The Official Linux g45 chipset MoBo Thread - Page 4

By the way, I saw that some people copied or linked their self-compiled drivers to /usr/lib from /usr/local/lib. That's unnecessary since one can tell X where the drivers can be found. Like this:

Phelin, that is indeed a great tip! Alternatively, when you compile the driver itself, you can set where it gets installed to...So for example on my ubuntu box, it wants them in /usr/lib instead of /usr/local/lib...I had no idea you could just call out the library dir so easily.
-Trouble

Don't bother with the latest drivers in the git tree.

- intel driver requires a version of drm that's only in the git tree, so you must compile drm;
- the drivers crash X. (low-rez mode window) Irrecoverable.

I don't understand how these made it into public view.

Looks like I got a fair way along the path without the pain the rest of you have gone through. I have the GigaByte GA-EG45M-DS2H, connected via a HDMI 1.3 cable to a 1920x1080p capable Samsung LCD TV. installed Debian Lenny, 64bit (ie amd64 arch), 2.6.26. As far as I can tell everything non A/V related just works with it.

I didn't expect the Lenny xserver-xorg-video-intel to work as it was 2.3.2, and I had read 2.4.0 kills the kernel. But I saw that 2.4.2 was the latest released version from xorg. I did toy with the idea of using git version, but then I noticed 2.4.2 had hit Debian experimental, so I used that. I just had to hack the build-depends a bit and it compiled without a hitch. And then just worked. 1920x1080 on the TV. Yipee! And this was with a default xorg.conf - it auto detects everything. Amazing.

OK, it doesn't "just work". The screen flickers a occasionally, even on static images. Looks like its loosing vertical sync. I am hoping a shorter HDMI cable will fix it. And something has decided I am blind and need huge font sizes - but the icons aren't scaled so they are tiny. But that won't be a driver problem, so I can probably hunt it down. And it later turned out The X server kills the machine on about every 10th restart. Still, its usable, and given the rate of change for the Intel driver I am confident that issue will go away.

So I have video, I presume (but haven't tested) that I have analogue audio. So I have a working system.

That leaves me with 3 issues. First is sound over the HDMI cable, which I understand requires some ALSA driver which doesn't exist, even in the unreleased 2.6.27. Is that correct? What should I see when it does work - a new sound device in /dev? Or will the existing Intel Audio automagically push sound through the HDMI?

Secondly, the G45 can do BluRay acceleration, on Windows on a good day anyway. I gather this is impossible under Linux now - the code doesn't exist. I am trying to make some sense of the thicket of acronyms that seem to be pertinent - XvMC and Gallium3D (1), and DRI2 (2). About the only thing I can make from all that is don't expect anything before X Server 1.6, which will ship sometime in 2009 - ie a year away. In the mean time, how feasible is playing BluRay content on a 2.4GHz 6600 CPU?

Thirdly, I have a BluRay drive, but playing BluRay disks looks to require of all things a closed source java application to strip the AACS, and it can't do streaming. Is there any other solution?

(1) www . phoronix . com/scan.php?page=news_item&px=NjU3Nw
(2) www . phoronix . com/scan.php?page=news_item&px=NjYzNw

Intel HDMI Audio is a fully compliant Azalia device, so the same ALSA drivers that work with other onboard codecs should work the same way to send LPCM over HDMI. However (and, again, it's been over a year since I read the source so I may be recalling incorrectly) the ALSA drivers enumerate all detected codecs on the bus and check them to determine their functionality. There's some code that checks the vendor ID and device ID, and if the Intel HDMI one is not present in the source I'm not sure how the ALSA driver reacts.

Certainly there was custom code to deal with a number of different sound solutions, and I suspect there will be some tweaks needed for Intel HDMI Audio as well. I can provide some help and guidance if there are developers out there who want to try to hack ALSA. Asking me to write code will result in buggy solutions and likely horrible noises-- not all of them coming from the sound card.


