RadeonHD 1.2.4 Driver Released

[phoronix]

Phoronix: RadeonHD 1.2.4 Driver Released

While the Novell developers responsible for the xf86-video-radeonhd driver have been busy working on OpenSuSE 11.1, we have managed to get a new driver point release before the end of the year. Matthias Hopf has this morning announced the availability of the RadeonHD 1.2.4 driver. Added since the RadeonHD 1.2.3 driver in October is HDMI audio support, support for the ATI RV710 and RV730 GPUs using DCE 3.2, rotation support, RandR 1.3 panning, and many acceleration and build fixes...

http://www.phoronix.com/vr.php?view=NjkyNQ

[bulletxt]

"but it's starting to look like we won't see it this year and perhaps not until FOSDEM 2009 in February. The code and documentation is still stuck in its legal review process. "

yea they can even release it in june nothing changes. they have lost all their credibility and don't deserve 1€ from me in the future.

[mityukov]

hmm.. Novell.. which driver will be there in OpenSUSE 11.1?

If radeonhd, it's bad, because my card (R500 family) has more features when used with radeon.

[bugmenot]

What is the state of RS780 (HD3200 exactly) in radeonhd?
Is it still better to use "radeon" driver or AMD catalyst drivers?

[bulletxt]

What is the state of RS780 (HD3200 exactly) in radeonhd?
Is it still better to use "radeon" driver or AMD catalyst drivers?

that's like asking if it's better to ride a donkey or an ostrich. oh, if you are able to ride one of them, please go directly to the AMD legal department,get all the documentation and publish it on the net. if you wait for them, you'll have to wait that new card series comes out in the market...there is no legal issue about that doc, it's all about business. they are just waiting their cards (2xx series and 3x series) get old. what liars holy sh%$.

[bridgman]

The package covers both 6xx (HD2xxx, HD3xxx) and 7xx (HD4xxx) so waiting for the 2xx and 3xx series to get old wouldn't help.

I guess we could have limited the package to only cover 6xx and been able to release a bit sooner but I didn't think that was the right choice. We ended up having to do a lot of the initial bring-up work on 7xx anyways, so cutting back to 6xx-only seemed kinda counter-productive.

EDIT - just noticed Michael was actually talking about "3D stack next year". That's probably right, not because of IP issues or anything, just that the 3D support is a *big* task. My SWAG estimates were a total of 28 developer-months to bring 6xx/7xx 3D up to the same level as 5xx, and that the equivalent of 18 developer-months of work had been done, either from Matthias/Alex/Richard writing code and working out problems or from internal code we were able to leverage. Adding 5xx 3D support was a much smaller task by comparison since the 5xx is a lot like a 3xx/4xx other than the actual shader instructions - the big fixed-function blocks like VAP didn't change much. Going to 6xx every block except the command processor is totally different.

The first code we push will be :

- basic EXA acceleration,
- Xv (hopefully with the colour-space-conversion shader working; everything is kinda purple without it ;)),
- DRM
- a demo program which shows how to make a 6xx or 7xx perform the major 3D functions.

With luck there will be a ~190 page register spec doc to go with the code.

[bulletxt]

I'm glad you replied with a serious post. However I am sure you understand my disappontment about all this. so much time for doc, fglrx driver that i'm not even sure how to define it, and so on.

I will be happy and change my mind only when I see facts. Hopefully this will happen soon. But untill that day I have more reasons to think bad about AMD then good.

But this is just for the Linux side, I can't complain of anything on the Windows side and that's already something.

[bridgman]

Serious question for you. Do you really think the time for 6xx/7xx has been unreasonably long ? We finished fixing up the 5xx documentation in April and the 3D driver work took until ~July, so that was maybe 7 months from start to finish. I think the 6xx/7xx will end up taking a couple of months longer than 5xx did, and the complexity is maybe 3 times as high.

I had definitely hoped that tcore would let us "shortcut" the documentation process and get something out early, but that didn't work out. Once we put together all the pieces we needed to show how to write a 3d driver (tcore, tgl and a bunch of other test programs) the resulting package was bigger than the finished driver would have been, and a lot harder to get through IP review.

Talking about tcore was a mistake. That's the last time I talk about something in public without doing a line-by-line inspection first :(

[Kano]

What do you tell somebody with a shiny new laptop and ATI card that locks up with fglrx when trying to watch DVB-T via kaffeine?

[bulletxt]

well, I will actually explain better. I'm not really disappointed for the doc releasing time, but from the beginning. till last year, fglrx driver was a real total crap, and you know this better than me. then amd finally released a better driver and opengl performance improved a lot. but there were(and still are) a lot of issues. then amd started releasing doc. but as of today fglrx is far from perfect, doc is incomplete. so in other words I am disappointed for all this put together.

