Windows 10 Radeon Software vs. AMDGPU On Ubuntu Linux

[phoronix]

Phoronix: Windows 10 Radeon Software vs. AMDGPU On Ubuntu Linux

Last week I posted results of Windows 10 vs. Ubuntu 16.04 when looking at NVIDIA's OpenGL performance. As those results were quite interesting, the next installment of our Windows vs. Linux benchmarking are some numbers for AMD Radeon graphics. Tested here were Radeon Software Crimson Edition on Windows 10 vs. Ubuntu 16.04 with AMDGPU vs. Ubuntu 16.04 with the new AMDGPU PRO driver stack.

http://www.phoronix.com/vr.php?view=23071

[juno]

Parity!
Looks good

[duby229]

Only 2 actual games benched and of them only the Xonotic runs are meaningful. All the rest of them are synthetic results with no other actual gameplay results to prove or disprove them. Synthetics are meant to measure known bottlenecks, but without actual gameplay results there is no possible way to verify whether those bottlenecks are being suffered from. It starts with understanding the bottlenecks you want to measure, then benching synthetics that actually measure them, and finally benching games that prove or disprove them in real world loads.

I'm pretty certain that disparity in performance between real world games will be greater than this, but according to these results you can't see that.

[bug77]

Only 2 actual games benched and of them only the Xonotic runs are meaningful. All the rest of them are synthetic results with no other actual gameplay results to prove or disprove them. Synthetics are meant to measure known bottlenecks, but without actual gameplay results there is no possible way to verify whether those bottlenecks are being suffered from. It starts with understanding the bottlenecks you want to measure, then benching synthetics that actually measure them, and finally benching games that prove or disprove them in real world loads.

I'm pretty certain that disparity in performance btween real world games will be greater than this, but according to these results you can't see that.

Well, there's the Unigine benchmarks. Synthetics, bit they do stress the GPU a lot. A nice showing and a marked improvement over what we've come to expect from Catalyst.

[Xaero_Vincent]

It would be interesting to see benchmarks of games that had driver bottlenecks with past Catalyst releases, such as Bioshock Infinite, to see if such bottlenecks still exist and how things might be running under RadeonSI.

[juno]

duby229: phoronix tests have always been like this. There is obviously not much real-world value in these benchmarks, especially when it comes to Linux vs Windows gaming. But it is still enough to indicate trends or reveal bigger problems.
But this is still the #1 place to go for linux benchmark results and therefore influential. When people consider Linux and read about drawbacks like poor gaming capabilities, they search the web for confirmation. They lose their interest and developers/publishers don't get interested at all. So imho, good results are important to see here

[tomtomme]

Only 2 actual games benched and of them only the Xonotic runs are meaningful. All the rest of them are synthetic results with no other actual gameplay results to prove or disprove them. Synthetics are meant to measure known bottlenecks, but without actual gameplay results there is no possible way to verify whether those bottlenecks are being suffered from. It starts with understanding the bottlenecks you want to measure, then benching synthetics that actually measure them, and finally benching games that prove or disprove them in real world loads.

I'm pretty certain that disparity in performance btween real world games will be greater than this, but according to these results you can't see that.

someone really needs to make those pts-profiles work on win10. thats the whole reason michael did only benchmark xonotic etc. sadly I am no programmer nor do I know how to use git etc.

[humbug]

To add to this article; based on my testing AMDGPU Ubuntu Vulkan performance is better than the current Windows 10 Vulkan performance.

Although still lagging behind Nvidia (according to Michael's benchmarks).

[jf33]

Were the Unigine benchmarks on Windows done with OpenGL?

[Nille]

Were the Unigine benchmarks on Windows done with OpenGL?

I guess they are with OGL, because AMDs OpenGL Implementation is much slower that the D3D one.