HOWTO: Jaunty Intel Graphics Performance Guide

[zevans]

OK thanks for the clarifications!

There will be absolutely no EXA support, even in x-updates PPA, very soon. This is because the pre-release driver touted in xorg-edgers has been completed and has now been pushed to 'stable' and will make it to x-updates very soon.

I agree this is good news and gives much better focus.

I was asking on behalf of users who use Jaunty as-shipped who are still suffering from freeze-bugs and so on. Are you saying that 2.8 will go into jaunty-updates? (This is not the same thing as x-updates, or did you mean "x" to stand for [jaunty|karmic] ?)

Users will be rid of their EXA support whether they like it or not, no matter which of the two PPA's they use.

I wasn't talking about PPAs - I was talking about Joe User who expects Ubuntu to work properly on his netbook without having to mess with additional PPAs...

[zevans]

We now have friends in high places...
http://xkcd.com/619/

[instant]

Hi Guys!

Great thread, nice and easily understandable.

I have a little problem with my laptop.

I recently installed Ubunut 9.04 so it is completely fresh.
I tried following the guide but for some reason this lowers my performance, ie in PPracer from 25-37fps down to 10-25fps

I reverted the changes but let the fixmtrr.sh still run when X starts, still no change, but after doing a chmod -x on the fixmtrr.sh everything seems fine and PPracer is back to 37fps.

With other words my issue is with the fixmtrr.sh which seems strange,
does anyone have an explanation to this since I would like to have even better performance than I already have.

I've got a:

instant@laptop1:~$ lspci -v | grep VGA
00:02.0 VGA compatible controller: Intel Corporation Mobile GM965/GL960 Integrated Graphics Controller (rev 03)

Thx,
instant.

[psyke83]

I reverted the changes but let the fixmtrr.sh still run when X starts, still no change, but after doing a chmod -x on the fixmtrr.sh everything seems fine and PPracer is back to 37fps.

Can you paste the contents of /proc/mtrr before and after running the fixmtrr.sh script?

[burkas]

Anyone had try xorg intel driver (2.8) in Jaunty ? I had try it and desktop didn't show off. But i was installed it completely. Anyone know what's my fault ?

[instant]

@psyke83


Before running fixmtrr.sh

reg00: base=0x0c0000000 ( 3072MB), size= 1024MB, count=1: uncachable
reg01: base=0x000000000 (    0MB), size= 4096MB, count=1: write-back
reg02: base=0x100000000 ( 4096MB), size= 1024MB, count=1: write-back
reg03: base=0x0bf700000 ( 3063MB), size=    1MB, count=1: uncachable
reg04: base=0x0bf800000 ( 3064MB), size=    8MB, count=1: uncachable

After manually running fixmtrr.sh

Extracing base address and memory size from lspci -v
d0000000
10000000

Supplying corrected MTRR ranges to /proc/mtrr
reg00: base=0x0c0000000 ( 3072MB), size= 1024MB, count=1: uncachable
reg01: base=0x000000000 (    0MB), size= 4096MB, count=1: write-back
reg02: base=0x100000000 ( 4096MB), size= 1024MB, count=1: write-back
reg03: base=0x0bf700000 ( 3063MB), size=    1MB, count=1: uncachable
reg04: base=0x0bf800000 ( 3064MB), size=    8MB, count=1: uncachable

I dont really see any difference between the two but just tried PPracer and I once again have lower FPS.

Thankful for any help,
instant

[ilna]

Can you paste the contents of /proc/mtrr before and after running the fixmtrr.sh script?

I also have the problem - up to date Karmic. After rebooting and being in KDE (I have added BEFORE/AFTER echos):

$ sudo ./fixmtrr.sh
cat /proc/mtrr BEFORE:             
reg00: base=0x080000000 ( 2048MB), size= 2048MB, count=1: uncachable
reg01: base=0x000000000 (    0MB), size= 4096MB, count=1: write-back
reg02: base=0x100000000 ( 4096MB), size= 1024MB, count=1: write-back
reg03: base=0x140000000 ( 5120MB), size=  512MB, count=1: write-back
reg04: base=0x160000000 ( 5632MB), size=  256MB, count=1: write-back
reg05: base=0x170000000 ( 5888MB), size=  128MB, count=1: write-back
reg06: base=0x178000000 ( 6016MB), size=   64MB, count=1: write-back
reg07: base=0x0af800000 ( 2808MB), size=    8MB, count=1: uncachable
Extracing base address and memory size from lspci -v
d0000000
10000000

Supplying corrected MTRR ranges to /proc/mtrr
cat /proc/mtrr AFTER:
reg00: base=0x080000000 ( 2048MB), size= 2048MB, count=1: uncachable
reg01: base=0x000000000 (    0MB), size= 4096MB, count=1: write-back
reg02: base=0x100000000 ( 4096MB), size= 1024MB, count=1: write-back
reg03: base=0x140000000 ( 5120MB), size=  512MB, count=1: write-back
reg04: base=0x160000000 ( 5632MB), size=  256MB, count=1: write-back
reg05: base=0x170000000 ( 5888MB), size=  128MB, count=1: write-back
reg06: base=0x178000000 ( 6016MB), size=   64MB, count=1: write-back
reg07: base=0x0af800000 ( 2808MB), size=    8MB, count=1: uncachable

Also /var/log/messages has this fragment:

Aug 12 19:35:06 localhost kernel: [    1.540207] agpgart-intel 0000:00:00.0: Intel 965G Chipset
Aug 12 19:35:06 localhost kernel: [    1.541047] agpgart-intel 0000:00:00.0: detected 7676K stolen memory
Aug 12 19:35:06 localhost kernel: [    1.543783] agpgart-intel 0000:00:00.0: AGP aperture is 256M @ 0xd0000000
Aug 12 19:35:06 localhost kernel: [    1.558561] [drm] Initialized drm 1.1.0 20060810
Aug 12 19:35:06 localhost kernel: [    1.566119] i915 0000:00:02.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16
Aug 12 19:35:06 localhost kernel: [    1.569486] mtrr: type mismatch for d0000000,10000000 old: write-back new: write-combining
Aug 12 19:35:06 localhost kernel: [    1.569489] [drm] MTRR allocation failed.  Graphics performance may suffer.
Aug 12 19:35:06 localhost kernel: [    1.689866] [drm] fb0: inteldrmfb frame buffer device
Aug 12 19:35:06 localhost kernel: [    1.689946] [drm] Initialized i915 1.6.0 20080730 for 0000:00:02.0 on minor 0

[GNUbee40]

Hello there!

Tried the Safe mode with good success. HD Youtube is no longer an issue. However, now I get logged out every time I send my machine to sleep and thus loose all my opened windows. Not sure if this is related to the X-upgrades, but I suppose it is some kind of problem with X.
Any comments or suggestions, what could have changed?

Regards

[PatrickVogeli]

yeah, I had that problem too. It was related to UXA, using EXA it didn't happen, if I recall correctly.

What I ended doing? upgrading to the optimal bleeding edge. Once I found a completely working update, I disabled the extra ppas, so I have everything I need working, nice speed and on breakage.

[instant]

I've been doing some reading about this whole MTRR thing here and I believe that my only issue is that I cannot change /proc/mtrr.

I checked the fixmtrr.sh script and tried the commands manually, but all i get is an error message about invalid argument.
I have seen this exact issue here, but cant seem to replicate the workaround mentioned in the report.

The workaround is exactly what the fixmtrr.sh script does, simply:

echo "base=0xd0000000 size=0x10000000 type=write-combining" >| /proc/mtrr

Anyone?

/instant