Gutsy: Intel should be preferred over 810

[trollord]

Binary package hint: xserver-xorg

I just removed after upgrade to Gutsy /etc/X11/xorg.conf and did
sudo dpkg-reconfigure -phigh xserver-xorg
to get a new default configuration.

It came up with i810 driver for me. This should *never* happen by default in any circumstances! It is absolutely intolerable if Gutsy will work out of the box like this. The default Gutsy installation must prefer intel over i810 in case both seem to support the hardware.

Reasons:

1. i810 is absolutely hopeless with widescreen modes. The only hope for getting those working is 915resolution, which has got its own bugs and is not installed by default. This means that the live CD environment and default desktop looks blurry and horrible no matter what. (I would like to point out that Intel has got globally ~60% market share of display adapters. Where I live 95% of sold displays are TFTs and of which some 50% are widescreen. We are not talking of a marginal group of people affected by this issue.) The intel driver however works wonderfully out of the box with widescreen modes.

2. i810 is notoriously bad with XV, causing the dreaded BadAllocs which crash the multimedia applications. You have to know to add stuff like
Option "VideoRam" "65536"
Option "CacheLines" "1980"
Option "LinearAlloc" "16000"
to xorg.conf or you are affected. Which then again can break compiz-fusion (the last line in fact does, but without that you have ca. 1/100 chance of video actually trashing XV until you restart when using gma950). These are old bugs that no one will ever fix because the attention moved to the new driver couple years ago already. The intel driver has got significantly better working xv which seems to be just working, and does not cause side effects.

3. Xrandr, enabling vga outputs and couple more advanced tricks are far more likely to work with intel driver than i810. Especially laptop users and people that own pivot panels are affected by this...

Please, do whatever to fix it. It reduces so much pain from ordinary users that you have no idea. Making sure intel gets used when appropriate should be considered imho as release CRITICAL issue.

[Bryce Harrington (bryce)]

Thank you for your bug report and for helping to make Ubuntu better.

You are correct that -i810 should be deprecated in favor of -intel. On my hardware (a Dell 1505n), when running sudo dpkg-reconfigure -phigh xserver-xorg, it correctly selects -intel. There is some hardware that we are not certain -intel will work properly.

Have you verified that your system works properly when you manually change "i810" to "intel"? If so, please provide the output of 'lspci -vvnn' so we can update discover-data accordingly.

[unggnu (unggnu)]

I can confirm this. Since Gutsy the Intel driver works fine on my Sony TX2 with 915 chipset so there is no problem to activate it per default.

[unggnu (unggnu)]

A driver update for this chipset/laptop would fix this bug report https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-i810/+bug/95867 too.

[trollord (trollenlord)]

In fact getting rid of i810 will "fix" all 915resolution bugs (16 of them + duplicates) as the package becomes entirely obsolete. There will be no need for it anymore. You can find also some ~50 "I got wrong resolution out of the box" bugs + duplicates that are assigned to random locations. Most of those caused by not supporting widescreen modes will be solved pretty much instantly.

I selected "intel" instead of "i810" and removed 915resolution. I got automatically the correct mode, and it works nicely. Why I got i810 when using dpkg-reconfigure, that I don't know. Xrandr and vgaout/tvout on-the-fly seem to be working (without restarting X or having a long and brutal fight with i810switch) which wasn't the case with 810.

I got still some BadAlloc issues with XV, but on that i810 and intel are identical. It is some separate issue that pops up only when the compiz-fusion is enabled and has plagued these cards for years. It's some sort of an issue with having too little amount of memory (after especially compiz-fusion has hogged some) for XV frame buffer. Or something like that. It should be handled separately. (See these: https://bugs.launchpad.net/Ubuntu/+bugs?field.searchtext=badalloc )

Anyways, in overall the experience when using intel over i810 is a lot better for me and I believe that it is for nearly everyone. Are there actually cards that are supported by i810 but NOT by intel? That would imho be the only reason for having the ability to use i810 included on default installations and on livecd.

[unggnu (unggnu)]

Ok, I have found one problem with the intel driver. According to this bug https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-ati/+bug/122979 the compiz xv problem has been fixed in i810 and intel driver but video playback doesn't work with new intel driver on some cards like i830, i915 but it works with i810 driver.

[trollord (trollenlord)]

Yes, i810 works better with xv+compiz, intel and xv+compiz seems to be failing. I will live without compiz then, because intel works nicely without it :-)

[Aaron Whitehouse (aaron-whitehouse)]

I have been fighting for years to fix the non-detection of my resolution. Every release I have tried to make 915resolution issues automatically incorporated into the install. Intel fixes these issues. Still, I'm not sure that I'm comfortable with moving the default to Intel for my card (855GM, 1400x1050).

Point 1 is absolutely correct and is the main reason that I hope Intel will make the prime time as soon as possible. It does, however, have the easy workaround of 915resolution. That workaround is only found by the computer savvy and most average users suffer with incorrect resolutions.

Point 2 I can't agree with. I have found intel to be more buggy than i810 (which is understandable given that it is "experimental") and the "fixes" in the bug reports for Intel are usually very similar to the ones that you complain about.

Overall, I have found Intel to be slower and more buggy. Far more so than combining i810 with 955resolution. I am running Intel to try and help find these bugs, not because I prefer it.

The compiz+intel problem has been a little glossed over here. Given that the intel cards can often do Compiz, it will be running by default on all the machines that we are talking about. That means that an important feature of the system is broken in the default install. Hopefully this is fixed by the final release.

I am sure that Intel will replace i810. My personal experience is that it isn't yet as stable as i810.

