What's the best Distro for gaming + desktop?

[Mike Morkleson]

(If it counts, my cpu is Intel i3 6100 and the gpu is PowerColor Radeon R9 380.)

I want a distro with a good desktop experience so I can do schoolwork and internet browsing but still pulls its weight in gaming.

[SRH]

โพสต์ดั้งเดิมโดย Zephar:

โพสต์ดั้งเดิมโดย Mountain Man:

I use Xubuntu, which is Ubuntu with the Xfce desktop environment. I was a big fan of KDE for a long time, but it's too bloated these days, and I find that Xfce offers much better performance across the board while still being a full-featured desktop.

100% agree with mountain man here.

+1

[Mountain Man]

โพสต์ดั้งเดิมโดย Maleko:

โพสต์ดั้งเดิมโดย Mountain Man:

I use Xubuntu, which is Ubuntu with the Xfce desktop environment. I was a big fan of KDE for a long time, but it's too bloated these days, and I find that Xfce offers much better performance across the board while still being a full-featured desktop.

Which compositor do you use with Xfce?

I don't know. Whatever comes default with Xubuntu.

But now I have to amend my comments. I did some experimenting last night with KDE because I haven't used it in many months and was curious if things had improved, and it turns out the secret to good gaming performance is to disable compositing for full screen applications. You can do this manually with, I think SHIFT ALT F12, or you can set it to happen automatically through the display options. Doing this, I found identical performance between Xubuntu and Kubuntu. I confirmed this running several Uningine and in-game benchmarks (Tomb Raider and Shadow of Mordor).

As far as the actual desktop environment, Xfce is definitely lighter weight and quicker to start up, but KDE seems a little snappier with applications starting quicker.

So I guess the conclusion is that if you set up KDE correctly then it works fine for gaming while Xfce works "out of the box" with default settings. As for which desktop environment you prefer, it's a matter of preference. KDE is a lot snazzier while Xfce is more no-nonsense.

[Diarrhea Larry]

โพสต์ดั้งเดิมโดย Mountain Man:

โพสต์ดั้งเดิมโดย Maleko:

Which compositor do you use with Xfce?

I don't know. Whatever comes default with Xubuntu.

But now I have to amend my comments. I did some experimenting last night with KDE because I haven't used it in many months and was curious if things had improved, and it turns out the secret to good gaming performance is to disable compositing for full screen applications. You can do this manually with, I think SHIFT ALT F12, or you can set it to happen automatically through the display options. Doing this, I found identical performance between Xubuntu and Kubuntu. I confirmed this running several Uningine and in-game benchmarks (Tomb Raider and Shadow of Mordor).

As far as the actual desktop environment, Xfce is definitely lighter weight and quicker to start up, but KDE seems a little snappier with applications starting quicker.

So I guess the conclusion is that if you set up KDE correctly then it works fine for gaming while Xfce works "out of the box" with default settings. As for which desktop environment you prefer, it's a matter of preference. KDE is a lot snazzier while Xfce is more no-nonsense.

That'd be correct. KDE is heavier, but they're both worth what they consume. Xfce's compositor can actually slow down games on some GPUs, but I think it's fine on Intel and Nvidia. If not, I'd say use Compton.
Plasma 5 is very light compared to the old KDE 4, though still not tiny it has enough features to replace Windows, including krunner, a very good equivalent to OS X's spotlight. If you like being able to search/index files, pick Plasma over GNOME, as GNOME's indexer (named Tracker) can put a strain on the CPU whenever it's "tagging" files. It made some of my machines pretty warm.
I'd pick Xfce if you don't need compositor effects other than shadows/translucency/Vsync, or if you don't want or need a full-feature file manager/search. If your machine has 4GB RAM and Intel HD 3000 or newer, Plasma 5 can still run Source games just as fast, if you prefer it. Tested both on a T420.

If you don't mind having a lot of applications, you can obviously just install both desktops and switch whenever you want. Hell, this isn't Windows.

[halifax]

โพสต์ดั้งเดิมโดย Mountain Man:

I did some experimenting last night with KDE because I haven't used it in many months and was curious if things had improved, and it turns out the secret to good gaming performance is to disable compositing for full screen applications. You can do this manually with, I think SHIFT ALT F12, or you can set it to happen automatically through the display options. Doing this, I found identical performance between Xubuntu and Kubuntu. I confirmed this running several Uningine and in-game benchmarks (Tomb Raider and Shadow of Mordor).

I've never noticed gaming performance dips due to DE's compositor. But I've never really tested, either. After reading your post, and remembering so many posts here from other linux users talking about DE compositors negatively impacting gaming performance, I did a small test.