I think the VAAPI project is the best bet on this, but I don't know what their roadmap looks like.

im pretty sure a 2.4 Ghz dual core can handle bluray playback, with overclocking, im sure it could handle it easily.


when i was messing around trying to get my xbox hddvd drive to work on my ps3 running ubuntu, i found a program that (i think) could rip blurays and hddvds to hard drive, but ill be damned if i can remember the name. and i never got it to work. haha tons of help, right? theres probably tons on the debain forums, i know theres some on the ubuntu ones.

Okay, you inspired me to look at the code again. There's probably a lot more to this-- I'm not certain the generic Intel HDA code is set up to accomodate HDMI, but certainly if you wanted to make the attempt with the currently available stuff you would need to add to the alsa-kernel/sound/pci/hda/hda_intel.c source code:

old:
/* PCI IDs */
static struct pci_device_id azx_ids[] = {
/* ICH 6..10 */
{ PCI_DEVICE(0x8086, 0x2668), .driver_data = AZX_DRIVER_ICH },
{ PCI_DEVICE(0x8086, 0x27d8), .driver_data = AZX_DRIVER_ICH },
{ PCI_DEVICE(0x8086, 0x269a), .driver_data = AZX_DRIVER_ICH },
{ PCI_DEVICE(0x8086, 0x284b), .driver_data = AZX_DRIVER_ICH },
{ PCI_DEVICE(0x8086, 0x2911), .driver_data = AZX_DRIVER_ICH },
{ PCI_DEVICE(0x8086, 0x293e), .driver_data = AZX_DRIVER_ICH },
{ PCI_DEVICE(0x8086, 0x293f), .driver_data = AZX_DRIVER_ICH },
{ PCI_DEVICE(0x8086, 0x3a3e), .driver_data = AZX_DRIVER_ICH },
{ PCI_DEVICE(0x8086, 0x3a6e), .driver_data = AZX_DRIVER_ICH },

new
/* PCI IDs */
static struct pci_device_id azx_ids[] = {
/* ICH 6..10 */
{ PCI_DEVICE(0x8086, 0x2668), .driver_data = AZX_DRIVER_ICH },
{ PCI_DEVICE(0x8086, 0x27d8), .driver_data = AZX_DRIVER_ICH },
{ PCI_DEVICE(0x8086, 0x269a), .driver_data = AZX_DRIVER_ICH },
{ PCI_DEVICE(0x8086, 0x284b), .driver_data = AZX_DRIVER_ICH },
{ PCI_DEVICE(0x8086, 0x2911), .driver_data = AZX_DRIVER_ICH },
{ PCI_DEVICE(0x8086, 0x293e), .driver_data = AZX_DRIVER_ICH },
{ PCI_DEVICE(0x8086, 0x293f), .driver_data = AZX_DRIVER_ICH },
{ PCI_DEVICE(0x8086, 0x3a3e), .driver_data = AZX_DRIVER_ICH },
{ PCI_DEVICE(0x8086, 0x3a6e), .driver_data = AZX_DRIVER_ICH },
{ PCI_DEVICE(0x8086, 0x2802), .driver_data = AZX_DRIVER_ICH },
{ PCI_DEVICE(0x8086, 0x2803), .driver_data = AZX_DRIVER_ICH },

This will at least tell the ALSA driver to use the Intel driver when it encounters the HDMI audio codec. Beyond that... it might be some work. I'm not clear from my brief look at the source whether it will properly create a PCM device on Intel HDMI. I notice that for other HDMI audio devices there are specific patches for the hardware.

Ah, another acronym to add to my collection (1). I suspect it is the only bet, as it is the replacement for XvMC. Since it is being worked on by Intel, there is a good bet it will be support the G45 I guess.

