The Truth About ATI/AMD & Linux

[Svartalf]

Well, the article describes things as I suspected them to be on the Linux side of things.

Personally, I don't expect a whole lot of things out of them for a while yet.

Here's a hint: My new laptop has a GeForce Go 7600 in it instead of an AMD part- sought out for that VERY detail.

[yoshi314]

i was expecting a development roadmap of sorts in the article. e.g. when do the maintenance releases end?

i'm betting that if something new is going to come with the drivers this year it will be on the very last day of december.

besides the article title is exaggerated. some of the truth was there, but before the slides, etc were censored by internal verification process in ati/amd.

[rolz]

i'm betting that if something new is going to come with the drivers this year it will be on the very last day of december.

Yeah .
"something substantial" announced in february, "delivered" in june, and they end with the phrase "AMD should have some interesting things in the pipeline for later this year...".


Lets just say i'm not holding my breath any more.

[Raven3x7]

The way i see this, the way they do releases has a lot of testing overhead and may not be the most appropriate way to develop a graphics driver for linux. I mean what's the point of having a driver every month when you cant support the latest linux related technology and features anyway? Its not like they have to optimize performance for the tons of linux games that get released every month . Im thinking a feature release every 3 months plus a bugfix release every month between feature releases would be a lot more effective. As for when we are going to see new stuff my guess is any new features related to OpenGL will go into the Orca codebase, so we're probably only getting bugfixes and Catalyst Control Center Improvements until that gets released.

[Vakilik]

Yeah, the information AMD/ATI gave for this article wasn't very suprising. I was expecting a little more detail and a roadmap, but we didn't really get any of that. Hopefully throughout the year they'll release something worth while. I mean, what good does anti aliasing and anisotropic filtering support in ATI Cataylst Control Center do when the performance is pretty much too poor to even use it anyway?

Its nice they're doing something at least, but they're not really focusing on the root of the problem in my opinion. But as they say, time will tell.

[bgreen]

OpenGL 2.0 support. NVidia has it (has for a while), ATI doesn't.
I've never seen this issue addressed, has anyone else?
I'd like to know where OpenGL 2.0 support (including GLSL) fits into ATI's development cycle.

[Svartalf]

Lets just say i'm not holding my breath any more.

Heh... Let's just say I wasn't holding my breath too much back when it was announced. Let's just say I'm not holding my breath for anything decent from them for at least 12-24 months from now.

Now, I might get surprised, but with their past performance on things, I doubt it.

[Svartalf]

OpenGL 2.0 support. NVidia has it (has for a while), ATI doesn't.
I've never seen this issue addressed, has anyone else?
I'd like to know where OpenGL 2.0 support (including GLSL) fits into ATI's development cycle.

The bulk of it is from a different group from Matthew's. No 2.0 support? Even more disappointing. Typical, though. While ATI's bunch shipped Vista OpenGL support before NVidia, it's missing a decent bit of the ARB extensions. At least part of the stuff that NVidia DOES provide and several games, in development, use. Oh well...

[givemesugarr]