"That's the last time I talk about something in public without doing a line-by-line inspection first "

I appreciate your detail post and suggest to don't do it again :) people just want to know when their card will work out of the box :) maybe I want to know more than a general user because I ask myself a lot of questions, but you said enough and it's fine like this :)

thanks for your reply ;)

[bridgman]

What do you tell somebody with a shiny new laptop and ATI card that locks up with fglrx when trying to watch DVB-T via kaffeine?

I manage the open source graphics project so you can guess what I would be thinking :D

Seriously, I don't have a good off-the-top-of-my-head answer for a DVB-T issue. The DVB/T standard isn't used in any of the countries where we do Linux PC driver development, so we would need to find a way to repro the problem with a different source. If the laptop owner can repro the problem without requiring DVB-T that would sure help.

There is a lot of work being done on the fglrx video stack now (the focus previously was primarily on the 3D stack) so this is one area where I expect to see ongoing improvement.

[Kano]

Well DVB-T is an example. I guess the xserver just locks because xv is tried to be used. Could not verify this yet, but Xorg 7.1.1 + xv seems be pretty unstable.

[DanL]

What is the state of RS780 (HD3200 exactly) in radeonhd?
Instead of ranting about a non-existent conspiracy theory, let me try and provide a real answer to your question.
The state of RS780 in radeonhd is very similar to the open-source radeon driver. They provide basic mode-setting and mesa software acceleration. If you want any kind of hardware video acceleration, you'll have to use Catalyst.

[korpenkraxar]

Just to make it a bit easier on potential customers, which series of cards use the RV710 and RV730? Is there a simple table available somewhere that helps one translate between chips and cards?

Well, too bad you AMD/ATI guys did not make it in terms of delivering spectacular video improvements for fglrx to compete with Nvidia or publishing the docs or a more competent open source driver for new cards in time for the holidays.

I just bought myself a new low-end graphics card for my HTPC based on the latest driver announcements and as much as I had hoped it to be an ATI card this time around, it wasn't. A case of ideology caving in for pragmatism and spouse approval.

But you know what, I still remain hopeful that 2009 will change much of that and that we by 2010 vaguely will remember the olden days when ATI cards were problematic in Linux and other *nixes. Nvidia should better have a secret open source project going already or there might be trouble on the horizon for them now as graphics innovation on Linux finally seems to get going and targets will probably move quite a bit for some time.

So bridgman & Co, keep the hard open source work going and see you again next year! Happy holidays!

[deanjo]

hmm.. Novell.. which driver will be there in OpenSUSE 11.1?

If radeonhd, it's bad, because my card (R500 family) has more features when used with radeon.

RadeonHD would be the default, but there are packages for radeon as well.

[bridgman]

Just to make it a bit easier on potential customers, which series of cards use the RV710 and RV730? Is there a simple table available somewhere that helps one translate between chips and cards?

RV710 is HD43xx and HD45xx, RV730 is HD46xx, RV770 is HD48xx.

Well, too bad you AMD/ATI guys did not make it in terms of delivering spectacular video improvements for fglrx to compete with Nvidia or publishing the docs or a more competent open source driver for new cards in time for the holidays.

I don't know about you, but we still have a lot of work to do before the holidays :)

So bridgman & Co, keep the hard open source work going and see you again next year! Happy holidays!

Thanks !

[bulletxt]

"I don't know about you, but we still have a lot of work to do before the holidays "

you are somehow able to make me continue dreaming...just be sure to make it become true :)

[korpenkraxar]

I don't know about you, but we still have a lot of work to do before the holidays :)


Me? I just sit here reloading the phoronix page in my browser all day ;-)

No I am actually on paternal leave (we have quite a bit of that here in Sweden) struggling to keep up with our 1.5 year-old so yeah I'm pretty busy too...

... and only check the page a few times a day these days :p

PS. With the risk of going slightly off-topic, I dare anyone here to grok the Swedish parental leave rights as specified over at http://en.wikipedia.org/wiki/Parental_leave

[Dieter]

> I don't have a good off-the-top-of-my-head answer for a DVB-T issue.
> The DVB/T standard isn't used in any of the countries where we do
> Linux PC driver development

You don't need to worry about DVB-T/COFDM vs. ATSC/8VSB. Those
are RF standards, the tuner/demodulator deals with that, not the
GPU. There are FLOSS drivers for some (not all) tuner/demod
cards for both ATSC and DVB-T. The HDHomeRun tuner/demod doesn't
need a special device driver (or a slot) just Ethernet. Whether
you use a card or the HDHR, you get a transport stream out.

