NVIDIA 381 Linux Beta vs. Linux 4.11 / Mesa 17.1 Radeon Comparison

[creative]

anything 4.4 does it no problem

[creative]

I searched and searched and searched and for some reason I cannot find a reason in ubuntu studio 16.04 that 381.09 with kernel 4.10 will not load the kernel module after building it against it.

[aufkrawall]

Metro LL runs decent on Linux, yes. Unfortunately it doesn't use anisotropic filtering right and unlike on Windows, you can't force it via drivers (at least it doesn't work with the Nvidia driver).
I just bought the Redux versions some time ago because there was a sale and Linux support, but I still prefer to play it on Windows without blurry textures.

[torsionbar28]

It's a pity the Radeon R9 Fury isn't more competitive in Metro Last Light Redux.

Its hard to take commentary like this seriously, when the graph shows the Radeon R9 Fury delivering 162 frames per second. Nobody looks at 162 fps and is disappointed. The GTX 1080 pulled 189 fps in the same benchmark. Can your eyes tell the difference between 162 fps and 189 fps? Mine can't. And is the $150+ price difference between an R9 Fury and a GTX 1080 worth a 15% increase in already-triple-digit frame rates? Not so much.

[aufkrawall]

Super smooth here, only a couple micro-stuttering at the beginning of the level. Almost all video options to the max, triple buffering enabled (GTX 1060, i5, nvidia 378.13). I don’t play this game so the shaders weren’t already in the nvidia cache.

I don't know then. It's definitely not an issue with my installation or CPU, for example Shadow Warrior runs fine with high fps.
No difference between Tumbleweed, Leap and Arch here.

[marek]

Yes, radeonsi still has some headroom and could be improved.

I'd like to see performance per dollar. That can be pretty interesting.

[stqn]

Left 4 Dead 2.

Super smooth here, only a couple micro-stuttering at the beginning of the level. Almost all video options to the max, triple buffering enabled (GTX 1060, i5, nvidia 378.13). I don’t play this game so the shaders weren’t already in the nvidia cache.

[cj.wijtmans]

In some cases the RX 480 already performs as it is supposed to: Between 970 and 980 but under 1060 - at least for OpenGL that's a valid result. With well-optimized AMDGPU + Mesa drivers as well as well-designes vulkan games it should beat GTX 1060.
In some szenarios it's even coming short of the GTX 960 which is a shame - there's still some room for improvement even though the open drivers really had come a long way from being unplayable on modern games to performing quite well on triple-A (AAA) titles.

Thanks for explaing that with triple-A you meant AAA, i would never have figured it out.

[cRaZy-bisCuiT]

In some cases the RX 480 already performs as it is supposed to: Between 970 and 980 but under 1060 - at least for OpenGL that's a valid result. With well-optimized AMDGPU + Mesa drivers as well as well-designes vulkan games it should beat GTX 1060.
In some szenarios it's even coming short of the GTX 960 which is a shame - there's still some room for improvement even though the open drivers really had come a long way from being unplayable on modern games to performing quite well on triple-A (AAA) titles.

[Nille_kungen]

The R9 Fury has more in common with the 1080 as its a 8.5tflop card, unfortunately drivers and software optimizations are limited to just Tomb Raider it seems. Shame but thats life.

No it's not, the R9 Fury is an 7.2Tflop card.
It's a difference between R9 Fury and R9 Fury X.