Radeon Linux 4.11 + Mesa 17.1-dev vs. NVIDIA 378.13 Graphics Performance

[gurv]

Really impressed by how far RadeonSI has gone, though I'm a bit surprised by the Heaven score as RadeonSI has always shown strong performance in it.
Gratz to the AMD/Valve team

But the most impressive performance is shown in Feral ports and I think these guys deserve a big shout-out.
They have shown incredible skill in their Linux ports, they've continued supporting them long after release and they contribute a lot to the community.

So thanks a lot guys! Keep up the good work and keep theses ports coming cause I'm gonna keep buying them
(for example Rise of the Tomb Raider or MGS V would be nice )

[plasma5]

Its impressive how good the open amd driver holds up against the nvidia propritary driver. I cannot wait to buy a little Vega card.

[pal666]

I'm a little disappointed with the results. I was expecting Team Red to place a little better with the shader cache enabled, which was the only reason for reading more than just the first page of this article.

shader cache does not increase fps

[pal666]

480 is faster than 1060 in dxmd. very nice

[smitty3268]

Wow. Take out the Vulkan results and the Mesa AMD results are very competitive with NVidia.

Still a few games that need some work, but the worst case wasn't that bad and about what fglrx's best case used to be. It's even beating NVidia on a few games (relative to expected hardware performance)!

[geearf]

shader cache does not increase fps

Well it can improve the minimal FPS, thus raising the average.

[LinX]

I'm thrilled to see AMD come back to form really. I know it still doesn't quite hold to the nvidia blob, but AMD is showing what can be achieved with the community. I will be glad to purchase a Ryzen + AMD GPU machine by years end. The future is looking bright.

[MrCooper]

shader cache does not increase fps

It can affect the PTS numbers for benchmarks which compile shaders during the measured period, because PTS runs each test at least three times (possibly more if the standard deviation between results is too high). In effect, if the benchmark reports different numbers for warm vs. cold shader cache, the average number reported by PTS is biased by at least 67% towards the warm shader cache numbers.

[leipero]

Mesa is going big steps for sure, I'm using old GPU, don't care much about (new) games, but very happy I can fulfill most of my use cases with OSS, and not having to bother with drivers (idstro people update it, makes sure everything works fine etc.). I'm not a programmer, but wouldn't be better (if possible) that Mesa people go into "framework" direction for drivers? Meaning that when new product comes out, it's just a matter of optimization for that particular product within the framework, or is it like that now? If possible, that would save a lot of time for people working on it.

[mibo]

Yes, I'm happy with my rx480.
But, for Linux gaming in general - there is a long way to go :-(
It seems to be fine if you happen to be an Ubuntu user...
For me gaming is not a priority, so I will not change from my openSUSE. But, there I can see the library mess that steam and some of the games produce. Also, using MESA (which also has dependencies on other libs) increases the trouble. (at least some of the library problems seem to be no issue with binary blob drivers).