Updated and Optimized Ubuntu Free Graphics Drivers

[d2kx]

Did some tests today with my AMD Radeon R7 260X (GCN 1.1, BONAIRE).

Counter-Strike: Global Offensive is 25-30% faster with stock Ubuntu 14.10 (Mesa 10.3, but HyperZ enabled manually) than it is with Ubuntu 14.10 + Oibaf PPA (Mesa 10.4-git, also with HyperZ enabled), so it regressed a little.

Independent of the Mesa version used, using Linux 3.18rc1 gives a small 8-10% performance boost over the stock Linux 3.16 kernel used in Ubuntu 14.10 (haven't tried Linux 3.17).

So Mesa 10.3 + Linux 3.18rc1 is the fastest combination for me with the opensource stack, but AMD Catalyst is still quite a bit faster.

[stfn]

Recent builds seem to break something with Intel drivers. I am not able at the moment to provide any info, since I am quite busy, and had to ppa-purge right away. Is there anyone who can confirm issues with intel cards, or someone who could provide some additional data about this?

I can confirm some issues with Intel hardware. A lot of flashing and blinking occurs in the terminal window, have not noticed issues with other applications.

I'm on a fresh 14.10 install. Terminal was working properly before I added the Oibaf ppa.

Intel i3 530 CPU: http://ark.intel.com/products/46472

[ 19.240] (II) intel(0): Using Kernel Mode Setting driver: i915, version 1.6.0 20080730
[ 19.240] (II) intel(0): SNA compiled: xserver-xorg-video-intel 2:2.99.916+git1410241932.d08a5f~gd~u (Oibaf <[email protected]>)

[oibaf]

Arm support would be nice, but I am a bit busy currently and it would require me some time updating the packages for it.

Nouveau issues are probably due to recent dri3 commits, if it's not fixed, file a bug, you should have read how to do in the ppa description.

Intel issues: file a bug for that also.

About radeon regressions I have no idea, you should try to see if it happens also when compiling from source and bisect to find what's wrong.

[gutigen]

Arm support would be nice, but I am a bit busy currently and it would require me some time updating the packages for it.

Nouveau issues are probably due to recent dri3 commits, if it's not fixed, file a bug, you should have read how to do in the ppa description.

Intel issues: file a bug for that also.

About radeon regressions I have no idea, you should try to see if it happens also when compiling from source and bisect to find what's wrong.

Is dri3 enabled for radeonsi in your PPA?

[sarnex]

Is dri3 enabled for radeonsi in your PPA?

No, I don't think so because DRI3 is not enabled by default for radeonsi, because Present is not yet implemented. If you use a small patch for xserver-xorg-video-ati, you can get it running, but there will be no present(eg no vsync).

The patch is very simple assuming you have xserver 1.16, see https://github.com/iXit/xf86-video-a...f0c7f8d422b7d5

If you have 1.15, contact me on irc and I'll send you a patch mannerov wrote for me.

[atcrank]

vdpau still pointing to nvidia.so

Hi everyone - many thanks to Oibaf for this ppa. I think this is going to be pretty easily resolved, but i am having a lot of trouble finding the culprit.

I'm on Ubuntu 14.04 now, but this is a machine that has been upgraded since 12.04, Back then, I had a GeForce 210 in it. Most of the time I just use it to watch MythTV. I upgraded to a 7790 (Bonaire, radeonsi) and want both some 3D and VDPAU. I tried out the PPA a while ago, but I backed out and went to Catalyst for a year. I've re-added the PPA, upgraded and installed the 3.17 kernel.

vdpauinfo:

display: :0   screen: 0
Failed to open VDPAU backend libvdpau_nvidia.so: cannot open shared object file: No such file or directory

glxinfo:

name of display: :0
Error: couldn't find RGB GLX visual or fbconfig

I have the impression that I've done what I need to if I was working from a blank slate. That makes me think I just need to delete some links or config file lines to get it working. Sorry this is rather basic.

dmesg doesn't report any errors.

