AMD Dropping R300-R500 Support In Catalyst Driver

[pawstar]

Those all work for me with my X1950. Still, I hope you've been filing bugs if everything's broken for you.

Using a dual-monitor setup? The driver isn't all that bad for a single monitor, but enabling dual-screens is what creates the mess for me.

However, I haven't actually been using the opensource drivers that much. Its been try it out -doesn't work, go back to fglrx.

With XAA, things are unbearable, enabling EXA the configuration get more usable. Part of the problem I think is due to my huge resolution for my main monitor (2560x1600) i.e. its using a dual-link connection. Dunno, there must be some bugs there. For one, I cant set a virtual resolution beyond 4060x1600 without getting the ridiculously slow redrawing on the screen - but this means I have overlap on my monitors

As an aside, I know I am not the only one having problems with the open source drivers:
http://forums.gentoo.org/viewtopic-t...c6b249680b38ad

some other threads as well.

P.S. I am just pulling in xf86-video-ati-6.12.0 which hit the portage tree so will see if it helps out.

[bridgman]

Pawstar, which GPU do you have ? AFAIK the 4K limit for total accelerated screen area (sum of monitors in an RandR environment) is a hardware limit, not a driver issue.

There is some work being done (look up Ajax's "shatter" work) to separate out the screens so that each screen can use the entire hardware limit for itself, but I don't think that is available anywhere yet.

[pawstar]

Pawstar, which GPU do you have ? AFAIK the 4K limit for total accelerated screen area (sum of monitors in an RandR environment) is a hardware limit, not a driver issue.

There is some work being done (look up Ajax's "shatter" work) to separate out the screens so that each screen can use the entire hardware limit for itself, but I don't think that is available anywhere yet.

R580 (Radeon X1950 XTX) and RV610 (Radeon HD 2400 XT)

[adamk]

Those all work for me with my X1950. Still, I hope you've been filing bugs if everything's broken for you.

Other people filed these bugs, but they do touch on issues pawstar listed:

https://bugs.freedesktop.org/show_bug.cgi?id=19215
https://bugs.freedesktop.org/show_bug.cgi?id=13405
https://bugs.freedesktop.org/show_bug.cgi?id=8056

The last one is an issue I would very much like to see resolved. Q4 is unplayable with s3tc support, and various other games would have mucn improved performance or textures with working s3tc.

Adam

[bridgman]

R580 (Radeon X1950 XTX) and RV610 (Radeon HD 2400 XT)

Ahh, OK. That explains some things.

AFAIK the multi-GPU support for Radeon cards recently added to fglrx (aka Multiview) was for HD2xxx and up GPUs only. Not sure what happens with a mix of pre-6xx and 6xx cards but I wouldn't expect it to work. If you need >2 screens then your best bet would definitely be a pair of similar cards, both HD2xxx or higher, and the fglrx driver.

The open source drivers are based on RandR multiscreen support, which currently doesn't extend past a single GPU. I don't *think* there is a way to usefully run multiple GPUs in a single desktop with the open drivers but maybe agd5f or one of the other devs can comment.

The problems you mentioned when running with a single GPU and the open drivers seem like things we should be able to get working today, so if you want to provide more details we can probably figure them out now.

[pawstar]

Ahh, OK. That explains some things.

AFAIK the multi-GPU support for Radeon cards recently added to fglrx (aka Multiview) was for HD2xxx and up GPUs only. Not sure what happens with a mix of pre-6xx and 6xx cards but I wouldn't expect it to work. If you need >2 screens then your best bet would definitely be a pair of similar cards, both HD2xxx or higher, and the fglrx driver.

The open source drivers are based on RandR multiscreen support, which currently doesn't extend past a single GPU. I don't *think* there is a way to usefully run multiple GPUs in a single desktop with the open drivers but maybe agd5f or one of the other devs can comment.

The problems you mentioned when running with a single GPU and the open drivers seem like things we should be able to get working today, so if you want to provide more details we can probably figure them out now.

Yes, please do help. I created a new thread for the topic with some more info http://www.phoronix.com/forums/showthread.php?p=66988 .

[pedepy]

well here's where I have a problem with this support drop: if the current state of the fglrx driver was just fine, I wouldn't have a problem. Thing is, I'm stuck with using this suck ass version of a driver with half-completed features like dri2 support and proper video acceleration, horrible showstopper bugs and issues which shouldve never made it into an official release, and so on.