Worry about mpeg transport streams. Step one: get Xv and XvMC working
in the FLOSS drivers. (working *correctly*, no tearing, purple, etc.
problems) That will take care of mpeg2. Step two: pick a API and
get mpeg4, H.264, etc. working in the FLOSS drivers.

Keep in mind that the transport stream might be corrupted by bad
reception. Crashing, hanging, etc. is not the correct response.

[bugmenot]

Instead of ranting about a non-existent conspiracy theory, let me try and provide a real answer to your question.
The state of RS780 in radeonhd is very similar to the open-source radeon driver. They provide basic mode-setting and mesa software acceleration. If you want any kind of hardware video acceleration, you'll have to use Catalyst.

Thank you, kind man.

[Zhick]

What do you tell somebody with a shiny new laptop and ATI card that locks up with fglrx when trying to watch DVB-T via kaffeine?

Tell him to use deinterlacing (the best option kaffeine has available, GreedyH or similiar). I had the same issue with my dvb-s card back when I was using fglrx and found out that it was caused by interlaced video and enabling deinterlacing fixed it (well, except for one channel, but I think that's because they partialy send interlaced and uninterlaced material and the deinterlacer doesn't know how to handle this.

[Kano]

The full system locks up, i guess because of xv. Nvidia drivers never had such an issue.

[Zhick]

The full system locks up, i guess because of xv. Nvidia drivers never had such an issue.

Use Xshm, enable deinterlacing, switch to Xv and it should playback without locking up.
Also your constant rambling about the quality of fglrx/nvidia gets annoying, y'know? I don't think anyone questions that the nvidia is in fact the better blob.

[bulletxt]

No, it's not annoying and it will never be. We always have to remind AMD they have to WORK if they want to earn respect. And I'll remind them always. NVIDIA works for linux since at least 7 years, and I respect them. And since I pay for AMD cards, they have to work too.

[mattst88]

Well DVB-T is an example. I guess the xserver just locks because xv is tried to be used. Could not verify this yet, but Xorg 7.1.1 + xv seems be pretty unstable.

Have you considered upgrading? I mean, X.Org 7.1.1... really?

[oibaf]

EDIT - just noticed Michael was actually talking about "3D stack next year". That's probably right, not because of IP issues or anything, just that the 3D support is a *big* task. My SWAG estimates were a total of 28 developer-months to bring 6xx/7xx 3D up to the same level as 5xx, and that the equivalent of 18 developer-months of work had been done, either from Matthias/Alex/Richard writing code and working out problems or from internal code we were able to leverage. Adding 5xx 3D support was a much smaller task by comparison since the 5xx is a lot like a 3xx/4xx other than the actual shader instructions - the big fixed-function blocks like VAP didn't change much. Going to 6xx every block except the command processor is totally different.


What about AMD contracting Nicolai Haehnle and Corbin Simpson to work on the 3D stack? :)
They did a lot of work this summer improving the R500 code.

[Kano]

@Zhick

You know that xshm is pure software rendering? That locks absolutely badly.

[nbi1]

Phoronix: RadeonHD 1.2.4 Driver Released

While the Novell developers responsible for the xf86-video-radeonhd driver have been busy working on OpenSuSE 11.1, we have managed to get a new driver point release before the end of the year. Matthias Hopf has this morning announced the availability of the RadeonHD 1.2.4 driver. Added since the RadeonHD 1.2.3 driver in October is HDMI audio support, support for the ATI RV710 and RV730 GPUs using DCE 3.2, rotation support, RandR 1.3 panning, and many acceleration and build fixes...

http://www.phoronix.com/vr.php?view=NjkyNQ

Anyone have this working under etch? Not looking for anything fancy, just want solid 2d for my 4850. 1.2.4 works great on Lenny, but I haven't found the time to make the complete switch from my etch installation yet.
Thanks in advance.

Happy Holidays!

[Kano]

No problem to compile with etch. Just take the one from Kanotix.

[nbi1]

No problem to compile with etch. Just take the one from Kanotix.

"the one"? what? Source package? .deb ?

I was all over the kanotix site and couldn't find anything relevant. No reference to radeonhd other than the script which I already have (and don't know how to use).

[Kano]

Maybe it is just in the repository ;) Read your PM.

[rob2687]

I am impressed with the development of these open source drivers.

[Zhick]

@Zhick

You know that xshm is pure software rendering? That locks absolutely badly.
He should switch to xshm just temoprarily, since IIRC you can only enable deinterlacing in kaffeine when a video is playing (might be wrong though). As I mentioned in my post, he can switch back after he's done that.