[   13.414225] [drm] Initialized drm 1.1.0 20060810
[   13.495565] [drm] radeon kernel modesetting enabled.
[   13.495677] fb: switching to radeondrmfb from VESA VGA
[   13.501783] [drm] initializing kernel modesetting (BONAIRE 0x1002:0x665C 0x1043:0x0452).
[   13.501810] [drm] register mmio base: 0xFE200000
[   13.501811] [drm] register mmio size: 262144
[   13.501816] [drm] doorbell mmio base: 0xD0000000
[   13.501817] [drm] doorbell mmio size: 8388608
[   13.501964] [drm] Detected VRAM RAM=1024M, BAR=256M
[   13.501966] [drm] RAM width 128bits DDR
[   13.507332] [drm] radeon: 1024M of VRAM memory ready
[   13.507334] [drm] radeon: 1024M of GTT memory ready.
[   13.507355] [drm] Loading bonaire Microcode
[   13.524262] [drm] Internal thermal controller with fan control
[   13.524423] [drm] probing gen 2 caps for device 1002:5a16 = 31cd02/0
[   13.531238] [drm] radeon: dpm initialized
[   13.536218] [drm] Found VCE firmware/feedback version 40.2.2 / 15!
[   13.536244] [drm] GART: num cpu pages 262144, num gpu pages 262144
[   13.538933] [drm] probing gen 2 caps for device 1002:5a16 = 31cd02/0
[   13.538939] [drm] PCIE gen 2 link speeds already enabled
[   13.553723] [drm] PCIE GART of 1024M enabled (table at 0x000000000078B000).
[   13.554402] [drm] Supports vblank timestamp caching Rev 2 (21.10.2013).
[   13.554404] [drm] Driver supports precise vblank timestamp query.
[   13.554479] [drm] radeon: irq initialized.
[   13.558733] [drm] ring test on 0 succeeded in 3 usecs
[   13.558829] [drm] ring test on 1 succeeded in 3 usecs
[   13.558850] [drm] ring test on 2 succeeded in 3 usecs
[   13.558944] [drm] ring test on 3 succeeded in 2 usecs
[   13.558953] [drm] ring test on 4 succeeded in 2 usecs
[   13.614613] [drm] ring test on 5 succeeded in 2 usecs
[   13.635627] [drm] UVD initialized successfully.
[   13.745176] [drm] ring test on 6 succeeded in 19 usecs
[   13.745191] [drm] ring test on 7 succeeded in 4 usecs
[   13.745192] [drm] VCE initialized successfully.
[   13.747215] [drm] ib test on ring 0 succeeded in 0 usecs
[   13.747360] [drm] ib test on ring 1 succeeded in 0 usecs
[   13.747503] [drm] ib test on ring 2 succeeded in 0 usecs
[   13.747649] [drm] ib test on ring 3 succeeded in 0 usecs
[   13.747793] [drm] ib test on ring 4 succeeded in 0 usecs
[   13.769449] [drm] ib test on ring 5 succeeded
[   13.790864] [drm] ib test on ring 6 succeeded
[   13.792374] [drm] ib test on ring 7 succeeded
[   13.795958] [drm] Radeon Display Connectors
[   13.795962] [drm] Connector 0:
[   13.795963] [drm]   DP-1
[   13.795964] [drm]   HPD2
[   13.795978] [drm]   DDC: 0x6530 0x6530 0x6534 0x6534 0x6538 0x6538 0x653c 0x653c
[   13.795979] [drm]   Encoders:
[   13.795981] [drm]     DFP1: INTERNAL_UNIPHY2
[   13.795981] [drm] Connector 1:
[   13.795982] [drm]   HDMI-A-1
[   13.795983] [drm]   HPD3
[   13.795985] [drm]   DDC: 0x6550 0x6550 0x6554 0x6554 0x6558 0x6558 0x655c 0x655c
[   13.795986] [drm]   Encoders:
[   13.795987] [drm]     DFP2: INTERNAL_UNIPHY2
[   13.795987] [drm] Connector 2:
[   13.795988] [drm]   DVI-D-1
[   13.795989] [drm]   HPD1
[   13.795990] [drm]   DDC: 0x6560 0x6560 0x6564 0x6564 0x6568 0x6568 0x656c 0x656c
[   13.795991] [drm]   Encoders:
[   13.795992] [drm]     DFP3: INTERNAL_UNIPHY1
[   13.795993] [drm] Connector 3:
[   13.795994] [drm]   DVI-I-1
[   13.795995] [drm]   HPD6
[   13.795996] [drm]   DDC: 0x6580 0x6580 0x6584 0x6584 0x6588 0x6588 0x658c 0x658c
[   13.795997] [drm]   Encoders:
[   13.795998] [drm]     DFP4: INTERNAL_UNIPHY
[   13.795999] [drm]     CRT1: INTERNAL_KLDSCP_DAC1
[   13.848953] [drm] fb mappable at 0xC098E000
[   13.848954] [drm] vram apper at 0xC0000000
[   13.848955] [drm] size 8294400
[   13.848956] [drm] fb depth is 24
[   13.848956] [drm]    pitch is 7680
[   13.849197] fbcon: radeondrmfb (fb0) is primary device
[   13.884977] radeon 0000:01:00.0: fb0: radeondrmfb frame buffer device
[   13.920771] [drm] Initialized radeon 2.40.0 20080528 for 0000:01:00.0 on minor 0

