To the AMD people: Money wasted.

[crazycheese]

I'd like UVD and OpenGL4 with top performance too, but you have a very strange definition of a "VESA superset".

AMD/ATi open source drivers are possibly the best and most featured FOSS drivers out there, period, at least up to r500. Intel drivers used to be good, but I hear they are slipping.

Can you name a better / more full-featured FOSS driver than r300g? r600g still needs time. It's not the perfect solution, but at least AMD did a lot to push it in the right direction, including documentation and lots of code.

I still have the very first Radeon card - the 3x2 geforce enemy

r600g and the rest do not need time, they need people. If you are for opensource or for amd on linux platform, the company should really have tools to know and understand your decision, and not develop drivers in the backyard.

The time advances, new features and calls appear. You have r600g full support in 5 years, but in 5 years 3 new generations roll out, new technology advances.

Its like getting propeller airplane ready in 2020 only by 3 engineers.
Im talking company policy seeing opensource as "second hand" solution.

That put, I seriously think, abadoning opensource, but leaving the documentation for just-for-fun engineers in free time, and fully focusing on working closed source driver is better than 2x 80% / 25% finished drivers. "Do one thing, but do it good" is linux.

[bridgman]

Why not approach #1, where workstation continues to focus on workstation and opensource focuses on consumer segment? 3000$ cards are for people with specific needs and demands, and in current situation I see pathway #2 - pushing proprietary fglrx in consumer area.

Sorry, I read your #1 as "dropping fglrx". If you meant pursuing a two-driver approach for #1 then I agree in principle.

The problem is that "consumer" covers a very wide range of usage scenarios, and some of those scenarios value workstation-like attributes more highly than the things a typical consumer or enterprise client user would value.

Our approach is to support both driver paths and work towards having enough overlap in capabilities that we don't have a "hole in the middle". Consumer and enterprise client users will tend towards the open driver, workstation and gaming users will tend towards the Catalyst driver, but I don't think saying *this* driver is for consumer, *that* driver is for workstation covers all the scenarios sufficiently.

Note that comments about "only N open source developers" are missing the whole point of open source drivers -- that progress is *not* limited to what AMD developers can do. The open source driver dev community is a lot larger than just our people.

[crazycheese]

The majority of users here I guess are Arch users. They pride themselves on simplicity - simple is the way to go. That having an rc.conf is sooo much better, clearer and transparent than debian's init. That debian is way to frustrating to tweak. Then they tell you how you are a noob or a "typical 'windoz' user" because you do not want to spend the rest of your natural life tweaking with ever experimental mesa and xf86-video-ati upgrades and xorg.conf settings. Go figure...

BTW, while I'm speaking of Debian, I love it how everybody nowadays tell you Debian is a rusty old distro that takes forever to release. Exactly what has happened in since Lenny was released in terms of hardware support? Development. Nothing finished. Here we are about to release Squeeze and still FOSS drivers are experimental and unstable. Those people smart enough to stick with Debian stable saved themselves all the hassle.

Im Gentoo guy and I think Arch is lacking in many ways.
Harder to maintain than Ubuntu, less options than Gentoo, a lot of "simple steps" every time I used it, low quality control as payoff for fresh releases. But they like it as it is and Im fine with that

Debian really lacked central wiki-like documentation. If you would have init system really good documented (overview, detailed, per step) there wont be no problem as obvious flaws would come out, or obvious misunderstandings.

[mirv]

because in reality most people need this because they wana run the windows apps/games on linux.

and you are wrong amd is fixing wine bugs and wine-opengL bugs but its really hard to catch the bug and makes clear what is the problem of the bug.

exampel i report amd bugs in the past and crossover fix that they fix it because amd pay for the support.

and no they don't fix it because of my 30? payed to crossover.

i really know it amd pay's wine guys for that kind of bugs.

Interesting....but you'll have to forgive me if I take a similar approach to the whole "steam for linux" thing: I'll believe it when I see an official statement.
Otherwise, I did mention that wine is good for stress testing OpenGL implementations, so I'm aware that AMD do take notice of bugs in wine for that reason. I just don't think that it's AMD's responsibility to be fixing wine (as opposed to fixing driver issues that are exposed by wine).

[pingufunkybeat]

The time advances, new features and calls appear. You have r600g full support in 5 years, but in 5 years 3 new generations roll out, new technology advances.

The whole idea is that this time (you claim 5 years) is decreasing very rapidly.

r600g was not just about writing a driver. It included a completely new driver architecture (Gallium3d), needed to reduce the complexity of writing drivers. It included a migration to kernel modesetting and kernel memory management (TTM), needed for even basic OpenGL 2. It included a huge amount of work in Mesa which simply cannot support OpenGL 3 yet -- there is no infrastructure. This is lots of work, and AMD has been helping the OSS community in making these steps.

Much of this is reusable and will help with next generations of cards and future drivers.

I agree that they need manpower. AMD is hiring another open source developer for this reason. Yes, I'd like it if they hired 10, but this is what we have, and it's loads better than the cold shoulder and animosity from Nvidia.

