AMD Radeon Graphics Performance On The Linux 3.18 Kernel

[phoronix]

Phoronix: AMD Radeon Graphics Performance On The Linux 3.18 Kernel

If you're curious whether the Linux 3.18 kernel will bring any performance improvements for users of the open-source AMD Radeon graphics driver, here's some benchmarks compared to Linux 3.17.

http://www.phoronix.com/vr.php?view=21135

[Ardje]

Nothing worse than a "It didn't change" review.
But good to do it anyway, thanks Michael!

[user82]

The R9 290 continues to be disappointing compared to the mid end AMD cards. Too bad, at least I hope it is more stable.

[dungeon]

Other then oibaf ppa should consider to use llvm3.6 git to better show radeonsi improvments in latest 2+1 that is actually ~3 months now... i am mostly satisfied

Read it as: we don't see what happened with llvm in last 3 months here

[darkbasic]

Other then oibaf ppa should consider to use llvm3.6 git to better show radeonsi improvments in latest 2+1 that is actually ~3 months now... i am mostly satisfied

Read it as: we don't see what happened with llvm in last 3 months here

+1 (stupid character limit)

[fhuberts]

no surprises here

Ghee what a surprise.
The kernel actually doesn't do very much.
Performance is mainly determined by mesa.

ANother click-baiting article

[Michael]

Ghee what a surprise.
The kernel actually doesn't do very much.
Performance is mainly determined by mesa.

ANother click-baiting article

There's still a lot of churn in the DRM drivers.... e.g. just from Linux 3.17 - http://www.phoronix.com/scan.php?pag...linux317&num=1

It's not click bait, it lets people know if/what's changed wrt to graphics performance with Linux 3.18.

[smitty3268]

There's still a lot of churn in the DRM drivers.... e.g. just from Linux 3.17 - http://www.phoronix.com/scan.php?pag...linux317&num=1

It's not click bait, it lets people know if/what's changed wrt to graphics performance with Linux 3.18.

Yeah, I've got no problems with this article. There have been a bunch of performance improvements for radeon SI in the past few kernels.

Wasn't there an issue with the Hawaii cards where they would randomly lose their reclocking ability and get very slow? I was wondering if this kernel fixed that, but it doesn't show up in the 3.17 results here either.

[bridgman]

The R9 290 continues to be disappointing compared to the mid end AMD cards. Too bad, at least I hope it is more stable.

It's ~40% faster on GPU-limited benchmarks, which roughly matches the difference in shader hardware, so that seems about right.

The slowest card tested is ~HD4870 class and average is more like ~2x that, so it's not unexpected that the less demanding apps are going to be CPU-limited.

[schmidtbag]

The R9 290 continues to be disappointing compared to the mid end AMD cards. Too bad, at least I hope it is more stable.

Yeah, I just bought a R9 290 yesterday and am waiting for it to arrive. Even though the results for it are disappointing, it should still handle everything I want it to and it will only improve. The 290 and 290X are a different architecture than the rest of the GPUs, and the 285 is different from everything else too. So I figure it'll take a while for those to improve. The 290 itself has dramatically improved over the past year, and even though it is far behind what it should be, it's still got some leverage over many other GPUs.