AMD GPU-PRO 16.20.3 Beta Linux Driver Released

[haagch]

Radeon already supports it of course, so no timeline there. For amdgpu, the timing is going to depend a bit on how much community developers are able to contribute. Right now we're trying to get a "basically working" all-open stack out there.

Is there a reason this is so low priority? Don't your customers with GCN 1 (.0) want Vulkan?

[alex79]

More like 90% closed source and 10% open source, if you count lines of code.

Not sure where you get "at it for ages" from, the hybrid stack is a relatively recent project with just a Vulkan preview and VI beta release so far. We talked about the overall roadmap a year or so ago at XDC but at the time the main development focus was getting the initial all-open stack ready.

Can you please recap the issues you are seeing ? AFAICS you are installing a beta driver that has not yet been QA'ed on your hardware, finding that it works pretty well except for tearing, but I wasn't able to figure out if you tried the setup changes that had been recommended by other posters (vein and atomsymbol among others).

Yes I did what vein asked and what atomsymbol wrote was already there so I'm not sure if I had to change something or what. Anyway, that didn't help tearing, desktop environment was all sluggish/choppy. Than there was an issue with youtube videos in Chrome, in Firefox was fine so I don't know what was the problem but it happened only with pro drivers.

[bridgman]

Is there a reason this is so low priority? Don't your customers with GCN 1 (.0) want Vulkan?

It's not low priority. It's just that with getting the pro driver into production, launching/integrating ROC and supporting a couple of new HW generations there are a lot of high priorities right now.

[bridgman]

I think you only think in this way because you think inside of the capitalistic way of thinking "business timeline requirements" and inside of this thinking maybe true but i think in the FlOSS world of software development it is not true.

No, I think this way because I'm looking at the components on each side of the ratio. On the closed source side we have the entire OpenGL driver (equivalent to most of Mesa not just the pipe/winsys drivers), the entire OpenCL driver (again not just HW layer), and the entire Vulkan driver (which is pretty big in our case).

Let's say we were able to shrink the closed-source components by 60%, ie reduce them to 40% of their current size. That would change the numbers from 90/10 to 80/20 closed/open, which I regard as not much of a change compared with alex79's 10/90 statement.

[bridgman]

In my point of view is would make a huge different? from 10 to 20% is like 100% more? on the OS side?

No change on the OS side, just closed source side getting a lot smaller.

also "alex79's 10/90 statement." is wrong but your "90/10" statement is also wrong but I think the 80/20 ratio is realistic.

Realistic in the sense of "what the ratio would be if the closed source code was substantially rewritten", not realistic in the sense of what we have today, right ?

[Dea1993]

i don't undertand if is possible install this driver on a notebook with carrizo apu a10 8700p (and r7 m360 dedicated).
carrizo radeon r6 is a GCN 1.2 GPU, but it isn't on the list of supported gpu...
someone tried to install this driver on carrizo?

[bridgman]

It's not "open source vs closed source" as much as "small / senior team developing without time constraints vs large team developing quickly".

[Dea1993]

i've installed AMDGPU-pro on a laptop with A10 8700p (carrizo).
the driver works

glxinfo | grep OpenGL
OpenGL vendor string: Advanced Micro Devices, Inc.
OpenGL renderer string: AMD Radeon R6 Graphics
OpenGL core profile version string: 4.5.13439 Core Profile Context
OpenGL core profile shading language version string: 4.50
OpenGL core profile context flags: (none)
OpenGL core profile profile mask: core profile
OpenGL core profile extensions:
OpenGL version string: 4.5.13439 Compatibility Profile Context
OpenGL shading language version string: 4.50
OpenGL context flags: (none)
OpenGL profile mask: compatibility profile
OpenGL extensions:
OpenGL ES profile version string: 4.5.13439 Compatibility Profile Context
OpenGL ES profile shading language version string: 4.50
OpenGL ES profile extensions:

i haven't tested performance yet.
but temerature are higher, glxgears has lower fps (~800/900 vs ~4000/5000 of open source amdgpu driver)

EDIT: i've tested CS:GO, fps are really lower than opensource amdgpu driver, i'll continue to use amdgpu driver and i'll wait a stable relase with official support of carrizo

[bridgman]

for example : http://www.heise.de/developer/meldun...r-2175954.html headline: Study on Software Quality: Open Source proposes proprietary

That study talks about quite small differences in defect density - IIRC just over .5 for open source in general, just over .7 for proprietary in general (anything under 1 considered good) and just over .6 for Linux (presumably kernel although it didn't say). Nothing about code density.

The point I'm trying to make is that while there are studies which compare results of open source vs closed source development they tend to assume that "openness" provided all the benefits while ignoring the fact that there are a lot of differences between typical open vs proprietary projects besides code visibility and/or review practices. A lot of the real differences boil down to team size and quality vs time tradeoffs, but I haven't seen that covered much in open vs proprietary studies.

[chithanh]

So, um... What's the best way of installing this on Gentoo? Should I wait for ebuilds/guides?

There is a vulkan overlay which is managed by a user, it contains ebuilds for amdgpu-pro:

https://github.com/Lucretia/vulkan-overlay