A Four Way Comparison For Alien: Isolation Shows Off The Performance Difference

Quoting: sigzAt best you will have near same fps as in windows in some cases, most of times you will get lower fps, and you will never have a 30% more fps on linux than windows.

You might if someone had to port a Linux game written around an OpenGL renderer to Windows using Direct3D, but that won't happen. You know, with OpenGL being cross-platform and all.

If you play on a PC long enough, you now that PC games in general are bad programmed in first place. Companies will only do bug hunting and optimizations only when is lucrative enough to then.

Feral, Aspir, VP, etc do not have deep pockets or much man power to do their jobs. They will do a job and need to make a profit. If go after that last 30% loss means loose money, they will not do it. Simple as that.

Games in general do not complies wit OpenGL or DirectX to the letter, so they put the blame to the driver developer. If a game is famous enough, they will got a driver patch to correct the errors made by developers. You see this all the time in Windows. But if you are a indie, or you learn your job really well, or your games will never see high performance.

The Nvidia driver is more tolerant to programming errors, that is their advantage. The AMD driver needs more compliant code to work well, but that almost never happens because people thinks NvidiaGL means OpenGL...

The hope with Vulcan is that there will be official compliant tests, so game developers will not have the excuse to blame the video driver to their game bad performance.


Last edited by M@GOid on 18 November 2015 at 2:06 pm UTC

Quoting: tuubi

Quoting: sigzAt best you will have near same fps as in windows in some cases, most of times you will get lower fps, and you will never have a 30% more fps on linux than windows.

You might if someone had to port a Linux game written around an OpenGL renderer to Windows using Direct3D, but that won't happen. You know, with OpenGL being cross-platform and all.

Well, Planetary Annihilation is a pure OpenGL game, it has been done with multiplatform in mind, I tested it under windows and I got 20% more fps than linux (Of course I talk about my system only, I didn't found any comparison benchmark)

Quoting: sigz

Quoting: tuubi

Quoting: sigzAt best you will have near same fps as in windows in some cases, most of times you will get lower fps, and you will never have a 30% more fps on linux than windows.

You might if someone had to port a Linux game written around an OpenGL renderer to Windows using Direct3D, but that won't happen. You know, with OpenGL being cross-platform and all.

Well, Planetary Annihilation is a pure OpenGL game, it has been done with multiplatform in mind, I tested it under windows and I got 20% more fps than linux (Of course I talk about my system only, I didn't found any comparison benchmark)

That'll happen, depending on the driver/hw and the amount of optimization a developer puts in for a given platform. Also, not relevant to my hypothetical scenario at all.

Quoting: GuestFeral's version is called "indirectx". Unless that's been changed in recent times. I'm sure Feral will correct me if I'm wrong.

Ah, thanks for the info.

Quoting: Guestdx9 might well be easy enough, but dx11 isn't quite so simple. And you don't want to strip apart an entire game codebase to replace everything if you can avoid it - something like togl is really the only way to go.
You will get performance hits because of trying to essentially emulate how DirectX does things, and then there's the mess of converting HLSL bytecode to something GLSL compatible.

This was pretty much my point. There will be overhead if you emulate an incompatible API, source based or not.


Last edited by tuubi on 18 November 2015 at 2:33 pm UTC

Quoting: dubigrasuView video on youtube.com

War Thunder (awesome game) is a wrapped game, or a native port?

Quoting: leillo1975

Quoting: dubigrasuView video on youtube.com

War Thunder (awesome game) is a wrapped game, or a native port?

native. has opengl option for windows version also

what about settings ? Sure people may not like to have to drop AA, but linux is behind in Multi sampling Antialiasing at the moment. Does it use it on this test ? If so then doing the test without may not be popular ( because people want the same quality across the board ) but it would at least identify how much of that 30% difference was down to a specific game setting.

Still, running 1080p 60fps with better than console graphics is entirely possible even with a 30% loss on a decent PC / Steam machine.

Quoting: mr-eggwhat about settings ? Sure people may not like to have to drop AA, but linux is behind in Multi sampling Antialiasing at the moment. Does it use it on this test ? If so then doing the test without may not be popular ( because people want the same quality across the board ) but it would at least identify how much of that 30% difference was down to a specific game setting.

Still, running 1080p 60fps with better than console graphics is entirely possible even with a 30% loss on a decent PC / Steam machine.

The settings were at max for all systems (yes AA enabled), read the video description for details.

I kinda don't care. Even on my aging GTX660OC it performs well enough that I can play it, and that's all I need. I mean... What's the alternative, not play it? Buy it for my dusty PS3? I'm sure as hell not going to install Windows again, after 13 years without it.

It runs really great on my Core i7, GTX660 and 16Gb of RAM with everything maxed out and I am quite impressed with the graphics.

Maybe I'd be singing a different tune if the performance numbers favored Linux, but I think people are making too big of a deal over this. No, performance is not where we'd like it or where it should be, but anybody with a decent computer will still have great performance in Linux. We're not talking a difference between playable and unplayable.


Last edited by Mountain Man on 18 November 2015 at 3:57 pm UTC

War Thunder is the proof that Linux performance can be on par with Windows, if the game is programmed in the right way. I play WT under both operating systems and I can't see any difference at all.

Unfortunately, later DX->OGL ports done by an external developer will, most of the time, be sub optimal because of budget constraints, which make impossible to rewrite the engine and so require the use of a translation layer. We must accept it as a fact of life.

