Shadow of the Tomb Raider Now Officially Available For Linux

[xcom]

It's 39fps LLVM vs. 45fps ACO here after loading the same savegame.
So if Michael won't benchmark with ACO, you can add a fair share of performance with at least Polaris.

How to enable ACO eg. on Manjaro with RADV?

[jrdoane]

Ok, we do disagree on that first part as for me using this source-level-translation from DX to OGL calls is somewhat important for the overall experience. If both worked equally performant, I also wouldn't care so much about this distinction, but unfortunately there is such a performance impact (and I have observed major differences on my hardware in Company of Heroes 2 which is also a Feral port).

Being a native binary doesn't automatically mean that performance is good. I can write crap code that performs like garbage, but still runs natively. Good performance isn't a requirement for an application to be running natively.

I understand your concern and it's totally valid, but that's not what running "native" means.

[aufkrawall]

It's 39fps LLVM vs. 45fps ACO here after loading the same savegame.
So if Michael won't benchmark with ACO, you can add a fair share of performance with at least Polaris.

[cute2dgirl]

Maybe DX12 - > Vulkan is easier, but the performance will be worse than DX11.

Not in this game, D3D11 is often CPU-limited:

https://www.youtube.com/watch?v=GjuPA6HrFIU

RotTR also runs better with D3D12 than D3D11 on AMD.

[CochainComplex]

DX12 games's performance is not the best currently, like Deux EX: MD, RottTR. DEVs are much better in DX11. Ergo I think DX11 -> Vulkan is currently the best option to port a game.
Maybe DX12 - > Vulkan is easier, but the performance will be worse than DX11.

Maybe, because Optimization was done by GFX Cards producers with drivers (under DX11). Now (DX12), Gamestudios have to do the low level work by their own - which was their wish btw. But this comes with a lot of work on the engine side. Hence, economists are cutting corners by still using dx11 as target api.

[xcom]

You will not see the same performance unless the game is rewritten from scratch with Linux and Vulkan as the intended target. For a game DX12 game like this the game engine have millions of lines of code that have been optimized for DX12.

DX12 games's performance is not the best currently, like Deux EX: MD, RottTR. DEVs are much better in DX11. Ergo I think DX11 -> Vulkan is currently the best option to port a game.
Maybe DX12 - > Vulkan is easier, but the performance will be worse than DX11.

[F.Ultra]

Another reason I am a bit grudgy here is that the performance delta between the Linux and Windows version of one of their ports which I play is massive on my hardware. I'd expect the same or at least similar performance from a "native port".

You will not see the same performance unless the game is rewritten from scratch with Linux and Vulkan as the intended target. For a game DX12 game like this the game engine have millions of lines of code that have been optimized for DX12.

[Ifr2]

People were quick to do the benchmark: https://www.youtube.com/watch?v=GjuPA6HrFIU

And the avg frame rate is only a 1.6% lower than on Windows. That is an awesome job devs did there.

[xcom]

For me if it can run without X or Xwayland it's native. Proton, Wine runs only with X so its not native. Use of sdl is also a good indicator

[lectrode]

Ok, we do disagree on that first part as for me using this source-level-translation from DX to OGL calls is somewhat important for the overall experience. If both worked equally performant, I also wouldn't care so much about this distinction, but unfortunately there is such a performance impact (and I have observed major differences on my hardware in Company of Heroes 2 which is also a Feral port).

Pretty much all of Feral's DX->OpenGL ports suffered a notable performance hit. Modern ports from Feral don't suffer from that particular issue, as Vulkan has proven to be much more performant for translating DX calls. The benchmarks linked to by sarmad show Linux native and DX12 battling it out at the top usually staying within 10 FPS of each other, with either outperforming the other depending on which part of the benchmark you're in. I'd say that's pretty damn performant. Feral has more than earned the right to call it "native".