VAAPI says it is not dependant on X, and indeed the way the pretty picture is drawn on its home page it looks like X doesn't talk to VAAPI at all, although VAAPI can use X to render its output. Instead the video player (eg Xine) uses VAAPI to do the decoding. I initially had hopes that means new version of Xine/MPlayer/Myth + VAAPI and I would get BluRay acceleration. But pictures lie. VAAPI talks to the kernel video interface - DRM, which in turn talks to the hardware. That would be the same interface X's DRI uses. I'd lay odds that VAAPI and DRI2 both require large changes to DRM. And I would also lay odds that those changes will be released together. Ie next year - when X.org 1.6 is released.


You are thinking of DumpHD (2), which is the Java program I was referring to. The author refuses to release the source, and I see some have resorted to editing the Java byte code directly to fix bugs. Some kind soul should put it through a Java decompiler, put the resulting crap on a repository somewhere so the world can reconstruct the source. If nothing else, that should prompt the author to do the right thing.

(1) VAAPI Home page: www. freedesktop.org/wiki/Software/vaapi
(2) DumpHD Home page: forum .doom9.org/showthread.php?t=123111

Does anyone have optical sp/dif out working on the dg45id? From what I have been able to determine, there is/was an ALSA bug that prevents this output from being detected/enabled at all. The hdmi output is detected/enabled, but that is not what I want to use.

I currently have my build to the point that I can start/use X reliably without crashes, however video output absolutely does not work (crashes mythfrontend) and sound does not work via the optical output.

I believe the best bet for anyone who wants Blu Ray on Linux would be to also have a Windows (virtual?) machine, with the AnyDVD HD product. AFAIK it is the only product which can handle both the AACS and BD+ schemes used on new Blu Ray products.

I don't think VA-API is going to help us anytime soon. The best bet IMO is retooling of XvMC to include additional acceleration methods, followed by whatever comes out of the Google Summer of Code project that is trying to implement generic gpu acceleration through Gallium3D. In the latter case, it sounds like specific hardware accelerations present in the various chips would be ignored, as the designed acceleration would use generic shaders and whatnot to achieve acceleration.

I've also seen that AMD/ATI may be working on releasing real usable video acceleration in their proprietary drivers soon, and there was an article on Phoronix yesterday about how Via is making an effort to suck less in this regard as well.

What does soon mean here? Does it mean "in the next few months", or "unlikely to be included in Xorg 1.6" (ie at least a year away), or "may not be included ever". If it isn't the first one, what causes you to think that?

**** that, windows ftl. i would rather die than spend money putting an operating system on an htpc. i highly doubt youd have enough power to run windows virtually while playing bluray thru it.


yeah, that guy really should release that source code, especially since its already so far along. either that, or get the **** off ubuntu so that someone else is motivated to do something about it. i dont get it when people put closed source code on an open source operating system.

you should get some sample videos off the web and trying playing them on that 2.4 of yours. elephants dream is a free 1080p movie you can dl, it was created entirely from open source software. ;P thats a pretty timid file though, so your best bet to see what it can handle is anything from here : http://www.hd-trailers.net/indexAll.php
also, linux developers have got to come up with a way to stream blurays. theres got to be a program (yes i know, something closed source) that would make the ****ing mpaa happy, that we could run on an open source operating system. either that, or someone should just hack it from scratch, and say screw the hdcp, but that sounds a lot harder for them.


i cant wait till this happens. i have a really good feeling about projects like this, its going to change everything. moving away from directx and opengl and GPUs in general is just about the best idea ive ever heard of in computing. i saw an article on slashdot, a guy from nvidia thinks the gpu is almost dead, i hope to god hes right.

Well I wasn't specific enough - the idea would be grab and decrypt a blu-ray image to a file, then play the file on the htpc.

I do not believe there is any capability for easily playing Blu ray on a PC right now, except using the AnyDVD product or having Windows and one of the commercial players. BD+ especially is not and may never be broken like CSS/AACS has been, such that you can just pop in a disc and play it without any additional work.