Quoting: Mountain ManMaybe I'd be singing a different tune if the performance numbers favored Linux, but I think people are making too big of a deal over this. No, performance is not where we'd like it or where it should be, but anybody with a decent computer will still have great performance in Linux. We're not talking a difference between playable and unplayable.

Yep, that reminds me the perfs comparison of Fallout 4 between PS4 and Xbox One. There are little loss of FPS on the XO, but nothing dramatic in game, until the frame freezes for half a second on XO. That is the kind of difference between playable and unplayable.

Quoting: GuestImpressively playable in all cases... Windows (presumably on the original DX11 code) has 50% better performance on nvidia than on Linux, whereas on AMD it has 100% better performance. But it works, in AMD's case I believe it's driver issues (AMDGPU just got a major update though, 280X uses the AMDGPU driver stack right? I wonder how the performance would have been with the powerplay update), in Nvidia's case I believe it's OpenGL issues, this game seems to be very well ported to opengl, no denying this. Guess we should just face it that OpenGL isn't really on par with DX11; unless they're using an outdated version of GL. It might be possible to optimize it to come close, but lets face it OpenGL is and has always been a mess, hopefully Vulkan won't be.

Disclaimer: I'm no 3D prog, but I've lived the change from OpenGL to DX.

OpenGL is not a bad API. A few years ago, when it was OpenGL 3.3, it was largely better than DirectX, then DX10 came out, simplified the communication with the GPU, ditched old ways of doing 3D prog and had better results. Khronos didn't make a move despite the fact that they were progressively losing game developers using OpenGL, and thus losing knowledge. They realized, too late, that they had lost their momentum.

And then, after years of no improvement at all, OpenGL 4.0 came out, and then 4.1, 4.2, 4.3, 4.4 and 4.5! Simplifying and ditching away the old API... In other words: to be on par with the DX 11 API.

So no, OpenGL is not a bad API, it has as much overhead as DX 11 has. It can do the same things as DX 11 can. It's just not used anymore by game developers: the skilled people went on the Microsoft side. That means that any tutorial or training now are made for DX, not for OpenGL. So it's more difficult now to write efficient OpenGL code, but still completely doable.

And Vulkan, as much as I'm waiting for it too, won't prevent the infinite ways of coding the wrong way. Moreover I don't think we'll see serious games on Vulkan until quite some time and there is the real threat that DX 12 will still keep its momentum, so I wouldn't get my hopes up already.


Last edited by Creak on 18 November 2015 at 5:33 pm UTC

Quoting: CreakOpenGL is not a bad API. A few years ago, when it was OpenGL 3.3, it was largely better than DirectX

I've read a very interesting and thorough history of D3D vs OpenGL, from the early '90s till today, on StackExchange:

http://programmers.stackexchange.com/questions/60544/why-do-game-developers-prefer-windows/88055#88055

D3D was substantially worse than OpenGL until DX8 came out. Then things changed forever and D3D won the battle.

Quoting: Creakthe skilled people went on the Microsoft side. That means that any tutorial or training now are made for DX, not for OpenGL. So it's more difficult now to write efficient OpenGL code, but completely doable.

That's true and is a very big problem. Documentation and programmers know how is orders of magnitude more on the DX side.
Also, the last advocate of OpenGL in the gaming world, John Carmack, is not in the industry any more.


Last edited by whatever on 18 November 2015 at 6:01 pm UTC

Well this video comparison actually gives us something. In stressful situations Steam OS performance doesn't change all that much, yet windows can drop up to 60fps! It basically means something is blocking performance rather than that there's more processing power being used. That together with the -O2 debian compile can indeed make sure that the steam os benchmark is slightly worse than the windows benchmark.

So if Feral interactive is reading this: could you please look if you can find the max fps issue? I'm pretty sure it would put a lot of people at rest about this.

Quoting: nattydread

Quoting: TheBoss

Quoting: nattydreadis it a native build or a port?

It was ported by Feral Interactive, and it is native. A port doesn't make something not native.

well what I mean perhaps is: does it use a wrapper to DirectX or does it directly use OpenGL? There must be some reason it doesn't perform as well on SteamOS.

Whether it's a wrapper or directly on OpenGL doesn't mean much. Porting it to OpenGL is something, and optimizing it for OpenGL is something else. Companies won't spend time optimizing the game unless there is a reason for that, and when it comes to this game there isn't much benefit for end users, since the game is already running above 60fps on current Steam Machines. Benchmarks have shown that when the game is optimized for Linux they run at similar performance to Windows. An example is L4D2.

Well shit, 60 FPS+ I guess all 4 options are acceptable to me.

Sure it brings a grin to my face when I drop $400 - $600 on a new GPU and get ungodly FPS boosts so that I'm outputting 120-260 FPS, but it's not a thousand percent necessary.

Meanwhile PS4 and XBONE are outputting at 30 FPS lol.

This benchmark is pretty well done, I wish the benchmark of Ars Technica would have been that professional instead or using an old rig at exaggerated resolutions that now make people believe that recent games on Steam Machines are unplayable on High setting which is simply not true. Here everyone can see that even if the game doesn't have the same performances that Windows have, on the lowest end Steam Machine it should still run well above 60 FPS on Very High setting, and on a game like that there absolutely no reasons to have more than 60 FPS if you are already on the maximum setting.

Quoting: ElectricPrismMeanwhile PS4 and XBONE are outputting at 30 FPS lol.

And sometimes not even 1080p...

I wonder what the difference is if you run Nvidia driver 355.11, since that's what's recommended in the README.