AMD has released "the fundamental Linux code" needed to develop open-source 3D-acceleration drivers for its R600 and R700 ATI graphic-processors series. This is good news not only for those wishing to create drivers for those potent GPU series - known to non-AMD watchers by their marketing names of ATI Radeon HD 2x00, HD 3x00 ( …
Closed drivers are a right PITA to keep properly aligned with the rest of the moving-target code-base that is desktop Linux (servers are much more stable platforms, but who needs a big graphics card on a server?). Reverse-engineered open drivers are never a touch on the official ones (through no fault of the OSS people who do a remarkable job considering what they have traditionally had to work with!)
My last GPU was Nvidia simply because their closed drivers were somewhat less technically unpleasant to work with than AMD's were at the time but this will likely make a difference to my next purchase (which won't be for some time yet).
It's about time these hardware companies start releasing drivers is source form. It's not like the hardware business model is driven by closed source drivers. The more people who buy it, regardless of operating system, the better.
They sell hardware after all, I'm tired buying new hardware only to discover that the latest official proprietary drivers are incompatible with the latest kernels, this has been especially problematic in 64bit systems. We should be able to unlock the hardware acceleration under the OS of our choosing, open source will let us do just that. Hopefully others follow suit.
The Register is getting it's news from Slashdot?
GOOD JOB AMD NOW KICK INTELS GIANT BEHIND
And if Nvidia don't keep up....
I'll be switching to ATI if things keep going this way. Nvidia's always been my choice due to ATI being dodgy with drivers. But since AMD took over it's started heading the other way.
GPUs for servers
If OpenCL works decently I would be happy to buy several big mean graphic cards for my front end servers: I like the idea of off-loading crypto work on to a number crunching machine like a GPU.
There is a whole new world out there!
Paris 'cause... well who doesn't like Paris' numbers?
drivers for windows as well?
Has anyone not realised that whilst this really does apply mainly to linux, this code would also be useful to create a driver for windows as well? I mean, it's hardware control documentation right? so if you can use it to create a driver for linux, then surely you just create a wrapper which allows you to swap it out for the win32 layer and bingo, windows driver?
anyone find a problem with that? Might help us get better stability from windows as well as linux :D
"anyone find a problem with that? Might help us get better stability from windows as well as linux :D"
I do...you don't deserve any stability on that crappy OS.
Re: And if Nvidia don't keep up....
It is more interesting if they release docs for older chipsets (both of the usual suspects). There are excellent pieces of hardware like the last generation of the HP thin clients with pre-R600 AMD graphics onboard. At present the choice for these is either the horrid flgrx driver (which is not available for most linux distros anyway) or Windows.
Same with Nvidia. Their binary driver locks up solid when used with older cards like 6xxx or 73xx on an SMP system. If the docs are released quite a few pieces of hardware that would otherwise be destined for the bin will have a new lease of life.
@drivers for windows as well?
> this code would also be useful to create a driver for windows as well?
Unless you're going to use this code to see how the hardware works and write Windows and DirectX bits... which you can download from ATI's site already. ;)
This will be useful for people running some unix + Xorg though. Hopefully we'll see decent drivers across the board (Linux distro's, BSDs, OpenSolaris...)
>I mean, it's hardware control documentation right?
Maybe this should have been pasted from the Slashdot article ->
"This code consists of a demo program that feeds the commands to the hardware, updates to their RadeonHD driver, and a Direct Rendering Manager update. With this code comes working 2D EXA acceleration support..."
So you get some example code that shows you how to talk to the GPU (generic) and then ATI's existing but slightly updated X driver (X/Xorg specific, esp. all the acceleration bits) and their interface to the Linux DRM stuff (Linux specific).
>anyone find a problem with that? Might help us get better stability from windows as well as >linux :D
One driver doesn't instantly rewrite your whole OS... I always wonder who actually rated the stability of "Linux" (Read; any semi-popular distro). Hell, it's sort of silly to even mention the word around third party graphics drivers as they're usually the reason a Linux based desktop falls over.
Been waiting for this for years. Nice one, AMD. Now lets get some proper Linux drivers hacked together.
radeonhd's efforts are extremely honourable, and they work .
im using radeonhd for a few weeks now and it does most the
common tasks required by a desktop.
It's experimental , unstable , needs a bit of research to make it work
but it's already for those common tasks, imho, better cause it resolves a set of
issues that were annoying under fglrx ( ctl alt f 1 2 .... didnt work . exiting x freezed the machine etc )
All around a good peice of work . much needs be done , but keep your ear to the ground.
mine's the coat : got to go shovel :(
@Mr. Gosselin. I think you'll find the drivers are very much part of the business model as they can make such a difference to the performance and, with respect, most hardware companies have got better things to do than chase kernel releases.
Open sourcing the spec makes a lot more sense in an OpenCL world than it does in accelerated graphics as it may well encourage new sales to customers keen to take take advantage of the cheap GPUs.
One more chance to get the HD2400M card in this Toshiba laptop running properly in OSX ;-)
I've been waiting for this to happen for a long time. Maybe now I won't have to turn off Compiz whenever I play a direct rendered game.
folding@home gpu functionality anyone?
@drivers for windows as well?
It'd be neat if someone could use this stuff to create independent Windows drivers, free of the proprietary "value adding" system-killing junk that has long been the trademark of ATI and NVidia hardware drivers.