I'm also worried that'll never be able to upgrade my kernel beyond 2.6.28 ever again ... wtf is it ?

This laptop is less than two years old. I can't change my video card ! .... IMO amd/ati is in no position to start playing that game againt nvidia/intel.


Isnt there anywhere we can complain to ? I think that hardware manufacturers ought to be obligated to provide software support to be able to use their products for a reasonable duration of time. Otherwise, imagine you buying a 500$ video card and be unable to use it with the latest hardware/software after a 12, 8 or 6 months ....people wouldnt go for that, but this is exactly what is happening.

[pawstar]

well here's where I have a problem with this support drop: if the current state of the fglrx driver was just fine, I wouldn't have a problem. Thing is, I'm stuck with using this suck ass version of a driver with half-completed features like dri2 support and proper video acceleration, horrible showstopper bugs and issues which shouldve never made it into an official release, and so on.

I'm also worried that'll never be able to upgrade my kernel beyond 2.6.28 ever again ... wtf is it ?

This laptop is less than two years old. I can't change my video card ! .... IMO amd/ati is in no position to start playing that game againt nvidia/intel.


Isnt there anywhere we can complain to ? I think that hardware manufacturers ought to be obligated to provide software support to be able to use their products for a reasonable duration of time. Otherwise, imagine you buying a 500$ video card and be unable to use it with the latest hardware/software after a 12, 8 or 6 months ....people wouldnt go for that, but this is exactly what is happening.

+1
Well said. Exactly how I feel.

[rbmorse]

Have you looked at AMD's financials recently? There's no guarantee the company is even going to exist in 12, 8 or 6 months (although I hope it is...the stock is bound to be worth something, someday).

[bridgman]

well here's where I have a problem with this support drop: if the current state of the fglrx driver was just fine, I wouldn't have a problem. Thing is, I'm stuck with using this suck ass version of a driver with half-completed features like dri2 support and proper video acceleration, horrible showstopper bugs and issues which shouldve never made it into an official release, and so on.

Sure, but dri2 support is still at a "science project" stage with all the drivers, not just AMD/ATI, and video acceleration is already working nicely in the open drivers.

I'm also worried that'll never be able to upgrade my kernel beyond 2.6.28 ever again ... wtf is it ?

It's possible, but it would require that (a) every active open source driver developer in the world simultaneously disappeared along with their PCs, and (b) every copy of the open source driver code disappeared off all the public servers at the same time.

It is a risk, but honestly it's a pretty low risk.

This laptop is less than two years old. I can't change my video card ! .... IMO amd/ati is in no position to start playing that game againt nvidia/intel.

Isnt there anywhere we can complain to ? I think that hardware manufacturers ought to be obligated to provide software support to be able to use their products for a reasonable duration of time. Otherwise, imagine you buying a 500$ video card and be unable to use it with the latest hardware/software after a 12, 8 or 6 months ....people wouldnt go for that, but this is exactly what is happening.

What exactly would you complain about ? That mean old AMD is going to be supporting your Linux system via open source drivers but you demand that we support you with the binary drivers you hate instead ?

Again, let's separate out two issues.

One is moving some families of GPUs to a lower support level after a few years. Nobody likes it, including us, but we have no plans to abandon you, just to reduce the rate of ongoing updates for all OSes, not just Linux.

The second issue is providing ongoing support via the open source drivers rather than continuing to develop fglrx for 3xx-5xx GPUs. I know you probably feel that is somehow cheating and that we should be MADE TO SUFFER, but as long as your laptop works well do you really care ?

Most of the areas where proprietary drivers have a sustainable edge don't really apply to 5xx and lower GPUs. The shader compiler in the 3D stack is one of the "crown jewels", but 5xx and lower GPUs have vector ALUs so the shader compiler can't do all that much optimization anyways. The 6xx and up have superscalar ALUs where a shader compiler can pack up to 5 independent instructions into a single clock if it's smart enough, and that's where open source drivers will have a tough time catching up with fglrx.

Same for video acceleration. UVD starts with 6xx and up (with one relatively rare exception) so you aren't missing out on XvBA with a 5xx or earlier GPU. I expect you will see shader-based video acceleration in the open drivers before you would see it in fglrx anyways; again, going forward I really do think you will be happier with the open source drivers than you would be with a reasonable continuation of our current fglrx development.