[asdfblah]

Hi everyone - many thanks to Oibaf for this ppa. I think this is going to be pretty easily resolved, but i am having a lot of trouble finding the culprit.

Do you have the "mesa-vdpau-drivers" package installed? If so, try setting "VDPAU_DRIVER" to "radeonsi":

export VDPAU_DRIVER=radeonsi

then run vdpauinfo.
Also, not sure if this helps, but make sure you have "libgl1-mesa-glx" and "libgl1-mesa-dri"...
If this doesn't help, I suggest to completely reinstalling xorg + mesa.
Not that I know much about how this all works, but that's how I solved problems in the past. It's either missing packages or leaving something is badly configured.

[atcrank]

Error 23

Thanks mesa-vdpau-drivers is installed. Directing VDPAU_DRIVERS changed things, so some progress, but:
vdpauinfo:

display: :0   screen: 0
Error creating VDPAU device: 23

I've had a look around and there are a few threads where this is coming up. Most suggest trying 'export DISPLAY:=0' (I tried this and got an error; but 'export DISPLAY=0' worked or didn't complain.)

After that:

>vdpauinfo
vdpauinfo: cannot connect to X server 0
>glxinfo
Error: unable to open display 0

[triune]

I'm not sure if this has been asked before, going through ~140 pages of replies to this thread would take ages! So...

Is there any way to get STABLE updates of Mesa in PPA form?

I've been using the OIBAF for two weeks now and have to say that I don't want to be on the bleeding edge of Mesa development and would much rather prefer to get stable updates of Mesa. Given that 10.3.x is now the current stable branch, could OIBAF maintain a separate repository for that latest stable Mesa releases? I'm sure I'm not the only 14.04 user out there who would love to upgrade from Mesa 10.1.3 but retain the stability of an LTS release.

I've noticed several system crashes (complete graphics lockup) while on OIBAF not to mention that packages need to be upgraded several times a week which all require a reboot. This is probably great for driver developers, but I'm not sure who else out there needs these sort of updates? Wouldn't the wider/larger audience best be served by stable updates instead of git trunk builds?

[axfelix]

I suppose that wouldn't hurt, but I think the reason it hasn't been done up to now is twofold:

- Ubuntu sources typically haven't been so far behind (and Mesa is usually pretty conservative about stable releases) that a new stable release typically makes much difference versus the default

- "Unstable" Mesa from git has always been 100% stable for me (on Sandybridge), so I'm surprised you're encountering so many issues! My system is configured to automatically fetch updates from apt-get once a week, and I'll usually go several weeks between reboots without any harm, so while I don't actually see these updates on the near-daily basis that the PPA provides, it still suits me just fine.