AMD's Open-Source Radeon Linux Driver: 2014 Was Incredible

[phoronix]

Phoronix: AMD's Open-Source Radeon Linux Driver: 2014 Was Incredible

With this week having delivered our 2014 Catalyst Linux Graphics Benchmarks Year-In-Review that looked at the evolution of the AMD Catalyst Linux driver performance over this year, now it's time to see how the open-source Radeon driver performance has evolved.

http://www.phoronix.com/vr.php?view=21274

[dungeon]

Cheers

Hopefully, we will see less improvments in radeon next year (it is imposibile to see more performance improments like in this year, of course still some smaller percentage gain is posibile) and more talks going on about amdgpu driver

[looserouting]

for me 3.17 is faster then 3.18 with my radeon hd7770

[duby229]

I am still using a 6850, and wow!

My experience has been really subjective. Basically I'm just not sure how to quantify it.. The display just feels smother. It seems like there is a lot less "jerkyness" The OSS driver has more or less always had great 2d responsiveness, but although 3d framerates haven't changed very much, 3d responsiveness "feels" better.

EDIT: I think it's probably likely that minimum framerates while in 3d are higher.

EDIT: I'm going to say it again... benching at hundreds of frames per second is totally pointless... What is the point in benchmarking settings that nobody is ever going to use?

[dungeon]

for me 3.17 ist faster then 3.18 with my radeon hd7770

Yeah a little bit more fps but with more stutter, 3.19 is also with no much more fps but even less stutter then 3.18, etc... Very high fps rate is not always good, when you consider render quality too

Hopefully si-scheduler will came in llvm 3.6, performance goes up in demanding cases with that... and a little bit more stutter - all is good if it does not introduce awfull stutter

[bridgman]

Hopefully, we will see less improvments in radeon next year (it is imposibile to see more performance improments like in this year, of course still some smaller percentage gain is posibile) and more talks going on about amdgpu driver

Remember that the article talks about the "radeon stack", not just the radeon kernel driver. Many (most ?) of the performance improvements are in the userspace bits, not the kernel bits, and those will be shared across radeon and amdgpu kernel drivers.

[dungeon]

Remember that the article talks about the "radeon stack", not just the radeon kernel driver. Many (most ?) of the performance improvements are in the userspace bits, not the kernel bits, and those will be shared across radeon and amdgpu kernel drivers.

Yeah i know radeon is kernel driver, xf86-video-ati (called radeon) is ddx, mesa drivers tested here are r600/radeonsi (no one call those radeon, because there is still original driver called radeon in classic mesa )... amdgpu is kernel driver, which will use same "radeon stack" called opensource and/or closed source UMD.

So same userspace "radeon stack" for amdgpu we will call "amdgpu stack"... blah, blah, i understand J.B. .

[89c51]

Things like HW accelerated video STILL need some attention.

[dungeon]

Things like HW accelerated video STILL need some attention.

Yeah, currently it is month broken by intel's guy https://bugs.freedesktop.org/show_bug.cgi?id=86837

[Nille]

Yeah, currently it is month broken by intel's guy https://bugs.freedesktop.org/show_bug.cgi?id=86837

Thats not the bad thing. The problem is thats this commit is not reverted :/