Windows 10 vs. Ubuntu 16.04 Gaming With NVIDIA's GTX 1070 & GTX 1080

[Kano]

I know that the option is there but did you compare the hair?

[mike4]

Yes I compared all by booting Win10 and have played most of the game on both. With exactly the same settings on both os I only remarked Win10 to be a little darker. Otherways I cannot spot any difference at all. Benchmark gives ~50 on Win10 and 70 fps on Ubuntu (1680x1050) but seems about the same in game.

[Cape]

Not sure what difference you saw, but from an X server / driver perspective there's no difference between the two cases. In both cases the game fullscreen window is displayed directly, without any involvement by gnome-shell / kwin.

I personally noticed a difference in CSGO (probably the only game i play without vsync) where i get around 260FPS compared to Gnome where i get 240.

The difference is that you take away any task that the GPU is taking care of for compositing, which also free up some video memory on your card.
I take this article from Martin Grassling (kwin developer) as a reference https://blog.martin-graesslin.com/bl...the-right-way/

[DJViking]

This was disappointing. I was hoping for better results on Linux.
The previous benchmark done on The Talos Principle showed it performed better on Linux

https://www.phoronix.com/scan.php?page=news_item&px=Win10-Linux-Vulkan-Early

What has happened since then?

[DJViking]

It looks like all ports from Feral are the ones performing much worse on Linux, while other ports don't have such issues.

I had the same thought... Though I am thankful for Feral porting games to Linux, they could do better with optimizing for performance.

[DJViking]

Also they tend to implement some kind of DX-OGL wrapper instead of a completely rewritten rendering engine. (Even Talos does something like this with Vulkan at the moment) Sometimes it works better, sometimes terribly.

Seriously... I thought Talos had native both OpenGL and Vulkan on Linux.

[Kano]

@mike4

You are blind! Compare the Linux benchmark on highest settings against

https://www.youtube.com/watch?v=KSZLOYuLEK0

On Linux the ponytail is never shown with so many hairs! If you don't see it, record it...

[Passso]

Nille has already declared that the linux version isn't running all the same effects.
You expect us to go against The Nille?

I just state that if there are some visual differences in both max quality I do not see them.

[gamerk2]

Two thoughts:

1: Using a DX to OGL wrapper is going to eat a lot of performance.

2: I'm still convinced the Linux scheduler is inefficient for these types of workloads. Michael, is there any way to peak at GPU utilization on Linux? I'm convinced the thread(s) that actually perform the rendering aren't running as much as they should, since Linux is treating every thread in the system equally. In short: I'm convinced the primary application threads are getting bumped by super-low priority background tasks.

[liam]

I just state that if there are some visual differences in both max quality I do not see them.

Heh.
I was actually having a bit of fun at the expense of Nille since they made declarations without giving us any reasons to believe them.
In general, Michael will point out issues with rendering,including when games don't render a scene in the same way.