Happy AMD are making the effort, but worried that it's ATi drivers. Last year I switched from ATi to nVidia due to the continuingly poor drivers. Personally, having been seriously bitten by ATi before, I'd wait for nVidia to catch up.
Re: Business model
>@Mr. Gosselin. I think you'll find the drivers are very much part of the business model as they can make such a difference to the performance and,
Absolutely, the drivers matter significantly. It's a terrible thing to have a decent card only to find that there are no accelerated drivers available in your OS for it. Hopefully having open sourced accelerated drivers means unaccelerated drivers are a thing of the past.
>with respect, most hardware companies have got better things to do than chase kernel releases.
This is true too, which is another great reason for them to open source these drivers.
>Open sourcing the spec makes a lot more sense in an OpenCL world than it does in accelerated graphics as it may well encourage new sales to customers keen to take take advantage of the cheap GPUs.
Disagree for reasons stated above. OpenCL actually could be delivered with proprietary binary drivers with the same level of success (or lack thereof) that proprietary graphics drivers offer. In other words, they are equally undesirable for the same reasons, regardless of the application.
I can't see how a hardware vendor or its users would benefit from closed source drivers.
"If the docs are released quite a few pieces of hardware that would otherwise be destined for the bin will have a new lease of life."
And I'd guess that exactly why they won't release them.
speedstep equivalent, finally
Maybe there will finally be a power saving mode for the GPUs. You're not running fancy 3D graphics while you're reading your mail or browsing (unless you're running "Aero" or "Enlightenment" or some other energy guzzler), so the GPUs should step down their power requirements.
If you don't want the junk (and who does?) then just download the driver installer not the full 50-odd MB omnibus. ATI and IIRC Nvidia both offer just driver installers. Or just uninstall the junk if you used the omnibus.
I think you're right; I was thinking of machines I've bought in the past (before switching from Windows) or used at work which came with the unending parade of useless junk preinstalled. Having cleaner drivers available from any source isn't going to help with that. Consider my comment withdrawn.
Does this mean we could get a 3D-aware VM guest working? (Without the phoney VMWare SVGA graphics drivers), or is more needed to be done than just tweaking the drivers...?
I love this business model
We make the card, but can't be arsed to design the drivers for it. Give out some info and let the other "enthusiasts" sort it out for us, all for free...
We sell cards, they do the work...
God I love communism!
Seriously a win-win situation for AMD
Well I'm currently thinking about getting a new graphics sub system. Currently the only decent choice for Linux is Intel. They have good on-board chipsets with good drivers. But that also means I will buy the whole board including the CPU from Intel.
Now if AMD provides usable (=open source) drivers I can not only buy a graphics card from them, but also buy the CPU from them. And I can even get multi-socket motherboards for several CPUs.
"I do...you don't deserve any stability on that crappy OS."
Glad to see 2009 hasn't seen the end of immaturity and fanboy-ness. Always find it amusing when freetards resent people being free to choose their OS.
On another note, if AMD would extend that to their mobos as well, I would be happy. My Asus M2A-VM (AMD chipset) has been really buggy under Ubuntu. Running under XP, it is much better behaved. My guess is that the BIOS is dodgy and XP is better at working around it. More likely, the BIOS is written just well enough to avoid going belly up with XP. More openness will surely help here.
As Lou stated: Guys, you are selling hardware, so stop being so dimwitted about firmware + driver specs. Many of us will give preference to companies who are more open (iPods being my guilty exception here).
I seriously doubt it's anything to do with the BIOS. Once a Linux kernel (and its initial RAM disk) is in memory, it has no further truck with the BIOS. That's why in 2.4 at least, you had to have certain drivers compiled right into the kernel. Modern 2.6 kernels can just load modules from the initrd.
Anyway, this is great news. Now, if only someone would pass a law requiring full Source Code disclosure .....
@ A J
Well... I would confess some ignorance of BIOS specifics, but I am not sure you are right. BIOS is the interface between the core Mobo hardware and the OS level at load time, innit? Afterwards...? Sometimes I guess the OS hits the HW directly sometimes not. Speaking as a application dev, I think it would make sense to use the BIOS to decouple HW + OS a bit. Or at least query the BIOS to ask the HW capabilities?
Regardless of the theoretical underpinnings, my M2A-VM flat out refused to boot under Ubuntu 7.10 until I had flashed its BIOS to a more recent version than what came in the box. After flashing, the same Ubuntu kernel managed to boot up. Since then, and even after Ubuntu 8.10 , things have been mostly OK, except for intermittent refusals to load USB sticks. Which, of course, some in the Ubuntu community blame on... the BIOS. Who's right???
But, seriously, the Ubuntu team is doing an _amazing_ job and is very, very, much the exact opposite of "crappy-OS-Alex" and his fanboy chums. I think they are greatly improving Linux, mostly because of the newfound example they are giving of respecting end users who aren't making a living as sysadmins.
- Geek's Guide to Britain INSIDE GCHQ: Welcome to Cheltenham's cottage industry
- 'Catastrophic failure' of 3D-printed gun in Oz Police test
- Game Theory Is the next-gen console war already One?
- BBC suspends CTO after it wastes £100m on doomed IT system
- Peak Facebook: British users lose their Liking for Zuck's ad empire