radeon 3000/4000 series card and unhappy with catalyst to try radeonhd and test it.

I have:
Intel C2D E6750
Intel Gigabyte P35 Motherboard
Powercolor Radeon HD4850
Archlinux (Xorg 1.5 + KDE 4.2.1)

[TechMage89]

The OSS driver still doesn't have 3d acceleration or power management for the hd2xxx and later cards. These are being worked on, and should be functional later this year.

Radeon vs Radeonhd shouldn't make a lot of difference, especially given that the r600+ code is virtually identical between the two drivers.

The lack of power management is probably why your fan is running so high: the chip isn't clocking down, so it's giving off a lot of heat even though it isn't doing a lot.

I don't have an r600 card myself to play with, but, budget permitting (I'm a bit strapped for cash at the moment) I'll be getting a hd4670 or perhaps even a 4850 in a few months. Hopefully basic 3d should be working by then, and they'll be working on shiny things like gallium3d support.

[borkdox]

Yea, I think in a few months radeon/radeonhd should be a solid pair of drivers for r600/700. I hope new ATI chips do not deviate much from the current architecture, so that the wait for opensource support is not as long as it has been for the current cards.

I wonder if there is any way i can modify the source and force it to clock down. Not really power management, just forcing the default IDLE clocks/fan-speed.

I was taking a look at the rhd_pll.c file, there are some power/clock setting functions here. I know C, but I am not familiar with writing drivers, or videocard internals. Maybe I learn a thing or two... hehe.

My card is indeed running at the maximum clocks

(II) RADEONHD(0): Default Engine Clock: 625000
(II) RADEONHD(0): Default Memory Clock: 993000
(II) RADEONHD(0): Maximum Pixel ClockPLL Frequency Output: 1200000
(II) RADEONHD(0): Minimum Pixel ClockPLL Frequency Output: 0
(II) RADEONHD(0): Maximum Pixel ClockPLL Frequency Input: 16000
(II) RADEONHD(0): Minimum Pixel ClockPLL Frequency Input: 6000
(II) RADEONHD(0): Maximum Pixel Clock: 400000
(II) RADEONHD(0): Reference Clock: 100000

[suokko]

Yea, I think in a few months radeon/radeonhd should be a solid pair of drivers for r600/700. I hope new ATI chips do not deviate much from the current architecture, so that the wait for opensource support is not as long as it has been for the current cards.

I wonder if there is any way i can modify the source and force it to clock down. Not really power management, just forcing the default IDLE clocks/fan-speed.

I was taking a look at the rhd_pll.c file, there are some power/clock setting functions here. I know C, but I am not familiar with writing drivers, or videocard internals. Maybe I learn a thing or two... hehe.

My card is indeed running at the maximum clocks

(II) RADEONHD(0): Default Engine Clock: 625000
(II) RADEONHD(0): Default Memory Clock: 993000
(II) RADEONHD(0): Maximum Pixel ClockPLL Frequency Output: 1200000
(II) RADEONHD(0): Minimum Pixel ClockPLL Frequency Output: 0
(II) RADEONHD(0): Maximum Pixel ClockPLL Frequency Input: 16000
(II) RADEONHD(0): Minimum Pixel ClockPLL Frequency Input: 6000
(II) RADEONHD(0): Maximum Pixel Clock: 400000
(II) RADEONHD(0): Reference Clock: 100000

I wouldn't touch clocks or voltge settings for your card unless you are ready to take high risk for destroying your graphics card. I jsut read a week ago about LM_sensors breaking some processors when they tried to detect system and touched buggy motherboard registers.

I would guess that is less likely with gpu from ati but you never know if you put in some bad values.

[highlandsun]

I've been running radeonhd from git for a couple weeks. It makes watching fullscreen h.264 video possible, which is a good thing. But yesterday I discovered a fatal downside - using OpenOffice 3.0 hangs the machine hard when I minimize an OO window, or do various other moves/resizes. So far OpenOffice is the only app I've used that exhibits this problem, and switching back to the default Ubuntu 8.10 radeon driver avoids the problem. Unfortunately it took me a while before I figured out the radeonhd driver was the culprit, and I had a deadline to meet on the presentation I was writing. what a pain...

[bridgman]

Can you try the latest radeon from git at some point ? That will tell us if the problem is in the acceleration code (which is mostly common between the drivers) or somewhere else.

[highlandsun]

I'll try to make time later this week to try that.

[highlandsun]

Well, I followed this post

http://www.phoronix.com/forums/showthread.php?t=15988

to install the Jaunty Xserver on my Intrepid system, and now I can't reproduce the problem. It's now using the radeon driver, not radeonHD. I was surprised to see this message when I started openoffice though:

violino:~> soffice
unknown chip id 0x95c4, can't guess.

I presume that it's some lower level library, and not OO itself looking at the video card's PCI ID ... ?

[agd5f]

violino:~> soffice
unknown chip id 0x95c4, can't guess.

I presume that it's some lower level library, and not OO itself looking at the video card's PCI ID ... ?

That message is from mesa (r300 3D driver). soffice must use GL for something. There is no 3D support for r6xx/r7xx hardware yet.

[bugmenot]

But yesterday I discovered a fatal downside - using OpenOffice 3.0 hangs the machine hard when I minimize an OO window, or do various other moves/resizes. So far OpenOffice is the only app I've used that exhibits this problem...

Yes, exactly! The driver is great, but sometimes locks up (with DRI enabled manually). I wrote an email to the radeonhd mailinglist:
http://lists.opensuse.org/radeonhd/2.../msg00148.html

I don't know if there is already a bug report in bugzilla for this issue. In any case, I can confirm this bug.