Slashdot Mirror
AMD Publishes New 'AMDGPU' Linux Graphics Driver
An anonymous reader writes: AMD has made available its new AMDGPU Linux graphics driver comprised of a brand new DRM/KMS kernel driver, a new xf86-video-amdgpu X11 driver, and modifications to libdrm and Gallium3D. This new AMDGPU driver is designed for supporting AMD's next-generation hardware with no support differences for currently supported Radeon GPUs. While yet to be released, this new AMDGPU driver is the critical piece to the new unified driver strategy with Catalyst where their high performance proprietary driver will now become limited to being a user-space binary component that uses this open-source kernel driver.
No good vaapi integration into mplayer like that of vdpau. Not to mention that the AMD linux drivers are still manifestly slower than the nvidia drivers.
"I don't know, therefore Aliens" Wafflebox1
Yep, that behaviour's sure gonna get you laid.
The last desktop I shopped for -- I'd liked it to be an AMD. But after this wobbly policy of AMD (do I need now binary blobs for AMD graphics or not?) I decided (I don't need gaming performance) that Intel with its integrated graphics seems the best bet at the moment.
AMD: you can stuff your "high performance proprietary driver" up any cavity of your choosing. I'll buy things from you again when you have a clear pro-free software strategy again -- if you're around by then at all.
Good riddance.
(captcha was "sadden", how appropriate)
After a decade+ of watching management at ATI/AMD screw the Linux users, with shite drivers (See numerous posts across www & various in particular @ http://www.phoronix.com/, with Astroturfing to boot!), I will refrain from holding my breath.
Pity, cause the hardware tends to show promise, again, and again, and again....
After those dumbasses @ AMD actually put some $$$$ into it, then I will look twice. Until then, the hardware will continue to underperform, video playback will be rubbish and the forums will be full of pain & paid for Astroturfing.
nVida maybe protect there IP, but at least there stuff works. Say, even optimus stuff works. I just CBF installing Linux on my lappie.
GreekGeek :-)
And I can hear the assholes rejoicing. Your point?
Everyone complains about AMD and NVIDIA drivers always being released for hippy Linux kernel gurus who run super fast advanced linux kernel based software. HELL THIS DRIVER SOFTWARE IS FOR LINUX KERNEL AND HARDWARE WHICH have not even been released yet! In the mean time us poor suckers that run the latest and greatest from Windows are stuck with drivers that at best might work only with directX drivers for the same hardware that is older than the hills. I guess you linux guys are the real lab rats when it comes to AMD devices and software!
This message was not sent from an iPhone because Peter Sellers really was a deviated prevert without a dime for the call
You should shave and bathe more often.
. . . well - perhaps 3D is better on some other card. But who even cares about 3D, if the 2D performance isn't good enough?
And this is where all fancy graphichs cards fall on their face. They can't match the 2D performance of the "simpler" Intel stuff. In theory they can, but not with existing open-source drivers. 2D is what you use for office work, for scrolling large windows. Or try playing openttd - a great 2D game. Nice and snappy with 4 cores and Intel graphics, sluggish with 8 cores & a new radeon. Older radeon is not so bad, although it can't beat Intel.
New radeon is worse, they use "glamor" for 2D, and openttd then gets too slow. As do anything else that do full-screen repainting in a loop.
Last time I bought a laptop, you couldn't get a full HD display without being burdened with radeon or nvidia. If I bought a laptop today, it'd be Intel graphichs now that it is available in high resolution. My work machine is like that, and it is actually better for 2D gaming.
And forget nvidia. I had this nvidia laptop where 2D sluggishness were bad enough to affect even word processing. It would run simple 3D stuff ok, if you bothered to set up the drivers just right. A horrible office machine. You would think they'd get to useable 2D performance in 8 years, but no. Never happened with that machine, neither open nor closed source. The display was kind of ok in 2007, but x.org "progressed" with EXA or whatever. It was impossible to stay with very old drivers. Newer drivers probably helped newer chips, but killed this one. At times I used plain framebuffer drivers - for performance :-( Nvidia didn't care about old machines. I cannot blame open source for not supporting everything, but it was sad they had to drop what worked.
Still, there are opensource implementation of VDPAU using VAAPI as a back-end.
and there are VAAPI implementation using VDPAU as a backend (useful also for opensource drivers which tend to implement VDPAU).
