Phoronix Lauds AMD's Open Source Radeon Driver Progress For 2014

Phoronix has taken an in-depth look at progress on AMD's open source Radeon driver, and declares 2014 to have been a good year. There are several pages with detailed benchmarks, but the upshot is overwhelmingly positive: Across the board there's huge performance improvements to find out of the open-source AMD Linux graphics driver when comparing the state at the end of 2013 to the current code at the end of this year. The performance improvements and new features presented (among them are OpenMAX / AMD video encode, UVD for older AMD GPUs, various new OpenGL extensions, continued work on OpenCL, power management improvements, and the start of open-source HSA) has been nothing short of incredible. Most of the new work benefits the Radeon HD 7000 series and newer (GCN) GPUs the most but these tests showed the Radeon HD 6000 series still improving too. ... Coming up before the end of the year will be a fresh comparison of these open-source Radeon driver results compared to the newest proprietary AMD Catalyst Linux graphics driver.

Comments

New drivers much better

[future assassin]

I run an A10-5800 and an A-10 7700K and the newest drivers since about 3 months ago give me way better performance now. I normally play Xonotic and with the new drivers with my 5800/DDR3 1833 I can max out the effects settings and still get 60-90 FPS on properly designated maps. If I lower the settings to Normal I get anywhere from 90-140 FPS. Before that I was having issues with Ultra settings and lots of weird movement lag.

Re:New drivers much better

[future assassin]

Eh I'm referring to the AMD propitiatory drivers. Haven't tested the Radeon ones yet, just waiting for Mint 17.1 to go into stable but there were not as good as the AMD ones when I tested them out about 5/6 months ago mostly with Xonotic.

Great news for OSS

[nicholas22]

Given NVIDIA's terrible reputation of open sourcing code (remember Linus' middle finger anyone?), I for one welcome our new GPU overlords...

Re:Great news for OSS

[Anonymous Coward]

And yet my A19-7700k Radeon SI drivers still crash every 2-3 months, and self-corrupt configuration files, and barely support OpenGL 3.3, and takes hours to setup/configure - sourcing weird builds of Mesa/LLVM from various PPAs that only support specific versions of Ubuntu, etc.

And yet my nVidia GTX 780 is running flawlessly, great performance, OpenGL 4.4 since day one, no stability issues, everything 'just works', and I can choose whatever distribution I want.

Re:Great news for OSS

[Zontar The Mindless]

The Nouveau drivers frequently lock up my desktop on 3 different machines with 3 different Nvidia cards, usually when I'm running VirtualBox. No thanks. Useless to me.

(I use OSS when it's viable. However, I also need to get work done.)

Re:Great news for OSS

[Anonymous Coward]

Then why not just use Windows? It gives you the best drivers, if open source is not important to you.

Perhaps there are at least some people who use Linux for real, practical reasons (however few of them there are), rather than purely ideological ones ? And those who really want a fully open system should ideally also use open hardware and firmware, in addition to 100% open source applications. Even the open source AMD drivers require proprietary firmware blobs, for example. How many FOSS advocates other than the FSF use obsolete hardware just to keep their system really open ?

Re:Great news for OSS

[drinkypoo]

I will welcome them when the drivers reach parity and they scrub the commercial driver, until then I've been burned too many times by ATI to even consider one of their GPUs. Not literally, either, just continual wrestling with shit drivers.

I'm told they've gotten better, but I've got no reason to change

Re:Great news for OSS

[msh104]

Well, there is a difference between eventual target and current reality.

As a sysadmin it is a joy to use linux on the desktop when maintaining linux servers.
NVIDIA is way beyond any other party in their linux support. ( equal in performance and features to their windows drivers )
It's simply the best you can get right now. Now many of our coders use linux as well. But i don't think any of them would consider running the open source drivers.
As much as i would love my drivers to be open source i get much better results using the binary drivers.

That being said, i do applaud major vendors building their open source drivers themselves.
It is a major requirement to getting decent video card drivers for the future.

Re:Great news for OSS

[aliquis]

Hardly GPU overlords.

I don't care if the driver is open-source.