A few things, although I certainly shouldn't have any more information to work with than anyone else here. VA-API is a good idea that hasn't managed to unite all the major players behind it so far - AMD/ATI is doing their own thing, nVidia is doing whatever it is they do, I believe I read an article where Intel folks were talking about extending XvMC, Via has done the same, the GSoC project is trying to implement generic acceleration through Gallium3D, etc. The vaapi webpage has not been updated since June and only contains API specifications and a proof of concept library. Even if it were done and ready to go today, we'd need driver support for it, application support for it, etc. and there is none. I may just have missed a bunch of info on it, but what I have seen so far looks like a project still in or just out of the planning stages.

I don't think 1.6 is a year away though. I thought I read that the plan was to have it out by the end of the year, or sometime in Q1 09. It would contain most of the things that were supposed to be in 1.5 but were pushed out due to the GEM/TTM switchover.

If you read my previous post, you'll see that my LCD display is working.

I installed lirc with patch for imon display and lcdproc with imon_lcd patches and it came up without a snag. You can't turn off the backlighting, though, without shutting power to the LCD which the power supply doesn't do.

Tearing is much better using the overlay video patch to the Intel driver.

I tried out archibael's suggested patch to get HDMI audio and it didn't work. I guess I'll have to tear into the source to see how it actually should work.

BTW, thanks a lot to archibael for getting the technical info to us.

Noticed that tearing was still as bad and xvinfo and shows only textured video.

According to the 2.4.2 driver source, G45 has no overlay hardware.

hey guys,
anyone try suspend/hibernate on the asus p5q-em yet, or any G45 board?

I just set up the Gigabyte G45 board with Ubuntu Intrepid alpha. I actually installed alpha 5, but have done a dist-upgrade. Suspend works fine on it.

After install, the login screen showed up no problem, but in order to get to the desktop, I had to add Option "NoAccel" to the Device section of the xorg.conf file (thanks to the Ubuntu forums for that workaround). It looks good over HDMI at 1080p!

This is with the 2.6.27 kernel, xf86-video-intel 2.4.1, xorg 7.4, and xserver 1.5.0. With the video workaround I can't watch any videos fullscreen (I'm assuming it's related to the workaround).

Now I'm just trying to get sound working over HDMI. Currently alsa 1.0.17, although I tried 1.0.18rc3 with no luck. I think I'm going to be a bit less ambitious and try to get sound over optical instead.

good to know, do you know what that NoAccel option does?

I think it disables 2D acceleration.

Also, audio through s/pdif works with no effort at 2CH to my receiver. I may play with it later to try to get AC3 output.

im assuming analog audio out works just fine? (rca jacks)

Haven't tried, but I would assume so - that would be the easy thing for alsa to deal with.

I finally got back to trying this tonight. Seems as long as I am using XAA accelleration method, I can finally startup X into whatever window manager I want. mpeg2 decoding of ATSC signals is jumpy all the time, and freezing myth at times. E8400@3Ghz CPU is maxed out. I also seem to be losing access to the sound driver once in a while playing any HD stream, causing myth to black out. Restarting X solves that problem.

Also, I have to set the XVideo option to false in order to have myth not blackscreen immediately at startup due to "insufficient resources...".

Anyhow, still haven't gotten around to patching a seperate audio cable to see how well it works yet...I am not really too excited about this as I usually get out-of-sync video versus audio without alot of adjustments if I don't go with HDMI through my receiver.

thec, you must not be using kubuntu 8.10 alpha5 and the Intel drivers.

Try the HD clip here:
http://nextcomwireless.com/R5000/samples.htm

... I get ~23% CPU utilization on the E8400.

yeah, theres definitely something wrong if a cpu that big is maxed out for OTA or cable hdtv. if it was a 2Ghz single core, i would say thats about right.

Don't know where I went wrong, but I was definitely running alpha5 with the intel driver. Anyhow, I just dropped alpha6 into the system...package checks afterwards show that I am still running 2.4.1 of the intel driver.