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
I know I'm ranting, but graphics cards are without doubt the most annoying piece of gear I own. I have used ATI cards so far but they always have problems, be it overheating at normal clock speed (my current R9 280), driver problems, crashes, flickering in OpenGL. So I'm thinking about switching to NVIDIA for my next machine, only to hear from reliable pro-audio people that these cards often cause problems with the audio drivers of professional external sound cards like horrible crackling noises.
Do I really need to buy two expensive PCs in order to work on GNU/Linux, play a flight simulator and record music?
Perhaps you're an AC trolling, but in case someone is actually interested in a reply, it doesn't seem that you're giving AMD nearly enough credit.
Now, I grant that Intel has a great policy when it comes to providing driver components, being open sourced For those who aren't going to use their GPU for anything more intensive than watching media, an Intel integrated GPU is probably sufficient. However, given that Intel's GPUs are orders of magnitude weaker than discrete modern GPUs, those who wish to use graphically intensive creative software, use GPU-preferred workflows for certain tasks (ie crypto-currency mining, video encoding, anything with OpenCL etc..), or just plain wish to play modern 3D games, Intel's hardware is likely insufficient.
Thus, users have the choice between AMD and Nvidia, and of the two AMD is MUCH more free software / open source friendly, as well as overall more ethically aligned than its competitor. To answer your question, especially if you're not gaming, you do not need a binary blob to have a working AMD GPU on Linux. The open source "ati" driver has been honed over the last few years to the point that for many it provides not only great 2D performance and support for video acceleration, but a bit of 3D as well. Yes, they offer the binary "Catalyst" driver as an option (which supports pretty much all the high end 3D acceleration that someone purchases a discrete GPU to use) so if you're going to be gaming and doing intense 3D work it is a better choice, but - aforementioned open source driver is always available.
AMD has actually gone out of their way to embrace open source and the Linux community. . Much of the bias against AMD is left over from ATI's disastrous drivers (even proprietary) back in the day. They knew they had a lot of ground to make up in Linux performance (not necessarily openess...) when they acquired ATI, and its pretty impressive how they've managed to catch up to Nvidia. Subsequently, they put lots of emphasis on contributing to the FOSS driver projects. They aren't keeping Catalyst proprietary so far simply because they're evil, but because of patents and other tech licensed to third parties that go back years. In their overall business plan, they clearly favor openness compared to Nvidia. Where NV creates proprietary implementations and engines like their G-Sync monitors and the frustrating presence of PhysX in gaming, AMD champions open standards and projects such as FreeSync, OpenCL, and Vulkan.
In fact, the release of AMDGPU and in fact their entire unified driver plan leads to more openness. As the article notes, this will mean that the vast majority of AMD's GPU driver components will be completely free software.! Where Catalyst used to be a completely different animal in every way, now it will only be just an additional binary component that can be added. This is one reason why it is being written with the next generation of hardware in mind (though it may be partially compatible with some current cards); they need to have some entirely new hardware etc....that doesn't require certain patented/licensed tech. This is a great advance for openness as well as driver quality. While it would be great if they were able to instantly fully open the entire Catalyst driver stack, one would be remiss to not see the current path as progress.
While I can only speak from personal experience (and I've owned both Nvidia and AMD GPUs. At current, I'm using an AMD 290x), running an AMD GPU under Linux is viable. Sure, there are issues that can arise same as dealing with any other non-kernel-included Linux driver. If you want to stay entirely Free Software, there is a FLOSS driver that will likely work well unless you plan to use a lot of 3D or OpenCL etc. If you want 3D acceleration and you're willing to go binary, Catalyst is an option. In the future, the openness, compatibility, performance, and ease of use will likely increase with AMD's unified driver plan we're reading about here. AMD is far from perfect, but they do seem to be interested in making FLOSS a part of their future, along with a whole slew of open technologies.
Except that the new driver is open source. See the official announcement for more info.
(do I need now binary blobs for AMD graphics or not?)
The whole point of AMDGPU is to simplify the situation.
Now the only difference between catalyst and radeon drivers is the 3d acceleration - either run a proprietary binary opengl, or run mesa Gallium3D.
All the rest of the stack downward from this point is opensource: same kernel module, same library, etc.
Switching between prorietary and opensource driver will be just choosing which opengl implementation to run.
I decided (I don't need gaming performance) that Intel with its integrated graphics seems the best bet at the moment.
If you don't need performance, radeon works pretty well too.