[Travis Reitter (treitter-dev)]

My laptop should also default to the "intel" X.org driver, too, I believe. Or at least 915resolution should be installed by default. Otherwise, I'm forced to 1024x768 on a native 1280x800 display (note that it's also a distorted aspect ratio, which is even more annoying). I was using the incorrect resolution the last few weeks, since I didn't realize either solution - X was just refusing my resolution configuration, so I thought I was out of luck.

As far as the Compiz/xv video problem, I have that with both the i810 and the intel driver, so that isn't a regression in my case (but a significant bug nonetheless; I'll file in the appropriate other bug(s)).

[Aaron Whitehouse (aaron-whitehouse)]

Does anyone else find that their machine runs a lot slower using intel
than using i810?

[Vincenzo Ciancia (vincenzo-ml)]

I have reported bug #137225 and bug #137234. Do some of you experience the latter, or does the second display work for you all? I am still unable to see videos using XV, but I must admit the driver is fast enough to watch them using "x11shm" method even with compiz effects on them. I favour the decision to make intel the default, but second head must absolutely work out of the box, as it did in dapper and edgy!

[trollord (trollenlord)]

Dapper and edgy both failed out of the box miserably with widescreen
modes...

[Bryce Harrington (bryce)]

Btw, I've sent in a change to switch all hw currently defaulting to -i810, to -intel. KyleM will be committing the change, however he's on travel currently so I don't know exactly if/when this will hit.

Note that we're aware of a ton of people for whom -i810 still works better than -intel on their hardware. But clearly -intel is the way forward, and we feel -i810 should just be used as a legacy option to switch to if -intel has not worked.

If after this change, people start having the reverse problem (Ubuntu install selects -intel and breaks, and you find -i810 works better), please file a new bug and include details of the failure plus the output from 'lspci -vvnn'. I can use that to individually blacklist your specific hardware to default to i810.

[trollord (trollenlord)]

I wonder if someone could give the xv some love, that would reduce the
amounts of those blacklisted ones..

[Bryce Harrington (bryce)]

discover-data has been uploaded with the s/-i810/-intel/ change.
See my previous note for what to do in reporting regressions caused in this change.

[Aaron Whitehouse (aaron-whitehouse)]

I suppose that it makes sense to go that way - hopefully we can iron
out all the bugs if more people are using it by default.

Thanks for all your hard work on this, Bryce.

[Alessandro Gervaso (gervystar)]

I report that I'm having issues like those mentioned in several bugs on two different pcs using intel graphics (and the intel driver):
xv output (totem, mplayer) doesn't work if desktop effects are active.
The strange thing is that it used to work until about a month ago.

References: #38389 #49360

[Alessandro Gervaso (gervystar)]

Sorry, I forgot to mention that I'm running a daily updated gutsy on both machines.

[unggnu (unggnu)]

Ok, the decision was and is correct to move to new Intel driver. The xv issue https://bugs.launchpad.net/ubuntu/+source/compiz/+bug/111257 has been fixed at least with the patch and in some days in rep I guess.

[Franck (alci)]

Well, saying the xv problem is solved is quite optimistic...
The workaround the said patch implements is _not to use_ xv but gl instead. So it kind of solves the problem for totem when using gstreamer, but in no way does it solve the problem with intel + xv. Just try any other video player, and you'll see it (say vlc for example) !

Or am I missing something ?

[unggnu (unggnu)]

At least for me XV works fine even with VLC or other playersl

[trollord (trollenlord)]

As far as I can tell you are not missing anything. There are still problems
using intel+xv+compiz simultaneously, at least on gma950. (badalloc+instant
crash for the media players)

[unggnu (unggnu)]

There is no updated drive package until now you have to compile a new Intel driver on your own with the debdiff posted here https://bugs.launchpad.net/ubuntu/+source/compiz/+bug/111257/comments/80 or wait some days as I posted above.

Imho the problem atm is this bug https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/127101 but it seems that not everyone is affected.

[legolas558 (legolas558)]

the intel driver is much more BUGGY than the i810 driver. I have lost 5 days trying to fix this bug: https://bugs.freedesktop.org/show_bug.cgi?id=12666

The solution was to use the xf86-video-i810 v1.7.4 driver, which does not use xf86-video-intel code

[Fredrik Wendt (fredrik-wendt)]

I have a laptop where the i810 driver properly detects my wide screen monitor and chooses 1200x800. Gutsy final beta with intel driver used 1024x768 instead. After trying to set up a second monitor (for a projector) X wouldn't start up and I had to remove xorg.conf and start over to get things working again.
The XV issue as described in the bug report really is a pita with i810, but that's possible to workaround. I'd prefer telling my grandfather that he'll have to wait for compiz/desktop effects to mature another 6 months than to get him to log in using the console, remove a file having him type in the name, running dpkg-reconfigure bla bla ...

I say stick with what is stable. That's why I chose and recommend Ubuntu - "things just work". Want the fancy stuff - go with some other distro.

[Fredrik Wendt (fredrik-wendt)]

Oh, forgot to say it's a ThinkPad Z61t (9443-4GG) with a Intel Corporation Mobile 945GM/GMS/940GML graphics card (lspci -n says 8086:27a2).

[Aaron Whitehouse (aaron-whitehouse)]

Did you file that bug Fredrik? if you can isolate the issue, a bug
report would be great. If it works in i810 and not in intel, it would
be worth mentioning that as well.

I have serious issues with the intel driver as well. I have filed Bug
#150109 and Bug #150114 , but what concerns me more is all of the
general instability that I can't replicate or isolate enough to file
useful bug reports. I would love to see those issues fixed, but it
doesn't look like we are likely to prior to release.

I don't know what the best answer is...

[unggnu (unggnu)]

Looks similar to bug # 138094 if the mouse is still moveable. But it happens for me with i810 too afaik.
Do you have played any Video while or before suspending?

[unggnu (unggnu)]

Sorry, last post was dedicated to bug # 150109.