I can understand your point of view, but i'd like to say some things:
first, some time ago i would have spend several hours or days to compile, install, configure the old ati driver. it was so buggy and with very low performance that many times i was induced to drop down linux and get back to windows. some other drivers of my notebook still make me think this way. now i have to say, that after a year that i'm using on a regular basis linux, at least the drivers don't give me problems when upgrading, they don't give me great problems when using them and they don't crash as they used to do some time ago. and on this point i don't think that someone could say something.
the second thing is that amd, in my opinion has chosen to focus on a very important aspect: the full rebuilding of the video driver from scratch. it has released the new windows and linux catalyst driver and in this one i can say that there aren't so many bugs as they where in the old version. or if they are, they are well hidden. this full rebuild will took quite some time and it will only support basic features. and i fully agree with amd on this point of view: why would you add aiglx support in a driver that hasn't still the power to support it?! it would be useless to use beryl or compiz (which i prefer because it has a more delineated progress) at 10fps and with the risk to break xwindow. it is a very stupid thing to do!! and why using somthing like beryl or compiz, in the first place?! isn't linux a terminal based os?! if you want to use this bunch of visual crap (which will take you to the hospital if you'll use it on regular 8-10 hour basis) use it with xgl, which was developed for this! i really think that amd is doing the right thing by not releasing aiglx or opengl 2.0 support (who needs it if the driver is not performant enough to support it) and by concentrating on developing a very stable base and a very easy to use installer. if you have a good base you'll be able to develop new features more rapidly than developing them on a instable base. think about gentoo: if you're using the stable tree you could be a little backward with the new features but you'll have a better system, that is less likely to broke down when you're using it. amd is doing the same: the old driver was so bad that it would have taken more time to fix it than to release a full new one. but you have to be a little patient and wait for it to grow up.
you're all right: the support isn't the one we want, the performance isn't the one we want, but i think that this time amd has taken the right path in developing the driver. maybe we'll have to wait some time for a very new feature but meanwhile we'll have a driver that is always bugfixed almost in real-time, that is easy to install in almost any distro, that is a unique version for all the platforms and that is almost developed for the newest kernel version. I think that nvidia has not this advantage. i use the nvidia drivers on a desktop and i always get mad when i have to wait some hours to build the driver, then install it, hoping that there aren't problems and so on. it is true that it has more features, but consider that it didn't needed to be rewritten from scratch as the amd's one and give amd about one year of development and then confront the advancement of amd driver with the nvidia one. i think that the real ati driver has come out the last month with the new catalyst, so on may 2008 we should check amd's efforts and see if they have been good or bad.
last thing: no one has ever constrained someone to buy an ati board so i think that this "i'll never buy an ati again" or "if they don't come out with this feature, my next pc will be a nvidia powered one" is the crying of a baby. if you use your pc for something useful you'll not use crap like compiz/beryl or gaming and so, this ati driver that is out now will only do for you needs. and if you really want good opensource drivers go pick and intel board, cause is cheaper that nvidia or amd-ati and it has full opensource support.
so try thinking about what i've said and try not pushing or threating ati support team cause you'll get nothing than more sorrow from this. let them work and develop and try supporting them with bug signaling and useful informations and not only with childish requests.

[Svartalf]

so try thinking about what i've said and try not pushing or threating ati support team cause you'll get nothing than more sorrow from this. let them work and develop and try supporting them with bug signaling and useful informations and not only with childish requests.

Oh, I'd never threaten Matthew's team- I've nothing but respect for them with what little in the way of support and resources that ATI's management has given him over the years on the task set before him. I know PRECISELY what he's been given to work with in the past and what he's facing in his near future. However...

It doesn't get their employer ANY slack whatsoever.

Imagine all the problems you state with the Linux driver with the Windows driver- how long do you think they would have been in business if they did that to that segment of the entire market? Not terribly long.

Why on earth should we be "thankful" to the company that once gave out technical information to allow a 3D driver codebase to be based off of their chipset that then closed EVERYTHING up so we are forever beholden to them and only the R100 and R200 series chips are actually properly supported? Why on earth should we be "thankful" for something that is precisely HALF as fast as the Windows side of things, which is known to be suboptimal in the OpenGL space in the FIRST place?

We shouldn't, really. I'm thankful for the work done up to this point, but have chosen to support other manufacturers because their employer has seen fit to NOT support us throughout all this time, instead chasing the DirectX boondoggle Microsoft has had everybody and his dog chasing. If the story changes, I'll change MY position. There's new executive management in place, we might even see changes that fix things- but I'm not seeing this come to pass for at least 1-2 years from right now with things as they're currently going. And, I will share with you that I've done my fair share of bug reporting for them- with no responses forthcoming on their RS480 support, not that it'd be all that great anyhow- they don't have hardware transform, clip, and lighting support on that chip, in spite of it being flogged as being comparable to the X300 in most of their literature.

What you see here is a large number of VERY dissatisfied customers venting on the subject of the sources of that very dissatisfaction.