Radeon have an opensource driver. It works best for a little bit older cards. Usually the latest gen cards lag a bit (driver is released after a delay, performance isn't as good as binary) (though AMD is working to reduce the delay).
Like Intel, the opensource driver is also supported by AMD (they have opensource developpers on their payroll for that), although compared to Intel, AMD's opensource driver team is a bit understaffed.
AMD's official policy is also to only support the latest few cards generation in their proprietary drivers. For older cards, the opensource *are* the official drivers.
(Usually by the time support is dropped out of catalyst, the opensource driver has caught up enough with performance to be a really good alternative).
The direction toward which AMD is moving with AMDGPU is even more reinforcing this approach:
- the stack is completely opensource at the bottom
- for older cards, stick with Gallium3D/mesa
- for newer cards, you can swap out the top opengl part with catalyst, and keep the rest of the stack the same.
- for cards in between it's up to you to make your choice between opensource or high performance.
If you look overall, the general tendency is toward more opensource at AMD.
- stack has moved toward having more opensource components, even if you choose catalyst.
- behind the scene AMD is doing efforts to make future cards more opensource friendly and be able to release faster the necessary code and documentation.
AMD: you can stuff your "high performance proprietary driver" up any cavity of your choosing. I'll buy things from you again when you have a clear pro-free software strategy again -- if you're around by then at all.
I don't know what you don't find clear, in their strategy.
They've always officially support opensource: they release documentation, code, and have a few developpers on their pay roll.
Open-source has always been the official solution for older cards.
Catalyst has always been the solution for latest cards which don't have opensource drivers yet, or if you want to max out performance or latest opengl 4.x
And if anything, they're moving more toward opensource: merging the to to rely more on opensource base component, to avoid duplication of development efforts,
and finding ways to be faster with opensource on newer generations.
For me that's good enough, that why I usually go with radeon when I have the choice (desktop PC that I build myself) , and I'm happy with the results.
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
Actually the footprint of binary is dwindling.
Before:
- either catalyst, which is a completely closed source down to the kernel module.
- or opensource, which is an entirely different stack, even the kernel module is different.
Now:
- open source stack is still here the same way as before.
- catalyst is just the opengl library which sits atop the same opensource stack as the opensource.
So no, actually I'm rejoincing. (That might also be because I don't style my facial hair as "neck bread" ).
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
An open driver allows a game producer to produce a game that works better on your graphics card.
Then the driver code is written by a larger talent pool, selecting for the best talent for the job.
Game engines can see what the cards do and see where their engine spends all their time and optimise the use of the graphics card hardware which they are unable to do with closed drivers.
You have simplified code because there's no shim + actual driver, only actual driver.
I usually try to call the graphics side ATI so that when say something like "fuck ..." it says ATI and not AMD, because I rather like AMD's processors. I'm still rocking a Phenom II X6 1045T and I'm very happy with its performance, considering I paid $120 for it quite a while back and have no plans to upgrade it anytime soon. Maybe another video card next year... I went from 450GT to 750Ti on the advice of Slashdot, and haven't regretted it.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Anyone got a moronix-less link to the story?
and it's like running windoze... bloated, slow, crap. The open source works better.
Serious question... I have a desktop w/HD5450 and three monitors on a desktop PC running RHEL 6.5. How do you build & install this driver without completely breaking RHEL's package management for everything else in the process, breaking the kernel configuration, or anything else that might render the computer nonworking?
I *tried* installing ATI's binary Catalyst driver a few months ago. Unfortunately, I think it made some naive assumptions about the underlying filesystem that aren't quite right when you have a spinning HD and a SSD, and both use LUKS for whole-drive encryption. I desperately want to upgrade the video driver, because the performance totally sucks (even for things like dragging a browser window to another monitor), but I don't want to end up burning another day undoing a failed upgrade adventure.
TL/DR: Want to build & install ATI's new driver. Running Red Hat Enterprise Linux 6.5. Not allowed to upgrade to 7, and a prior attempt to directly install ATI's Catalyst binary driver rendered the system unbootable [most likely, because of issues with LUKS whole-disc partition encryption]. Yes, I'm root.
I'm no native, but "comprised of" is ugly English to me.
I'm so tired of hearing news story after news story about how they're improving their drivers. You never hear this from Nvidia because their shit _just works_
So does this mean we'll be able to get kernel mode setting with the proprietary Catalyst driver? Because that would be real sweet..:)