[Hannes The Hun]

nvidia and VDPAU is recommended for *painless* GPU-assisted HD acceleration. VAAPI for AMD/ATI makes good progress, so there's definitely enough "sense" in the Fusion platform even with linux, but we've had a lot of talk about this already in this thread.

[sylvio2000]

Ok, I dont wanted to state this as a fact

But to sum it up, atm the better solution is VDPAU and zacate would not be the ideal solution (although I would prefer AMD )

[Hannes The Hun]

yeah, that's about right

[Robotica]

If you look at the experiences all over internet it is no problem to have hardware acc. with Mplayer (only since recently and not that perfect as VDPAU but workable). XBMC doens't have a good track record like Mplayer since there isn't much development. Have a look at TRAC: http://trac.xbmc.org/log/trunk/xbmc/core...?rev=29053

This is the case while (XBMC) users reported workarounds and fixes for the problems.

However, VAAPI is actively developed since then: http://www.splitted-desktop.com/~gbeauch...yer-vaapi/ and even Xvba is developed: http://www.splitted-desktop.com/~gbeauch...vba-video/ Same for libva: http://www.splitted-desktop.com/~gbeauchesne/libva/

Some related and unsolved XBMC problems:
http://forum.xbmc.org/showthread.php?tid=81685

So for now, VAAPI of libva aren't the problem; XBMC is. But that is open source. It takes time for quality patches to get this working. Currently, teamXBMC has no interest in developing this feature and therefor some competitors are ahead.

And knowing history, I can imagine that there is no interest in workarounds and quick fixes.

[sylvio2000]

Mmh, maybe I havent found the answer, but is there any reason for teamXBMC to not develop vaapi support?

I dont want to blame anyone. You guys do a great job. I just would like to know why?

[Temar]

I think the reason it hasn't shown up with the open source AMD driver is simply because those who are developing that have barely gotten the basic 3D down- x264 decoding is icing on the GPU cake.

No, the reason is much simpler: ATI/AMD did not release any specs on the hardware which is used for video decoding (UVD). If we will ever see an open source implementation for hardware video decoding it will be purely based on shaders as XvBA is only available when signing a NDA.

On the Phoronix forums some of the ATI open source driver developers are very active. The basic problems they are facing are:

• The specs are incomplete => no video acceleration until someone writes an implementation based on shaders
  
• They are not allowed to use any code of the catalyst driver as many parts of that driver are third party ip
  
• They have to build the driver as a clean room implementation.
  
• They had to focus on the gallium architecture first before writing the actual driver.
  

As Gallium3D is mostly done now, the pace of the open source driver development should increase. However we will never see a driver which is on par with the catalyst, as the open source developers simply do not have access (or are not allowed to use) the full specs.

[Temar]

For those interested in Fusion & Linux:

http://www.phoronix.com/scan.php?page=ne...&px=ODc5Mw

There already is an open source driver and the catalyst will also support this chip from day one:

For those considered about maximum performance and fully taking advantage of Fusion, AMD's Catalyst Linux driver should also support this hardware from day-one. It's also not known to what extent this initial open-source Fusion driver is supported. Kernel mode-setting support from AMD is now a given, but it will be interesting to see if this initial code drop also contains Gallium3D support, which would be awesome. It would be amazing if there was also video acceleration support, but due to the same problems barring AMD from opening up their UVD engine on the Radeon GPUs, it's probably unlikely to see that support officially come from the company.

[Robotica]

Great info Temar!

Zacate (E350 / E240) will hit the market in December: http://news.ati-forum.de/index.php/news/...menden-apu

(but we also have to wait for new motherboards with the new sockets and those motherboards need to be mITX; it will probably take a few months for HTPC-usage)

[sylvio2000]

I hope not! I cant wait so lang for fusion

But I have read that Asus is already working on a mini-itx board based on Ontario.

This news is from august, but in german:
http://www.computerbase.de/news/hardware...md-fusion/

[Temar]

I hope not! I cant wait so lang for fusion

I really can't understand why you are so anxious to get your hands on this new APU.

