Fixing and improving drivers is good (usually). The number of opensource AMD driver improvements have been huge lately. They are significant just for the number of changes that have gone in. And they are significant because AMD is taking the linux platform seriously and putting resources into making the experience better.

There is a another benefit that comes out of this. The opensource driver getting closer to the propietary driver means that increased visibility and usage will provide AMD with more feedback and will increase the speed of the iterative process of improving the drivers.

I hope the perfomance will be on par with the Windows drivers soon

Think it will take 1-2 years still until that is acheaved

Originally posted by Kendji:

Think it will take 1-2 years still until that is acheaved

Please share what you know. Why 1-2 years? I haven't heard that timeline.

Originally posted by *****cubed:

Originally posted by Kendji:

Think it will take 1-2 years still until that is acheaved

Please share what you know. Why 1-2 years? I haven't heard that timeline.

know != think

Look them up in a dictionary.

Originally posted by Dusk of Oolacile:

know != think

Look them up in a dictionary.

Thank you for the semantic correction. Too bad you have nothing to offer in the real conversation here. Are you afraid to look (even more) foolish by getting involved in discussion?

What do you think about the 1-2 year timeline?

yea Dusk of Oolacile is clueless... thinking and making assumptions requires some knowledge, thus its justified to ask of what he knows.

Anyway its just AMDGPU kernel drm driver for better support of some r9 and future GPUs, too early to expect something more from it right now imo. So far its good for opensource drive developement.

It's my educated guess based on everything I've read, my IT knowlage and the state of the current drivers. The work is going along, but not fast enough for me to think it is coming within a year, but if progress keeps going at this pace you might have nVidia standard drivers within 2 years, imo.

My short motivation:
-July 2015: The AMD drivers not up-to standard. http://www.phoronix.com/scan.php?page=article&item=amd-tonga-linux42&num=1
-March 2015: Promise of new AMD drivers. http://www.phoronix.com/scan.php?page=news_item&px=amd-catalyst-linux-march-2015

-AMD made it's drivers open-source in a hope that they would need to only put in minimal effort into developing them. Though there haven't been any big open-source community development, drivers still not good.

-AMD drivers have been awful atleast since HD 5850 (which is my earliest AMD GPU)

-Thus I conclude that AMD need to make the effort to develop their own drivers like nVidia is doing. Considering the economical situation AMD currently is in, which is bad, and considering that AMD will start feeling the economical upturn in 2017 after their Zen processors etc. have been a while in the market already. Assuming their product launches in 2016 will be successful.

-In adition to this there is a very low Linux adoption rate among PC enthusiasts, enthusiasts are significant because they create hype, recomendations etc. and influenses other potential customers purchase decicions. In other words, there would not currently be much economical gain from developing drivers for Linux. SteamOS etc, could change this.

-Conclusion: I have a hard time believing that we will have drivers capable of competing with nVidia's drivers withing the next 12 months and within ~24months (by august 2017) are an optimistic estimate. This time table is dependent on the 2016 launches being succesful by AMD, compared to the current state of the drivers and the work needed to fix them.

I am running a R7 360 right now on steamos at was playing Metro: 2033 Redux and performance seems good to me. But I have no comparison personally since this rig was built just now and for steamos. I don't have a windows pc to compare to.

My point is that performance with the latest AMD drivers (which I am running) is fine right now.

You have to be careful treating phoronix as completely balanced. Plus he's talking about the tonga which just released and the driver's not mature yet because a lot of devs didn't have one yet. It's completely new technology.

We don't agree and time will have to tell the story.

I am happy with this new $100 R7 360's performance.

Originally posted by Kendji:

-AMD made it's drivers open-source in a hope that they would need to only put in minimal effort into developing them. Though there haven't been any big open-source community development, drivers still not good.

They never made their driver open source, everything that has been commited directly by AMD has been written from scratch - everything else is pretty much impossible due to legal reasons.
Don't confuse the driver itself and the OpenGL library working on top of it, either - there's a lot of code inbetween an OpenGL call and its required hardware commands, large parts can be split up into independent components - which is exactly what the free drivers have been doing.

The transition over to using the amdgpu kernel module is not something you can realize on a single weekend, it's a ♥♥♥♥♥♥♥♥ of work. Gallium uses an Intermediate Representation language similar to Vulkan's SPIR-V for a somewhat hardware-independent direct access, pretty much the whole core of fglrx's OpenGL library will need to be rewritten to make use of it. Once it's ready, you'll face a Catalyst that will have a pretty much newborn OpenGL library without all the old hacks that were once the reason for its bad reputation.

The whole decision wasn't made to save money. It was actually something they wanted to do from the very beginning (after the acquisition), but both fglrx and the free driver simply weren't ready, yet. Don't forget that most of radeon's core developers already have been on AMD's payroll for years, this step will actually bring more of Catalys's core developers over to work on the open source parts. Even the published hardware documentation cost them a lot of money, everything needs a very time consuming legal clearance process.

In the end any addition made to the amdgpu module in order to offer new features with fglrx actually have to be usable with the free driver as well, else they would get rejected - Linus doesn't want code that only exists to serve proprietary software in his kernel.

Originally posted by Letalis Sonus:

Originally posted by Kendji:

-AMD made it's drivers open-source in a hope that they would need to only put in minimal effort into developing them. Though there haven't been any big open-source community development, drivers still not good.

They never made their driver open source, everything that has been commited directly by AMD has been written from scratch - everything else is pretty much impossible due to legal reasons.
Don't confuse the driver itself and the OpenGL library working on top of it, either - there's a lot of code inbetween an OpenGL call and its required hardware commands, large parts can be split up into independent components - which is exactly what the free drivers have been doing.

The transition over to using the amdgpu kernel module is not something you can realize on a single weekend, it's a ♥♥♥♥♥♥♥♥ of work. Gallium uses an Intermediate Representation language similar to Vulkan's SPIR-V for a somewhat hardware-independent direct access, pretty much the whole core of fglrx's OpenGL library will need to be rewritten to make use of it. Once it's ready, you'll face a Catalyst that will have a pretty much newborn OpenGL library without all the old hacks that were once the reason for its bad reputation.

The whole decision wasn't made to save money. It was actually something they wanted to do from the very beginning (after the acquisition), but both fglrx and the free driver simply weren't ready, yet. Don't forget that most of radeon's core developers already have been on AMD's payroll for years, this step will actually bring more of Catalys's core developers over to work on the open source parts. Even the published hardware documentation cost them a lot of money, everything needs a very time consuming legal clearance process.

In the end any addition made to the amdgpu module in order to offer new features with fglrx actually have to be usable with the free driver as well, else they would get rejected - Linus doesn't want code that only exists to serve proprietary software in his kernel.

Yes true, my fact was wrong :). But from what your telling me the end-user won't still get as good of drivers as from nvidia within a years timeframe. I am planning to buy an new graphics card at the end of summer and the only viable alternatives right now is nVidia. Really hoping AMD will get their drivers sorted out.

Originally posted by Letalis Sonus:

Linus doesn't want code that only exists to serve proprietary software in his kernel.

Great post. Linus has sort of lost the battle though to a degree. However the tide is definitely changing.

The notion that code that can be reviewed by 1,000s and is good for the code and everyone involved as opposed to 10 or 20 reviewing it is catching on in certain circles where they focus on value.

This AMD move is smart and it will force competitors to do the same. Intel must be paying attention here. Not sure about nvidia because I don't have any nvidia hardware anymore.