A new benchmark video shows Dota 2 with Vulkan performing better on Windows than Linux

Quoting: liamdaweTo the people asking about the fullscreen mode used, it was stated in the article:

QuoteI spoke to the person who did it on twitter, he mentioned both Windows/Linux were done in borderless fullscreen mode.

It sounds like he's confused, because "borderless fullscreen" isn't a thing. There's fullscreen, and there's borderless window which can occupy the fullscreen, but just plain "borderless fullscreen"? No such beast exists.

The fact that he was running in a borderless window presumably with the compositor crunching away in the background makes me feel better about these results. Someone needs to do another benchmark using actual fullscreen mode or with the compositor disabled at the very least.


Last edited by Mountain Man on 15 June 2016 at 4:11 pm UTC

Quoting: dubigrasuWell, the options in Dota are Exclusive Fullscreen, Windowed and Borderless Window.

Not trying to impose anything, but is it possible for you to do a benchmark for Dota2 Vulkan? You are my favorite source when it comes to Linux/Windows benchmarks. You are doing an awesome job. :)

http://steamcommunity.com/groups/steamuniverse/discussions/1/541907867770822642/

It might not be the problem, but I feel, too, that using a brand new GTX 1070-1080 may not give a fair portrait of Linux performances. Usually it requires a couple of driver's iterations to get it's full capacity.

Sigh. This again..

Unless someone with Linux as a primary system who knows how to use the Linux Desktop properly then benchmarks something im never taking these as literal.

Why do they need it as a primary ? Well for one there is less confirmation Bias ( yes i believe Linux users tell things the way they are more in general than a windows centric press, they have nothing to hide and nothing to lose ). Also what compositor is used ? what settings on the compositor ? what desktop environment ? were multiple tried? what power settings / config settings on the driver ? Was there something other or weird about the testing process ? What kernal/Distribution ? How long has the card been out given that windows gets support for new hardware first ?

Im not denying that X11 is complete pants in 2016 (we have wayland with AMDGPU-PRO) and that windows does have a faster pipeline ( that's what billions of $ and two decades of dominance does ) , but the fact remains unless the above is met it cannot be a fair performance comparison.


@Liam it would be nice to get some testing between the two from some people on here who have both OS. Or in some performance articles because if we always take our benchmarks from Phoronix or the MSM gaming press there probably never going to be right.


Last edited by on 15 June 2016 at 4:40 pm UTC

Quoting: Mohandevir

Quoting: dubigrasuWell, the options in Dota are Exclusive Fullscreen, Windowed and Borderless Window.

Not trying to impose anything, but is it possible for you to do a benchmark for Dota2 Vulkan? You are my favorite source when it comes to Linux/Windows benchmarks. You are doing an awesome job. :)

http://steamcommunity.com/groups/steamuniverse/discussions/1/541907867770822642/

It might not be the problem, but I feel, too, that using a brand new GTX 1070-1080 may not give a fair portrait of Linux performances. Usually it requires a couple of driver's iterations to get it's full capacity.

Quoting: Mohandevir

Quoting: dubigrasuWell, the options in Dota are Exclusive Fullscreen, Windowed and Borderless Window.

Not trying to impose anything, but is it possible for you to do a benchmark for Dota2 Vulkan? You are my favorite source when it comes to Linux/Windows benchmarks. You are doing an awesome job. :)

http://steamcommunity.com/groups/steamuniverse/discussions/1/541907867770822642/

It might not be the problem, but I feel, too, that using a brand new GTX 1070-1080 may not give a fair portrait of Linux performances. Usually it requires a couple of driver's iterations to get it's full capacity.

I'm doing one now for my own curiosity, but not so extensive (with 4K and so on), just 1080. I might add a graph for it if you're curious, but video no.
Don't have much time and I think I'll wait a while for things to settle down (with Dota itself & drivers) before doing a video.

Another Comparision

It`s all about optimiziation... still opengl can get the same results like directx...

you should give a shit on a video from someone in the internet... it`s just one result... it says nothing...

Quoting: dubigrasuI'm doing one now for my own curiosity, but not so extensive (with 4K and so on), just 1080. I might add a graph for it if you're curious, but video no.
Don't have much time and I think I'll wait a while for things to settle down (with Dota itself & drivers) before doing a video.

Great! I'd be happy if you could share us your graph. Thanks!

Quoting: mr-egg@Liam it would be nice to get some testing between the two from some people on here who have both OS. Or in some performance articles because if we always take our benchmarks from Phoronix or the MSM gaming press there probably never going to be right.

As much as I would like to, it is very time consuming especially when I myself don't know the first thing about the Dota 2 command line benchmark tool.

If someone can PM me with details on how to do it easily, with a really heavy scene, then I can look at running a test on my 980ti on OpenGL, Vulkan across Windows and Linux myself.

...Besides...

Slice of life:

I'm working for a CAD company and we just upgraded our workstations to Win10... Not new workstations. Just a full and clean Win10 reinstall...

Conclusion:
if the situation apllies to gaming too, I can live with lower fps if it permits to avoid loads of BSODs... Crap! I've never seen that much BSODs in my entire life! Not one workstation... On all worksations with diffrent hardware configurations ranging from i7-2600 to i7-4790 with quadro gpus. Yuk! Yuk! Yuk!

End of slice of life, but still funny. :)

Just did the test with his demo file on:

Distribution: Manjaro KDE
Driver version: linux44-nvidia 1:364.19-6
Kernel: 4.4.13-1-MANJARO
Graphics card: GTX 760
CPU: Intel(R) Core(TM) i7-2600 CPU @ 3.40GHz
Resolution: 1920x1080,
Composit modes: disabled & enabled
Window mode: exclusive fullscreen & fullscreen borderless window
Graphics mode: vukan&opengl
Number of runs: 10 each graphics mode & OS

Result:
- No difference between exclusive fullscreen and borderless window fullscreen. They seem to set the X flag there.
- No difference between composite enabled or disabled (would hit it that they set the flag, kwin is checking for it to suspend composite)

Vulkan: 58-117
OpenGL: 38-115

OpenGl stuttered from time to time, Vulkan was completely fluent.

================
Result on Windows10, driver 368.39:
DX: 83-113
Vulkan: 71-122


So I can back his data, looks pretty similar here too, though, I have a lot weaker graphics card and bet my CPU is pretty much less up-to-date than his ;-).

I usually play with vsync, so in my particular case: I couldn't care less in dota2, but there is work to do if our driver should be as good as the Windows for other games reaching the same performance.


Last edited by STiAT on 15 June 2016 at 6:04 pm UTC

It seems only fair to run the tests on SteamOS in my opinion. Seriously Ubuntu is trash