NVIDIA GeForce GTX 1050 Ti Linux Benchmarks

[oooverclocker]

To be fair, things might be different in Vulkan/DX12. I haven't checked out how well perf there lines up with the TFLOP values. Perhaps async shaders and other tech lets the GPU be fully utilized.

It looks like you have actually understood what the problem is. TFLOPS are the product of ALUs and their clocks so it's absolutely not like comparing just MHz. It's the performance the GPU can theoretically deliver and is dependent on the driver and the application running. On old stone age games and casual games that are poorly implemented you get of course more fps with less but higher clocked ALUs. Although AMD cards deliver more performance for the same piece of silicon in comparison with current Nvidia GPUs that's one point they suffer from. There are other pieces in a GPU but their performance impact is not significant and they have about the same quantity no matter which brand.
So when it comes to Linux we see a mixture of immature drivers, games that are not using the full capabilities of these drivers, and games that don't stress the GPU enough to get its best performance.

Well optimized implementations that get the best out of every GPU are for example DOOM with Vulkan on Windows and Xonotic. FarCry Primal also shows a reasonable performance when you help the GPU usage a bit by playing in high resolutions. Quantum Break and Hitman(DX12) also use the GPU well but they partly perform better for AMD cards than one should expect. There are currently about no games at all that stress the cards optimally than Xonotic on Linux.

There are other posts writing against my comment. If someone writes something like a game "prefers team red GPUs" or "prefers team green" it makes no sense to answer him because his logic is different from mathematical logic, and he usually has infinite time to argue in his logic while my time is limited and I can do something more useful instead. This is what I learned from writing and reading in many forums in several years. So for anyone who don't likes to understand what I wrote or has its own metric for calculating GPU performance - just believe it or not. I won't waste my time trying to convince you.

[SaucyJack]

...Why?

Because of the obvious. AMD cards pack more advanced hardware that's only beginning to get utilized. Game studios have been wasting their potential so it 'appears' that nvidia cards are better in utilizing its tflops when that isn't the case at all. Google some Vulkan benchmarks for doom and add in ID software said themselves that they could have done a lot more with it.

[eydee]

Xonotic again... Nothing represents current-gen graphics in a test of a brand new video card. Unigine needs to hurry up with their new benchmark.

[dungeon]

Xonotic again... Nothing represents current-gen graphics in a test of a brand new video card. Unigine needs to hurry up with their new benchmark.

Unigine better to do some Vulkan benchmark at this time... but Xonotic is fine, if someone play it, it show him very good performance on opensource radeosi driver.

BTW, Marek announced yesterday on irc that multithreading is there for early testers, so i guess people who like to test early stuff might wanna try it

https://cgit.freedesktop.org/~mareko/mesa/commit/?h=glthread-5-rebased&id=970a000921a75c2f3c8e5736106e7eae7a625145

Expect additional boost with it... see there Borderlands 2 goes 70% faster ... and likely bugs too, but that is how development goes

[nadro]

@efikkan
If you want to see a nice implementation of Vulkan, just run Doom. In short... NV Maxwell/Pascal arch is much worse than latest GCN for low level APIs and this is a reason why AMD is better for D3D12/Vulkan. You can't just fix in software some lacks in hardware...