[arch-general] Kernel 2.6.32 and Radeon KMS

Sven-Hendrik Haase sh at lutzhaase.com
Tue Dec 8 11:33:12 EST 2009


On 08.12.2009 11:13, Xavier wrote:
> On Tue, Dec 8, 2009 at 6:52 AM, David C. Rankin
> <drankinatty at suddenlinkmail.com> wrote:
>   
>> On Monday 07 December 2009 09:05:18 and regarding:
>>     
>>> I'm confirming this as well. Mine fails at radeon/R200_cp.bin, though.
>>> Loading in rc.conf works alright but the GPU appears to get very hot. I
>>> have no diode on it but it just tries to burn a hole through my laptop.
>>> Again, KMS performance is just as bad as with kernel 2.6.31. I get
>>> better performance without KMS and the system runs a lot cooler in idle
>>> as well. My card is a puny rv250 (aka Radeon Mobile 9000).
>>>
>>>       
>> Sven,
>>
>>        I have the RS690M (x1200) card and the gpu used to get so hot with earlier additions of the 1.25x radeonhd driver that it was uncomfortable to rest my palms on the laptop palmrest. AMD really fsck'ed all of the laptop users when it dropped all support for pre-2400 series cards in March automatically converting many fairly new laptops (Q1 2008) to "Legacy" boxes. The fglrx driver did such a marvelous job with gpu powerdown and downclocking that thermal was never an issue and performance was great. Thankfully the radeonhd driver has gotten much better in power management in the past 3 months that I can use the palm rest again, but performance is still way low. Best we can  do is keep our fingers crossed -- or go shop nvidia gpu based laptops :p
>>
>>     
>
> I am not sure that's a solution at all.
>
> The only way to be sure about long term support is with open source
> drivers. And AMD/ATI position certainly looks better than Nvidia for
> that.
> Just compare both :
> http://en.wikipedia.org/wiki/Graphics_hardware_and_FOSS#ATI.2FAMD ->
> specs for all? the latest chipsets
> http://en.wikipedia.org/wiki/Graphics_hardware_and_FOSS#NVIDIA ->
> nothing,void,nil
>
> Now I know nothing about ati fglrx proprietary drivers, but maybe you
> should know the situation with nvidia.
> 190.42 drivers : only supports geforce 6 !! and newer
> 173xx drivers : support geforce 5 (and newer but no point)
> 96xx drivers : support geforce 4 and some geforce 2
>
> Note that for the recent Xorg 7.5 release, the 173xx and 96xx lagged
> two weeks behind to get compatible, leaving many users in doubt during
> that period.
>
> And anything older than that is considered legacy and no longer
> supported. It was the 71xx series which stopped working with Xorg 7.4
> release, quite a while ago...
>
> But maybe in term of open source support and power consumption, intel
> is doing better than ati ?
> I have to admit I don't know these two well. And I am sure things are
> not perfect there either.
> Just don't assume that everything is shiny in Nvidia's world, far from that.
>
>   
Well then, can I actually expect KMS to run slower than "regular" mode
(because that's what it does) or would it be faster if it worked right
at the driver's current state? I tested both kernel 2.6.31 and 2.6.32
and both give me about 500 FPS less in glxgears (for what it's worth).
Also, KMS frequently causes kernel panic and all kinds of other trouble.
Changing from vt1 to X is fast, though.
Any tips on this or is this actually the current state of KMS? Remember,
I'm using xf86-video-ati.

-- Sven-Hendrik


More information about the arch-general mailing list