[pingufunkybeat]

And before I forget, there has been a complete rewrite of the GLSL compiler in the meantime, also with the attempt to make it more future-proof. This all takes time and delays free drivers.

[crazycheese]

Consumer and enterprise client users will tend towards the open driver, workstation and gaming users will tend towards the Catalyst driver, but I don't think saying *this* driver is for consumer, *that* driver is for workstation covers all the scenarios sufficiently.

I cannot imagine Joe the barber, not using PC for games or anything 3D related (3D card, no?).

Of course, unless Joe has a laptop(or similar low power device), for internet, IM, reading emails etc, for which Software Rasterizer and Framebuffer would do; just that cpu can do correct power management.

Or unless he is a hacker(a white hat, in good sense), not barber and he keeps Linux for hacking and Windows for games.

But if Joe the barber has a PC and wants Steam or similar gaming platform, which is powered by opensource ecosystem, and he has 600$ budget, from which 400$ will be invested in gfx upgrade, and he goes to the market and sees
- a working green card with proprietary driver which means reduction in security(he is not a hacker or on mobile), reduction in hardware support life(he can manage to pay another 600$ for hw and games once in 8 years), reduction in base software choice(he is not a hacker);
- a lacking in features red card, which is useful for gaming only in windows; and capable of gaming in linux with lots of bugs and absence of features compared to green card, where both drivers share same proprietary disadvantages.

The choice is obvious unless he really loves opensource and can play his games on 5970 with 10 fps.

This would mean neither of two companies can combine good drivers(definition in my above post) with opensource model.

Derivated from this, only green company has good drivers(not opensource) now and already with history.

Mr Bridgeman, please allow AMD card buyers transfer 10% of card cost to AMD to linux driver development crew. Make this process transparent, so people can vote with extra money.

If linux is not included in default money package, add it as a feature.

People will vote money, if they know 10,000 people with 15$ from each will be sufficient for opensource video accel as example.

Provide a live counter as Wikipedia did. There are people that are capable to pay for opensource results. And people with no money which would pirate or use what they have anyway, just because they have no money.

[pingufunkybeat]

I cannot imagine Joe the barber, not using PC for games or anything 3D related (3D card, no?).

What are you talking about?

Almost EVERY 3d app for Linux works fine with open drivers. There are some exceptions for which you have to run Catalyst, but 3d works.

Joe the Barber who wants to run Windows games typically runs Windows. You don't blame AMD that your Ubuntu can't run PS3 games, do you? 3d Linux drivers provide an OpenGL API for Linux programs. If Joe the Barber wants a Windows games using the Direct3d API, then Joe the Barber installed the wrong operating system.

I understand that running WINE is a legitimate wish of some Linux users (a very small minority!) but nonsense such as "VESA superset" and "nothing 3d related" is just nonsense FUD.

[crazycheese]

What are you talking about?

Almost EVERY 3d app for Linux works fine with open drivers. There are some exceptions for which you have to run Catalyst, but 3d works.

Joe the Barber who wants to run Windows games typically runs Windows. You don't blame AMD that your Ubuntu can't run PS3 games, do you? 3d Linux drivers provide an OpenGL API for Linux programs. If Joe the Barber wants a Windows games using the Direct3d API, then Joe the Barber installed the wrong operating system.

I understand that running WINE is a legitimate wish of some Linux users (a very small minority!) but nonsense such as "VESA superset" and "nothing 3d related" is just nonsense FUD.

Why do you shift 3D to DirectX or Windows(directly or indirectly via Wine)? Whats a reason for Joe to consider Linux, if Windows does just fine? A lot of Joes consider Linux. Huge portion of them use nvidia for 3D, waay not only via wine.

Working native 3D stack is one problem less between Steam and Linux or Newgrounds-like community for 3D games and Linux. Working (see definition of good driver above) opensource 3D stack is something for Joe to consider switching the gfx card vendor. See here.

[DarkFoss]

The whole idea is that this time (you claim 5 years) is decreasing very rapidly.

r600g was not just about writing a driver. It included a completely new driver architecture (Gallium3d), needed to reduce the complexity of writing drivers. It included a migration to kernel modesetting and kernel memory management (TTM), needed for even basic OpenGL 2. It included a huge amount of work in Mesa which simply cannot support OpenGL 3 yet -- there is no infrastructure. This is lots of work, and AMD has been helping the OSS community in making these steps.

Much of this is reusable and will help with next generations of cards and future drivers.

I agree that they need manpower. AMD is hiring another open source developer for this reason. Yes, I'd like it if they hired 10, but this is what we have, and it's loads better than the cold shoulder and animosity from Nvidia.

Thank you...
I really feel lot of peeps simply don't grasp how much work was and still is being done behind the scenes in order to modernize the linux graphics stack and bring it up to a near parity with Window's Direct X..and how much still needs to be done.. Without AMD's and Intel's contributions to OSS we'd still be at OpenGL 1.4.. I guess it's easier to complain.

All Nvidia does is sit back and reap the benefits yet nothing but praise for them and their just works driver!