Talk:Intel graphics

From ArchWiki
Revision as of 17:41, 29 July 2014 by Indigo (talk | contribs) (Kernel Module options: re)
Jump to: navigation, search

Setting gamma and brightness

Intel offers no way to adjust these at the driver level. Luckily these can be set with xgamma and xrandr.

Adding acpi_backlight=vendor to the kernel boot line allows the backlight to be changed on the Intel HD 3000 via xbacklight without root access. Perhaps a note could be added to provide this option. papabean (talk) 02:40, 14 January 2013 (UTC)

acpi_backlight=vendor stopped working a few kernels ago. (No. I did not revert to be certain).

However, intel-gpu-tools in the community repository includes the intel_backlight utility which allows one to manually set the backlight level.

papabean (talk) 04:36, 11 February 2013 (UTC)

Best performance options

In my experience (i5 CPU w/ HD4000, Z77 mobo) the best performance can be had with the following options:

  • SNA acceleration method
  • Vertical sync enabled
  • TearFree option enabled
  • i915.semaphores=1 kernel option turned on

Using the ck kernel specific to the CPU architecture also seems to make the video performance more consistent, but not necessarily any faster.

Disable VSYNC

I suggest that the restoration of this entry, it was very useful

To disable VSYNC, adjust /etc/X11/xorg.conf.d/20-intel.conf in Section "Device" the string Option "SwapbuffersWait" "false".

Alternatively, set vblank_mode to 0 in ~/.drirc and make sure that driver is set to dri2:

<device screen="0" driver="dri2">
   <application name="Default">
   <option name="vblank_mode" value="0"/>

Weathered Colors

Is there a way to do the equivalent of xrandr --output HDMI3 --set "Broadcast RGB" "Full" without having X running? I have two identical monitors (Asus VH222H), one connected to my dvi port (HDMI1), the other to the hdmi port (HDMI3). Both support full range rgb, but the one connected to hdmi always boots set to limited range. Its easily fixed with the xrandr set command, but I don't like running that all the time and it also doesn't work if I go straight to a vt after booting. It would be nice to put it in a systemd job.

Perhaps I should file a bug report with the kernel? The auto detect is more sophisticated than just HDMI == limited range tv, right?

Corrupted fonts after resume from ram

I've finally found searching across many forums the solution for my i915 graphic card.

My card is: "Intel Corporation Mobile 915GM/GMS/910GML Express Graphics Controller (rev 04)".

The problem is: After resuming from ram, some letter have a missing orizzontal line, for example I see ¦ instead of |. Another thing is that sometime, after resuming, the cpu goes to 100% of usage with no apparent reason.

The solution is: Create the file /etc/modprobe.d/local.conf (if it doesn't already exist) and add to it the line "options drm_kms_helper poll=N".

I cannot understand this mystery but now I can use "Suspend to ram" as many times as I want.

Confusing TIP - Disable Vertical Synchronization (VSYNC)

This section is in Tips and Tricks and the way it is written, at first I thought this is recommended setting.

From 1 month I am using that setting, today I realised that the section was mentioned just for testing/benchmarking purpose.

May be there should be clear indication that, these settings are not required.

Amish (talk) 08:59, 11 June 2014 (UTC)

Kernel Module options

I was chatting with some Intel devs on Freenode (#intel-gfx) this evening and came across this:

  • <gwmngilfen> yeah, i have /sys/module/i915/parameters/enable_rc6=7
  • <gwmngilfen> let me change that
  • <bwidawsk> don't do 7
  • <bwidawsk> ever
  • <bwidawsk> gwmngilfen: I will try to find you a real link, but bit 2 there, rc6++ is known to not work

Apparently, we should be setting rc6=0 unless it has been shown to work for a given chip - in general, it's broken. I can update the main page if no-one objects?

There should be "a real link" first... -- Lahwaacz (talk) 06:27, 16 July 2014 (UTC)
I agree, in general, but bwidawsk said he couldn't find one in the public Intel docs, it was an internal thing. Assuming this is the same bwidaswk as, then this is coming from an Intel employee and thus I'm inclined to believe him. Here's the rest of the IRC transcript from that chat:
  • <imirkin> bwidawsk: is the official stance that it's broken or have you guys just not been able to get it to work?
  • <bwidawsk> rc6++ is broken, it has never worked
  • <bwidawsk> if it works for you, you may use it
  • <bwidawsk> but i'm pretty sure with any modern GL it will not
  • <bwidawsk> perhaps pre HW context GL can work
  • <bwidawsk> yeah, i think we publish nothing about those registers
- Gwmngilfen 10:26, Jul 22 2014
I've added an Accuracy flag that points to this discussion, let's wait for more opinions. -- Kynikos (talk) 04:32, 27 July 2014 (UTC)
I believe what is referenced by him as "rc6++" above is the so-called "deepest" powersave mode you can set along with all others (with rc6=7). He does not imply rc6 is broken and has to be disabled completely (with rc6=0). Clearly it is chip dependant and the more modes/options you try to enable manually, the more trouble might arise. Perhaps the section (after checking what is enabled automatically; the accuracy flag) might be changed to advice turning on the basic (safest) mode first (just rc6=1) and experiment setting it higher with the other rc6 modes/module options. (Most Arch users probably try maxing it out with "rc6=7 et al" anyway:) --Indigo (talk) 17:41, 29 July 2014 (UTC)