Nvidia = better.

Re:Great news for OSS

[aliquis]

(I use OSS when it's viable. However, I also need to get work done.)

That's up to you I guess..

Not really relevant for the rest of us who don't pick the shitty alternative.

I've been playing valve games on Debian Testing

[complete loony]

I've been playing a fair amount of Dota 2 and TF2 on Debian testing, out of the box. The frame rate is good enough that I haven't noticed the difference between the proprietary driver, or booting into windows (ew).

Great

[Anonymous Coward]

Maybe if the open source drivers get good enough they can port this stuff to Windows and replace AMD's mess there as well.

Re:Great

[ustolemyname]

It would seem your baseless assumptions are wrong.

Sure, that was 3 years ago. But you seem to be ignoring the following facts:

• AMD's proprietary driver shares most of its code between Windows and Linux.
• NVIDIA's proprietary driver shares code between Windows, Linux, Solaris, OSX and assorted BSDs.
• The majority of a device driver has to do with the device itself, and all kernels are trying to get the device to do the same thing and so will expect similar hooks. It is annoying to support multiple targets, but certainly not difficult.

Re:Great

[TheRaven64]

No. The nVidia drivers share around 90% of their code between all platforms (Windows, Linux, FreeBSD, Solaris) and the open source ones all use the Gallium framework, which is designed for portability from the ground up.

Modern GPU drivers require a set of services from the kernel, mostly related to memory management. They need to be able to get access to the device's I/O range in the physical address map and they need the kernel to grant access to texture memory in both main memory and the device. That's about all that they need from the kernel.

At the top, they need a state tracker that manages 3D API state (which is fairly minimal on modern APIs, as they aim to be stateless for performance reasons) and that translates the shader programs into some intermediate representation.

The majority of the device-specific driver code lives between these two layers, which are usually handled by abstraction layers so that they can be plugged into different APIs. You use the same Gallium driver with an OpenGL 2, OpenGL 3, OpenVG or Direct3D state tracker.

Re:Great

[goarilla]

I don't get it neither. Especially now when their graphics department is basically keeping them afloat.
It's prudent to have more resources dedicated to gpu driver development.

the obligatory....

[Masked Coward]

Does this mean it's official? 2015 is The Year of the Linux Desktop?

Re:the obligatory....

[Kjella]

Does this mean it's official? 2015 is The Year of the Linux Desktop?

My crystal ball might be a little wonky, but I'm pretty sure that 2015 is not the year of the AMD desktop whether they run Linux or not. I suspect Q4 is going to be another bloody quarter for AMD, apart from the console sales they haven't had any killer CPUs/GPUs for the holidays. So if YotLD happens, I suspect their Linux drivers had very little to do with it.

Re:the obligatory....

[Zontar The Mindless]

No, 2015 caps the Decade of the Linux Desktop.

It's finally getting ot the point where I literally can't help people with their Windows machines, because I'm forgetting how Windows does things. At long last, thank $$__DEITY__.

Same source says AMD still not so good

[laing]

This article from only a week ago indicates that the Linux AMD drivers are still quite terrible. I think maybe this current article is an attempt at some damage control.

Re:Same source says AMD still not so good

[Anonymous Coward]

"For this article, the proprietary AMD Catalyst and NVIDIA Linux graphics drivers were used."

This article is talking about the OPEN SOURCE DRIVER progress.

Could you at least rtfm the articles you link to?

Re:Same source says AMD still not so good

[Mal-2]

Yeah, this seems a little like saying "Our cafeteria food has improved. It used to be completely inedible. Now it just sucks."

Performance isn't everything

[JohnboyHolmes]

My old radeon 3650 based laptop runs hot on the open source driver compared to old FGLRX drivers that support it. So, hot that at times the thermal protection shuts it down when running computational intensive processes for 10+ minutes. So performance isn't everything at times.

Phoronix

[Narishma]

There are several pages

Of course there are, this is Phoronix. I don't think they know how to make a single-page article.

Re:Phoronix

[stoatwblr]

Of course they do.

A multi-page article gets more ad impressions.