ATI R800 linux

[Panix]

Loud voice is different from whining.

In the Linux community, the loudest voices are found in git and svn commits. The drivers are open, anyone can help. That's how Linux works.

If you can't help, you can sponsor a developer, file bug reports, bisecting, writing test cases, testing apps for compatibility, help with the wikis, help running the irc channels, helping other people with installation and bug reporting, package drivers for your favourite distribution, etc.

If you insist on seeing Linux as a product, you will always be disappointed, because it simply is not, and will never be, a product. And that's why an open source driver changes EVERYTHING. It makes ATi hardware a first-class citizen in the Linux ecosystem. The performance is not great yet, but it works for most everyday tasks. The rest will have to evolve, like the Linux kernel evolved, like OpenOffice evolved, like Mozilla evolved. with time, it will happen.

So, you want *me* (could be anyone) to help but AMD/ATI won't spare some change towards the projects? What's wrong with this picture????

[pingufunkybeat]

But AMD is funding a huge number of developers (only 3 working on the OSS drivers, the rest on catalyst), and spending a lot of money on releasing documentation, etc.

Sure, I'd like them to drop a billion $ and hire 100 developers full-time to work on the OSS drivers, but this is unlikely. A couple of more developers would be great and doable, though, but it's hard to find people as capable as the ones they are currently employing. Bridgman, if you're reading this, hire a couple of developers now

I don't know why they're struggling to provide a stable binary driver, and it's a good question, but I guess that desktops are simply not their priority (workstations are).

At the same time, there is not reason why you (or anyone) could help. Marek is doing a great job, and MostAwesomeDude did lots of great work in the past. That's the best way to improve the (open) drivers.

[Kano]

The OpenGL devision must be really small, because 10-7 had not working tesselation with Heaven and with 10-8 Heaven DX11 tesselation (normal) is 25% faster than Win OpenGL. As the Linux OpenGL variant is faster the Win group should fix some OpenGL speed issues too. Nv cards are not much different in speed when you use OpenGL on both plattforms (maybe in case of sli).

[Kano]

I told you even ATI Win OpenGL is slower than Linux OpenGL. For a shared source driver this is really weird. I can understand that DX11 is faster but no the huge difference between Win+Linux and OpenGL 4.

[bridgman]

OK, so let me get this straight. If OpenGL runs faster on Windows than on Linux we get blamed for ignoring Linux, right ? If OpenGL runs faster on Linux than on Windows then we're *still* doing something wrong ? Hello ? Would you be happier if we just ignored most of the Linux graphics environment and ran Windows code everywhere so you could get *exactly* the same frame rate on both OSes ?

The Linux graphics environment is *very* different from Windows. There are going to be differences in performance between the environments - some things will be faster, others slower. We can bring them closer by wiping out part of the Linux environment but I'm not convinced that is a Good Thing for the market.

[Kano]

You usually would expect equal performance (well +/- 5%) using OpenGL. Buy a Nv card to see what i mean - single card config. When you get something different you must have a problem in your code. Or do you want to tell ppl that AMD is not even capable of doing correct Win drivers when Win rules the market share? Too funny. It is certainly NOT the case that AMD binary drivers are too optimizied for Linux as they always fail the KDE 4.4.5 function test and all those video/kernel/xorg issues...

[Loris]

You usually would expect equal performance (well +/- 5%) using OpenGL. Buy a Nv card to see what i mean - single card config. When you get something different you must have a problem in your code. Or do you want to tell ppl that AMD is not even capable of doing correct Win drivers when Win rules the market share? Too funny. It is certainly NOT the case that AMD binary drivers are too optimizied for Linux as they always fail the KDE 4.4.5 function test and all those video/kernel/xorg issues...

Kano, isn't the ATI proprietary driver reusing more components in the Linux's graphics stack than the Nvidia driver? Could it be that, since more of the code path is preserved between Nvidia's Windows and Linux drivers, the resulting performance is more likely to be similar?

I always thought of the Nvidia driver as a parallel infrastructure, more of a black box than other approaches.

Say, could it be?

[rohcQaH]

Nah, that explanation is too obvious to be true. It's because ATI sucks.

[Kano]

NV replaces more parts of the X stack, therefore the binary is much smaller. AMD ships for each X server that has a different abi new files, that means you are over 100 mb for the driver package where NV is still not even at 40 % of the size (32 bit only is just 25 mb). But i don't think that has impact on OpenGL performance.

[Sjonny]

Why is everybody so obsessed with OpenGL speeds, when even simple 2D stuff is unbearably slow. I can't work normally in Linux. I have a 5870, with 10.7 installed. If I enable the new 2D code, I have black or gray planes allover (mostly when using firefox or thunderbird). And switching desktops is really slow, as it needs to redraw the whole screen. (I simply have metacity as wm, nothing fancy).

Is there a roadmap of the driver development somewhere? I'd like to know if this is hopeless, or if this will be fixes withing the next 2 or 3 driver releases.