Poor graphical performance

[Beralus]

[root@compname:/mnt/ext4/Downloads/sauerbraten] # lspci | grep VGA
00:02.0 VGA compatible controller: Intel Corporation Mobile 945GME Express Integrated Graphics Controller (rev 03)

---------------

[root@compname:/mnt/ext4/Downloads/sauerbraten] # rpm -qa |grep intel
xorg-x11-drv-intel-2.9.1-1.fc12.i686

[me@compname:~] $:uname -a
Linux compname 2.6.32.12-115.fc12.i686.PAE #1 SMP Fri Apr 30 20:14:08 UTC 2010 i686 i686 i386 
GNU/Linux
HP-530 notebook PC

My problem is that:
very poor graphical performance, which slows down everything(or white or colorod big triangles on screen ) when i use some applications (3D or similar , for example Google Earth)
Any suggestion for solution?

[tripleninez]

hmm...

well first off, you should know that the Intel graphics stack in linux sucks...

however, that being said you still might be able to increase performance of the card.

myself, i use the radeon OSS driver. if i setup xorg.conf (which fedora doesn't use by default, as it
auto-configures x11 on boot) then i am able to change around settings and switch modes.
I can't offer you tips on your hardware directly, but if you google around i am sure there are Intel graphical
tweaks and different modes for that driver/xorg....

I too am having some GFX problems right now, i was hoping by fedora 13 (which i just installed) would have fixed them, but apparently not.... i currently have to disable 3d acceleration as of kernel 2.6.33 ,
i have been waiting for a while to see Radeon and kernel-mode-setting to be fixed, but it is still garbage.

are you using 2.6.33???

if so, you are probably using KMS - which by all popular reviews isn't all that good yet. in some cases,
people are experiencing upto a 25% LOSS in graphical performance...

i believe that you can still use UMS - user mode-switching, which may speed up your graphics. however, because i do not use intel, you will have to search around.

I can't use either right now

i wish upstream hadn't pushed KMS through so fast. now, so many people are having big issues
under the new methods/drivers/implementatons. i am disappointed with the linux kernel team right now, and if these problems aren't resolved soon, i will stop using linux on my laptop, until a buy a new one...

sorry i can't help more, but i just don't know enough about Intel graphics.

n

[Zanpactou]

Originally Posted by tripleninez

well first off, you should know that the Intel graphics stack in linux sucks...

More precisely, the i915 driver has sucked since it was switched to kms.

Originally Posted by tripleninez

are you using 2.6.33???

if so, you are probably using KMS - which by all popular reviews isn't all that good yet. in some cases,
people are experiencing upto a 25% LOSS in graphical performance...

i believe that you can still use UMS - user mode-switching, which may speed up your graphics. however, because i do not use intel, you will have to search around.

Nope. ums for i915 was removed before 2.6.32 was released. A while ago.

Originally Posted by tripleninez

i wish upstream hadn't pushed KMS through so fast. now, so many people are having big issues
under the new methods/drivers/implementatons. i am disappointed with the linux kernel team right now, and if these problems aren't resolved soon, i will stop using linux on my laptop, until a buy a new one...

Eventually, things will get better because the xorg developers seem to have a strong lead and appear to have got their act together of late.
Yes, I read that Linus was unhappy about the nouveau developers removing ums from their driver. I only wish someone had spoken up when the Intel developer removed ums from the i915 driver. I think it was the fact that he did it sneakily and didn't announce it in his commits which he pushed into his next branch ready for Linus to merge, that no one noticed or realised that it was a problem until too late. Some Intel cards did not and still don't work with kms.

There is a newer xorg-x11-drv-intel which has been released which might provide a tiny bit of joy for you but you'll also want the latest lib drm. I don't know if either are in Fedora testing yet but the kernel drm and driver code which works best to match the latest xorg-x11-drv-intel driver is in the 2.6.35 rc kernels right now. It's maybe not essential but may well also provide more improvement when kernel 2.6.35 is released.

It's a waiting game basically and plenty of others are in the same boat, including me.

Thanks,
Zanpactou

[tripleninez]

i hope this guy finds your info useful. i didn't realize that Intel had fully-removed UMS. I was able to use it with radeon.
2.6.33 it doesn't work though, if you enable compiz, mutter, etc. stops dead in it's tracks. i am FORCED to disable any real acceleration.

I'm really surprised that they took this approach, so many users have been completely screwed over. i have filed bug reports in several locations, talked to devs and after all this time. still no solution. :|

(and i do not consider "nomodeset" or any other boot parameter to be a solution, i consider that to mean that
the system is broken, which it is...)

it wasn't a huge deal, as i use Fedora for proaudio, and GFX and realtime audio aren't a good pairing, although they SHOULD be, as they are in any other OS.

however, it bugs me that i don't have any choice, or "freedom" which people often associate with using linux,
and the open source idealogy. don't get me wrong, i ABSOLUTELY love linux. but at the moment, i am not impressed
at all.

freedom, ya right. how about "stuck" instead, because things were pushed upstream, and into distro's that aren't
even close to being ready...

...oh well.

[Zanpactou]

Yeah, for the developers, it's like they're damned if they do and damned if they don't.
By not removing ums from their drivers they are stuck maintaining code that they don't even use, split into two different things which do similar things but you can only use one at a time.
On the other hand, removing ums when kms doesn't work for the hardware breaks what was once working.
It's not a great situation but kms is a pretty important software technology which users want.
It must have been too difficult to fork the working ums drivers I imagine. That doesn't at all excuse it though.

Thanks,
Zanpactou

[tripleninez]

Originally Posted by Zanpactou

Yeah, for the developers, it's like they're damned if they do and damned if they don't.
By not removing ums from their drivers they are stuck maintaining code that they don't even use, split into two different things which do similar things but you can only use one at a time.
On the other hand, removing ums when kms doesn't work for the hardware breaks what was once working.
It's not a great situation but kms is a pretty important software technology which users want.
It must have been too difficult to fork the working ums drivers I imagine. That doesn't at all excuse it though.

Thanks,
Zanpactou

True enough. especially the "damned if they do, damned if they dont" part.
it's too late now, but i wish they hadn't made KMS manditory until 2.6.35 or something.
because ATI doesn't make drivers for my card (firegl) and the new stack for me is garbage. which hopefully, will change.

it must be so difficult for the kernel developers sometimes....but, im sure it will all be worked out soon!

later

ninez