The State Of Open-Source Radeon Driver Features

[curaga]

The only thing that might be better with an Atom is video playback (E-350 with r600g certainly does not have hardware acceleration, not sure about Atoms).

They don't. Even the newer desktop chipset, GM45, still doesn't have it properly working.

(I don't know whether the blobs for the PowerVR atoms support any. But who wants to inflict those on themselves?)

[manmath]

I beg to differ. The 3150 - a repackaged 945 - is not faster than the E-350/450, both running open source drivers.

Well, I've not benchmarked. But from pure perception (i've both and well configured) E-350 on opensource driver is unable to play HD video smoothly, if at all. Same is true for intel 3150. Whereas on pure graphics hardware muscle everyone is aware that e-350 is better. Now the point is what's the value of running E-350 on opensource driver and getting somewhat the same (or just little more) performance as an atom??? And what's the use of "OpenSource Driver" if it's feature incomplete and offers only 25% performance of the proprietary drivers? Besides, if anyboy have tracked the development of opensource radeon drivers, it's quite clear that it's been one of the MOST slowest developing project...

[Ansla]

What's the use of running an Atom and getting "only slightly" less performance then the E-350? The E-350 is cheaper, uses a similar amout of power and has better performance even with the incomplete mesa drivers. So I'm happy with my E-450 based laptop, it's a huge improvement from the crappy Atom based one I had before.

[disi]

Well, I've not benchmarked. But from pure perception (i've both and well configured) E-350 on opensource driver is unable to play HD video smoothly, if at all. Same is true for intel 3150. Whereas on pure graphics hardware muscle everyone is aware that e-350 is better. Now the point is what's the value of running E-350 on opensource driver and getting somewhat the same (or just little more) performance as an atom??? And what's the use of "OpenSource Driver" if it's feature incomplete and offers only 25% performance of the proprietary drivers? Besides, if anyboy have tracked the development of opensource radeon drivers, it's quite clear that it's been one of the MOST slowest developing project...

This, all the statements of Atom being faster are not true and the AMD chip is cheaper.

On Linux the open drivers are much better integrated into the system e.g. work natively with new xorg-releases, xrandr etc. Nobody has to wait for AMD or NVIDIA to update their drivers for new features or bug-fixes.

[Gusar]

E-350 on opensource driver is unable to play HD video smoothly, if at all. Same is true for intel 3150.

This has 100% nothing to do with graphics. The open source radeon driver only has hardware decode for mpeg2 (and I doubt your HD video was mpeg2). The intel 3150 doesn't have any video decode unit.

People do this a lot, talk about "GPU video playback" in scenarios where the video is decoded entirely in software.

I don't know whether the blobs for the PowerVR atoms support any. But who wants to inflict those on themselves?

I believe EMGD (poulsbo driver) has VAAPI, don't know about cdv (cedarview driver). Though you're 100% correct about the inflicting part. ValleyView can't get here soon enough.

[brent]

The Radeon open source drivers have many issues with APUs. The biggest is probably that the lack of power management permanently forces the GPU to a low clock, limiting performance to about 30% of what it should be. Additionally, 2D rendering is quite unoptimized, and doesn't performance at all like UXA or SNA. I have noticed sluggish video playback (with software decoding) too, a slow Xvideo implementation appears to be the cause.

In practice, E-350/450 graphics indeed perform just like GMA950 in most cases, if not worse.

[manmath]

In practice, E-350/450 graphics indeed perform just like GMA950 in most cases, if not worse.

Yes, exactly that's what I wanted to say. In practice fusion apus on opensource amd graphics driver is comparable to pinetrail atom graphics. It's really pity and very bad of amd radeon opensource efforts.

[Panix]

I got a Radeon HD 6450 to use with a TV so with an HDMI connection. No, not some complicated TV Tuner card or anything too drastic. It CANNOT detect anything properly. The resolution is not detected properly. This is with VGA. If I use it with HDMI, the sound is scratchy and the resolution is detected but it cannot detect the size of the TV. It interprets a 72" TV when it's 24"! LOL!

I am so impressed with the features of the FOSS driver!!!!!

[droste]

I got a Radeon HD 6450 to use with a TV so with an HDMI connection. No, not some complicated TV Tuner card or anything too drastic. It CANNOT detect anything properly. The resolution is not detected properly. This is with VGA. If I use it with HDMI, the sound is scratchy and the resolution is detected but it cannot detect the size of the TV. It interprets a 72" TV when it's 24"! LOL!

I am so impressed with the features of the FOSS driver!!!!!

This has to be a problem of the driver and not some faulty EDID of the TV, right?

[ChrisXY]

It interprets a 72" TV when it's 24"! LOL!

I am so impressed with the features of the FOSS driver!!!!!

Have you tried dumping the edid data from vga and from hdmi and comparing them, possibly with another machine where it works?

Where is your bug report?
AMD doesn't have that many linux graphics people and the open source team is even smaller. Sure, you can make fun of them, but you could also help them by telling them when something doesn't work.