OS: Debian 8.4 Jessie
DE: Plasma 4 / KDE 4
Graphics Driver: nVidia 352.79
Graphics Card: GTX 770
Game: Dying Light

I found a spot outside on a communications tower, where my framerate is sustained lowest looking out over the town. using SHIFT+ALT+F12, and ALT-TABBING in and out of the game to verify compositor switching off and on:

Compositor on: 41 fps sustained
Compositor off: 41 fps sustained

Is it different video cards? PCs from the late 70's you're still on, why are you guys convinced this is an issue, am I missing something? X-) Is it something besides sustained/avg FPS in games you're noticing?

[Drako Frost]

โพสต์ดั้งเดิมโดย halifax:

โพสต์ดั้งเดิมโดย Mountain Man:

I did some experimenting last night with KDE because I haven't used it in many months and was curious if things had improved, and it turns out the secret to good gaming performance is to disable compositing for full screen applications. You can do this manually with, I think SHIFT ALT F12, or you can set it to happen automatically through the display options. Doing this, I found identical performance between Xubuntu and Kubuntu. I confirmed this running several Uningine and in-game benchmarks (Tomb Raider and Shadow of Mordor).

I've never noticed gaming performance dips due to DE's compositor. But I've never really tested, either. After reading your post, and remembering so many posts here from other linux users talking about DE compositors negatively impacting gaming performance, I did a small test.

OS: Debian 8.4 Jessie
DE: Plasma 4 / KDE 4
Graphics Driver: nVidia 352.79
Graphics Card: GTX 770
Game: Dying Light

I found a spot outside on a communications tower, where my framerate is sustained lowest looking out over the town. using SHIFT+ALT+F12, and ALT-TABBING in and out of the game to verify compositor switching off and on:

Compositor on: 41 fps sustained
Compositor off: 41 fps sustained

Is it different video cards? PCs from the late 70's you're still on, why are you guys convinced this is an issue, am I missing something? X-) Is it something besides sustained/avg FPS in games you're noticing?

I don't think it works on the fly, you have to close the game and launch it again. Composition causes a big impact in performance, I lose about 20-30 FPS if I play with composition enabled.

[halifax]

โพสต์ดั้งเดิมโดย Drako Frost:

I don't think it works on the fly, you have to close the game and launch it again. Composition causes a big impact in performance, I lose about 20-30 FPS if I play with composition enabled.

20-30 fps?!? That would make me track this issue, if I was seeing differences like that. I see sub 1 fps differences, doesn't matter if I exit the game and disable composting first, I still get the same fps as with compositing on, in Dying Light anyway.

20-30 fps though, I almost want to call bullsh** - that's unimaginable the compositor would be taking that much GPU bandwidth away from games ;-) Not quite as bad as OpenGL didn't load so the game falls back to software renderer, but it's similar. The compositor is not working your graphics card hardly at all, it's not taking enough bandwidth to really even slightly matter. My GTX 770 stays at an idle temp of 30 degrees Celsius on the Plasma 4 desktop whether the compositor is on or off.

Does your video card run at hotter than idle temps with the compositor on?

[Drako Frost]

I did the test a while ago, I used an AMD Radeon HD 7870, Xubuntu (because you can easily disable composition) and Left 4 Dead 2. I did the test in the last map of the game, because right in the begining of the map there's a waterfall, and the FPS drops quite a bit there. I had almost 30 FPS less with composition enabled.

Before Steam for Linux, Unity and Gnome didn't even support the function "unredirect fullscreen windows", they introduced the feature because of Steam.

I can't test this on Ubuntu, Unity doesn't have an option to turn composition on and off manually.

[Mountain Man]

I did some brief testing this morning in Kubuntu. Here are the results:

Tomb Raider
Compositing off
Min 41.7
Max 60.5
Avg 58.1

Compositing on
Min 34
Max 60.3
Avg 58

So a pretty substantial hit to the minimum FPS. Maximum FPS tops out at 60FPS in both cases probably because I had vsync enabled.

The next one was a bit surprising.

Shadow of Mordor
Compositing off
Min 16.69
Max 77.67
Avg 33.43

Compositing on
Min 16.83
Max 83.06
Avg 33.72

No idea why compositing on achieved a higher maximum framerate, unless I got my tests mixed up and was running with compositing off when I thought it was on and vice versa. I was in a bit of hurry and didn't have to time to run multiple tests to double check my numbers. But the fact that the minimum is effectively identical is interesting.

I'll have to play around with it more after work, maybe run some tests without vsync enabled and see what happens. But at the moment, I'm really not seeing any reason to not use KDE if you want a full-featured desktop environment so long as you're aware that leaving the compositor enabled all the time can impact performance.