To make a long story short, I can now play ATSC mpeg2 with just 23% of my CPU.

I also found out that instead of switching to XAA acceleration method, I can turn off DRI to workaround the X startup problem. I am not sure if something else has changed in Alpha6 to allow this, but the X server still "freezes" at startup without changing something. I haven't tried turning off DRI before and thought I'd just try it...turns out to be working well.

Now I can turn my attention back to getting audio through HDMI. Any one have progress on this end yet?

Has anyone noticed any differences in running 32-bit vs 64-bit linux with this chipset? Specifically with the intel video drivers.

Hi all,

New here and all, but stumbled across this thread a while back and read it with interest - finally decided to contribute.

I've currently got an Intel DG45FC running openSUSE 11.1 beta 1 (x86_64) with MythTV running "ok" hooked up to my home theatre. Video is over HDMI, Audio is optical SPDIF. HiDef DVB-T streams (720p) require a disproportionate amount of CPU (~40% on an E8500!) and I need to look into that.

Pretty much stock install (although that is a looong story) except ALSA drivers. The ALSA drivers that got installed with the 2.6.27RC5 kernel from the build service didn't seem to match with the ALSA 1.0.18RC3 packages. Grabbing the ALSA source and compiling and installing the drivers gave me many more "digital" options to play with and thus got audio over SPDIF working. Another things of note is that audio is not configured to use PulseAudio. I want to get audio over HDMI working, but to date have had no joy. I will try archibael's "patch" above when I get a chance.

I've not bothered looking at any other features of MythTV or the board/case yet (Antec Fusion Remote Max), but will do once the important (i.e. AV) stuff is sorted.

I'll post more info as time goes by and I get time to fiddle (kids, life and all that stuff).

Ah nice, I'll have to try this. I have the same board and have been using the alsa drivers in kernel 2.6.26 and haven't been able to get digital out to work.

Have you got lm_sensors to report the temps of your mobo yet? I'm having a hard time getting it working. I'm been using lm85 module, but it just reports all zeros. And I get dmesg warning: i2c-adapter i2c-0: Unrecognized version/stepping 0x68 Defaulting to LM85.

sounds like im doing better than most of you! tee hee~~

i just got my asus p5q-em a few days ago, and aside from the seemingly impossible struggle to install ubuntu ibex alpha 6 from usb flash, everything is working... well, flawlessly. i just actually finished the install today, so ive yet to test minor things, but currently im running my e7200 underclocked at a multiplier of 6, (1.6Ghz) and only .85 volts! thats the lowest setting that my motherboard allows. the whole computer, not including monitor, only draws about 50 watts at idle. running at this speed, mythtv uses about 30% of both cores to decode OTA hdtv. (it was the debates, and i didnt watch for long, but it looked like 1080p) seems stable so far, but like i said, i havent had it up that long. i havent edited my xorg.conf file, like some of you had to do. after installing ibex alpha six it came right up to full resolution, and it runs the desktop effects on full just fine. it looks great! im really impressed with the ubuntu team. they never fail to amaze me with their ability to do this, as well as they do it, for free.

heres my xorg.conf file, for those of you with the noaccel option, or the dfi off.

seems like there should be more to it than that? ive never had such a short one before.

btw, i have nothing fancy to digital out my audio to, and i dont even have a splitter for my analog yet, so i cant help with that.
i run over hdmi at 1920x1200 on a dell 24 inch monitor, through an hdmi switch.
lm-sensors reports around 40C in both cores at idle, but the bios reports less than 30C, and i tend to believe the bios. i dont have windows (pc probe) to compare it to.

there is a howto on the ubuntu forums, seach for temperature applet, or something like that, and it should tell you how to set it up. the only one it could find for me is the module coretemp, which i had to manually modprobe but the script in the howto added to /etc/modules.