Poofyhairguy explained it pretty good. For XBMC this APU only is an option if you want to use Windows (which in my opinion is the worst choice of operating system for a standalone XBMC box).

[Robotica]

- I like this hardware to hit the market so there will be some competition to Intel Atom.
- Linux is able to decode 1080P with this platform.

[Temar]

- I like this hardware to hit the market so there will be some competition to Intel Atom.

I agree with that.

- Linux is able to decode 1080P with this platform.

There is a huge difference between being able to decode 1080p and having a nice user experience. Did you ever try ATI+VAAPI+Linux? I recently had to and I can tell you from personal experience that the ATI driver still has lots of problems when it comes to decoding 1080p content. It's not about CPU usage, but about distorted images.

Often the colors just shift a little bit, so i.e. a dark green becomes a bit lighter and all that within the same scene of a movie. For a viewer this is very irritating. Also there are many files out there which just don't play at all. You get huge distortions on the screen and sometimes the whole image gets shifted to the left or right.

These are just the problems when playing 1080p content. Before even thinking about watching a movie, you have to fix the overscan problem to get a fullsize image on your screen (you can use aticonfig for that). This has to be done on each reboot or each time you resume from standby. While this is not much of a problem when booting your media center, as you can add these commands to your XBMC startup script, it really gets annoying when resuming from standby. If you resume from standby you either have to restart XBMC or you have to add these aticonfig commands to you pm-utils resume scripts. This however will crash your X server if the commands are executed to early (the ati driver is not yet fully initialized).

When talking about suspend/resume it even get worse. After you figured out the correct timing for restoring the correct screen size, you still won't have a reliable S3 mode. Every 10-30 resume operations the driver will just crash. Stoping the X server before you suspend is also not an option because the driver will also crash if there is any framebuffer driver like uvesafb active. And if you can't use uvesafb you won't have a nice plymouth bootsplash.

ATI on Linux really is a pain in the a**.

[Robotica]

I didn't try this yet but I want try this very soon. I am aware of the problems you describe. Although, you didn't mention which driver you used. Also taking in account things Poofyhairguy said.

On the other hand I know things are possible all be it very difficult and time-consuming ATM. For HTPC scenario this can work out. See this link how to setup hardware acc. for AMD/ATI using flgrx.

Besides, I believe more progress will be made if this platform is a competitive deal to ION. Check this topic for development and success stories.

So before writing this platform down on forehand let's see how things develop. This was the conclusion of this topic a few pages back in this topic.

ps. this topic has some interesting stuff about shaders and xbmc: http://www.phoronix.com/forums/showthrea...071&page=6

[Temar]

I didn't try this yet but I want try this very soon. I am aware of the problems you describe. Although, you didn't mention which driver you used. Also taking in account things Poofyhairguy said.

I tried many different combinations of XvBA and the fglrx driver. None worked without flaws.

On the other hand I know things are possible all be it very difficult and time-consuming ATM. For HTPC scenario this can work out. See this link how to setup hardware acc. for AMD/ATI using flgrx.

Believe me, I know very well how to set up ATI hardware acceleration. I tested all the latest code which was available at that time. :-)

Besides, I believe more progress will be made if this platform is a competitive deal to ION. Check this topic for development and success stories.

The linked topic is a testimony to the failure of ATI on Linux. That was the topic I got my information from, which driver/xvba combination was supposed to work. In fact these combinations do work but there is always a BUT...

So before writing this platform down on forehand let's see how things develop. This was the conclusion of this topic a few pages back in this topic.

While I can't share your optimism, I wish you all the best. I justed wanted to share my experience, so less experienced people are warned. Someone who wants a working setup should not buy ATI atm.

[Anastrophe]

<comment from the left field>it's funny how we're talking about ARM+Crystal etc. - I wonder if Roger/Sophie Wilson had anything to do with the Crystal HD?</>

It wouldn't suprise me in the least.