XBMC Gains Crystal HD 1080p Decoding Support

[monraaf]

Wow, this is great news. Now this is how it's done, AMD should take note.

EDIT: There's even the source code of a gstreamer plugin included in the zip file, so totem should also be able to take advantage of this. To bad I don't have the hardware (yet) to test it.

[MartjeB]

This sounds absolutely great! Can't wait to see it in action!

[Ant P.]

Broadcom hardware that actually works? It must be a cold winter in hell this year.

[linuxjacques]

I didn't realize Broadcom had started releasing source code.

They used to be my best example of difficult to work with hardware companies.

Good job Broadcom. Please keep it up!

[zhark]

Great news! Big thanks to Broadcom!

[Fixxer_Linux]

That sounds great, but I don't wanna add any new card in my box as I already got a powerfull ATI HD card (4870), with UVD2 enabled that is *just* waiting a driver to work.

I'm no dev at all, but why this broadcom chipset that comes from nowhere could have a driver faster than my 4870 for which the full specs have been released by AMD more than a year ago ??
Is it so hard to write a driver for a GFX card rather than writing a driver for some MPEG extension card ?
And why all major multimedia linux apps could use instantly this broadcom driver while they can't do it with ATI ?

I guess this comes from the lack of XvBA documentation from ATI. But why is it so difficult, as I guess there is something strong enough to prevent ATI to gives the clues to use it.
After all, ATI has made huge efforts to support open-source, so why is it so difficult ?
Why, as an ATI customer of latest GFX cards, should I be "humiliated" by an expansion card costing 20$ that could do more than my monstruous-powerfull ATI card ???
Should I imagine being forced to buy such card if I just want to read some AVCHD or blu-ray on my linux distro as I already got all that should be enough inside my mobo ?!?

There is things I definitely don't understand. 8|

ATI I don't even blame you as I don't know really why XvBA has been so secret (certainly good reasons, as I guess this is not making you laugh of developing something that no-one can use), but do you listen this ?

[Jorophose]

Wow, that's pretty effing amazing coming from broadcom! I remember when we used to curse that name. Heck, I was cursing it just yesterday as I tried to get a wifi chip working. :P (but it did work. seems ubuntu 9.10 botched the hardware driver app thing)

(as a side note: what's the story on the broadcom-sta driver? is it considered non-free because it loads closed firmware, or is it just source-inspectable and not copy[left/centre]?)

Is there any chance we'll see benchmarks or a review of this? Anyone want to volunteer results? I remember Gwenole Beauchesne said that the performance would be worse (or was it just less pretty?) than VDPAU, but I wonder if that's true? Or if it would just be the occasional wrong colour problem?

Unless the game has changed with ARM chips, no, this broadcom card isn't in them. They have NEON built-in to the CPU's die, along with a GPU, but it's a lot like Poulsbo. I've never gotten the chance to play with a beagleboard or similar product, so I can't say for sure if the driver is closed or open, or a mix of both, but I know PowerVR is typically really secretive.

EDIT: Coincidentally enough, I was just reading this:

http://www.logicsupply.com/blog/2009/11/16/the-little-pcie-card-that-could/

Yesterday and commented to a friend that open support would be nice. I guess there is a dog. :P

EDIT2: As for ATI/AMD's support for XvBA, I'm glad they decided to back VA-API... I don't know if we'll see any open drivers doing video acceleration any time soon. I think the best solution is a shader-based backend (and maybe an OpenCL app!) in Gallium3D, so that for example, Radeons could accelerate VC1, Xvid, Theora, Dirac, etc. (at the moment I think they can only do h.264 and mpeg2, right?) as opposed to a lot of different backends (VDPAU, XvBA, VA-API) that nVidia seems to like.

It could be that AMD is prepping a documentation release, or it could be that ATI really, really sucked hard with fglrx and it's taking this long to reverse it.

[chaos386]

Is it so hard to write a driver for a GFX card rather than writing a driver for some MPEG extension card ?

I would assume so. In Broadcom's case, there's only one chip to write a driver for, and it only does one thing.

In ATi's case, they have to work on video acceleration, 3D acceleration, 2D acceleration, video output, GPGPU, multi-GPU, power saving, etc. for multiple generations of chipsets, multiple GPUs in each generation, and different variants for each (discrete cards, integrated mobo chipsets, notebook GPUs, etc.).

What I'd like to know is if Broadcom found an easy way to separate HDCP/encryption from acceleration, since I believe that's a major reason why the specs for UVD/UVD2 are taking so long to come out.

[whizse]

Are there any full size PCIE cards using this?

All I could find on Ebay was mini PCIE (and the adapter cards are more expensive than the decoder).

[Jorophose]

What I'd like to know is if Broadcom found an easy way to separate HDCP/encryption from acceleration, since I believe that's a major reason why the specs for UVD/UVD2 are taking so long to come out.

Does it even cover HDCP? I assumed that was always done in CPU, isn't it? Or do the radeons/geforces do it in GPU?

@whizse: have you checked dealextreme? I'd really like this in PCI or PCIEx1 too.

(But I thought they were electrically compatible?... (x1 and mini))

EDIT: Here's some adapters, they seem fairly common:

http://cgi.ebay.ca/Mini-PCI-Express-USB-SATA-Coverter-Adapter-Card-Compu_W0QQitemZ250496885880QQcmdZViewItemQQptZLH_DefaultDomain_0?hash=item3a52c72478

(although I have no idea what performance over USB would be like, and I don't think SATA is meant to bend that way)

http://cgi.ebay.ca/Mini-PCI-E-to-PCI-E-adapter-3-antenna-Shielding-Case_W0QQitemZ170422041954QQcmdZViewItemQQptZLH_DefaultDomain_0?hash=item27adf1fd62

(you could probably remove that backplate and hack on a more generic backplate. or rig the antennas to something else. or leave them I guess.)