[halifax]

I did a second test on a much more modest gaming PC I have in the house:

OS: Debian 8.4 Jessie
DE: Plasma 4 / KDE 4
Graphics Driver: nVidia 352.79
Graphics Card: GT 730 (much more low power card than the GTX 770)
Game: Saints Row: The 3rd

Exact same render scene in each test, I would not touch the mouse at game start in the same crib.

comp off: 25-26 fps (FPS counter constantly flips between two numbers)

comp on: 24-25 fps (FPS counter constantly flips between two numbers)

So, 1 FPS difference on a machine that is under serious graphics load for its hardware.

I think testing the compositor being off and on is a valid clean test in Debian 8, because it is night and day on the desktop, ALT+SHIFT+F12 clearly toggles compositing fully on and fully off - Plasma 4 has a lot of compositing visual effects that cue you if it's on or not. Ex: the KWin "Toggle Present Windows" hot-key command doesn't even work with compositing off.

[Mountain Man]

To make things even more confusing, it's my understanding that software can automatically disable compositing if the programmer chooses to use that feature, but not every program does it, which could account for oddities like one program showing a substantial difference in performance while another doesn't.

[halifax]

โพสต์ดั้งเดิมโดย Mountain Man:

To make things even more confusing, it's my understanding that software can automatically disable compositing if the programmer chooses to use that feature, but not every program does it, which could account for oddities like one program showing a substantial difference in performance while another doesn't.

I think from the tests I've seen and test you've directly done here, there is one takeaway IMO. If you are experiencing your Linux DE compositor causing a serious FPS performance impact on your high end 3D games:

1) Your distro most likely has software bugs
2) The established "industry standard" norm in Linux is that the DE, no matter how fancy, should not use more than a negligible amount of GPU bandwidth to render.
3) If you're not seeing a negligible amount of GPU bandwidth used by your DE compositor, refer to point 1, and pursue fix actions accordingly. Up to and including abandoning that particular distro as not worth its weight in salt :-)

[Mountain Man]

โพสต์ดั้งเดิมโดย halifax:

โพสต์ดั้งเดิมโดย Mountain Man:

To make things even more confusing, it's my understanding that software can automatically disable compositing if the programmer chooses to use that feature, but not every program does it, which could account for oddities like one program showing a substantial difference in performance while another doesn't.

I think from the tests I've seen and test you've directly done here, there is one takeaway IMO. If you are experiencing your Linux DE compositor causing a serious FPS performance impact on your high end 3D games:

1) Your distro most likely has software bugs
2) The established "industry standard" norm in Linux is that the DE, no matter how fancy, should not use more than a negligible amount of GPU bandwidth to render.
3) If you're not seeing a negligible amount of GPU bandwidth used by your DE compositor, refer to point 1, and pursue fix actions accordingly. Up to and including abandoning that particular distro as not worth its weight in salt :-)

The problem, I think, is that not all software interacts well with a compositor running in the background, which isn't necessarily a fault of the compositor.

[Hwkiller]

I think plasma has a decent solution to the compositor issue (no idea if this works well with wayland yet though... wayland requires a compositor properly, but compositors can disable features in an analogous way).

Plasma/kwin will automatically fall back to X11 2D, non-compositing if a window is flagged as 'fullscreen', which most games are.
With wayland, I'm guessing the solution would be to disable various compositing features when any 'fullscreen' windows are active.

[Diarrhea Larry]

โพสต์ดั้งเดิมโดย Hwkiller:

I think plasma has a decent solution to the compositor issue (no idea if this works well with wayland yet though... wayland requires a compositor properly, but compositors can disable features in an analogous way).

Plasma/kwin will automatically fall back to X11 2D, non-compositing if a window is flagged as 'fullscreen', which most games are.
With wayland, I'm guessing the solution would be to disable various compositing features when any 'fullscreen' windows are active.

Hopefully, the Wayland-based desktops add that equivalent functionality before the protocol becomes the new de-facto. I tested Source games on GNOME 3.18's Wayland session a while ago and it was horrid.

[Hwkiller]

โพสต์ดั้งเดิมโดย autstanding:

โพสต์ดั้งเดิมโดย Hwkiller:

I think plasma has a decent solution to the compositor issue (no idea if this works well with wayland yet though... wayland requires a compositor properly, but compositors can disable features in an analogous way).

Plasma/kwin will automatically fall back to X11 2D, non-compositing if a window is flagged as 'fullscreen', which most games are.
With wayland, I'm guessing the solution would be to disable various compositing features when any 'fullscreen' windows are active.

Hopefully, the Wayland-based desktops add that equivalent functionality before the protocol becomes the new de-facto. I tested Source games on GNOME 3.18's Wayland session a while ago and it was horrid.

It's not going to be the same. Many games require Xserver, so you're still running an X session within wayland. Don't compare the two until more programs actually support the wayland protocol as opposed to the X11 protocol nested within wayland.