Tweet
This is an experiment of mine, where I want to compare the performance in games between popular distros.
This experiment is motivated by the appearance of the Nobara project, which is a distribution targeted for games, and the fact that I have been slacking off in upgrading my Ubuntu 21 installation.
This system has a Ryzen 7 5800X/Radeon RX 6800. As one does, I just thought the natural course of action, when presented with the big Ubuntu 21 to 22 update, would be to compare performance between the two, and then install Nobara and Garuda from scratch and compare results with those too.
I planned to test Garuda last, since I have tried it before, and I know I'm most likely to stick to it for a while.
The "performance of the distro" is to be measured "off-the-shelf", meaning without any tinkering, no custom kernels unless they come with the distro.
The "performance in games" is meant to be realistic. This is why I will try to include rather popular games like Dota2 or a custom-made profile for Final Fantasy XIV.
Here are what I found the most significant results.
2022_07_26_games_distro_comparison.png
Like I said, I put Garuda last because I expected to stay on Garuda. The distro seemed very consistent the last time around: I made a similar comparison on January. Back then, I tested Clear Linux out of curiosity as well, but it's rather complicated since some games won't run and one must use the Flatpak Steam. The performance was not so different from other distros, in any case.
Not all the results were uploaded, there is such thing as too many results in PTS:
Below is a more detailed explanation of why I chose the test profiles that I chose, and the setbacks that I encountered.
Dota 2 major 2022
My first setback came when realizing that my Dota2 installation seems not to be able to load the demo (dota2-pts-1971360796.dem) for the old PTS test profile pts/dota2. Demo was running fine last year, so this must have happened at some point last year, with some of the game updates.
If I use a different demo, it won't be the same results anymore, so I may as well create a new test profile. This new profile will be basically same as pts/dota2 but running a new demo, which is compatible.
The name of the test profile will be "local/dota_major_2022".
The demo used will be "6582379623.dem", Game 4 of the finals of Dota 2 Major in Stockholm, 24 May 2022, 19 minutes in, 6 minutes of gameplay.
Can be run manually this way:
steam -applaunch 570 +con_logfile <logfile>.log +timedemoquit replays/658237962 3 +demo_quitafterplayback 1 +cl_showfps 2 +fps_max 0 -nosound -noassert -console -fullscreen +timedemo_start 60000 +timedemo_end 66000 -autoconfig_level 3 -testscript_inline "Test_WaitForCheckPoint DemoPlaybackFinished\; quit"
Total War Warhammer 2
My second setback came when TWW2 was crashing, after doing the update to Ubuntu 22 (segmentation fault). Steam is unable to run the game either. Since both the test profile and the game were running fine on Ubuntu 21, and I didn't change anything from the configuration, I count that run as a "failed".
I am sure that probably the Proton version may be working for this game, but it is reasonable to expect players to be running the Linux native version, as this game has a lot of single-player content.
Otherwise, the game runs fine on Fedora/Nobara and Garuda.
Company of Heroes 2 - Proton
Like the games previously mentioned, Company of Heroes 2 remains still a very relevant and very popular game today. This hasn't changed due to the launch of new Relic titles Dawn of War 3 nor Age of Empires 4, and it is very unlikely to change after the launch of Company of Heroes 3.
Now, CoH2 is an example where perhaps the opposite of TWW2 applies. Some knowledge of the game and a quick glance at its ProtonDB page will reveal that most players go for multiplayer mode and, since cross-platform play is not working, they run the Proton version.
This was made worse by the fact that the game must be run on Steam Linux runtime anyway in some distros, already with worse performance than Proton, and performance is what I was here to measure.
Taking all this into consideration, I will create a custom profile local/coh2_proton. Again, this is a copy-paste from original test profile, but the game is running under Proton.
The main difference to consider is that the results will now show up under pfx directory created by Proton:
I was not able to find an xml file for preferences, other than the leftover one in Feral directory. So the test preferences are configured manually to be all maxed, as I normally use these presets in the native test profile.
Other benchmarks
Unigine Superposition and GCC compilation performance are not that relevant per se, but these two happen to get me the highest values in GPU and CPU temperatures respectively, so I am mainly interested in temperature values in this case.
This experiment is motivated by the appearance of the Nobara project, which is a distribution targeted for games, and the fact that I have been slacking off in upgrading my Ubuntu 21 installation.
This system has a Ryzen 7 5800X/Radeon RX 6800. As one does, I just thought the natural course of action, when presented with the big Ubuntu 21 to 22 update, would be to compare performance between the two, and then install Nobara and Garuda from scratch and compare results with those too.
I planned to test Garuda last, since I have tried it before, and I know I'm most likely to stick to it for a while.
The "performance of the distro" is to be measured "off-the-shelf", meaning without any tinkering, no custom kernels unless they come with the distro.
The "performance in games" is meant to be realistic. This is why I will try to include rather popular games like Dota2 or a custom-made profile for Final Fantasy XIV.
Here are what I found the most significant results.
2022_07_26_games_distro_comparison.png
Like I said, I put Garuda last because I expected to stay on Garuda. The distro seemed very consistent the last time around: I made a similar comparison on January. Back then, I tested Clear Linux out of curiosity as well, but it's rather complicated since some games won't run and one must use the Flatpak Steam. The performance was not so different from other distros, in any case.
Not all the results were uploaded, there is such thing as too many results in PTS:
Code:
Would you like to upload the results to OpenBenchmarking.org (y/n): y Would you like to attach the system logs (lspci, dmesg, lsusb, etc) to the test result (y/n): y [PROBLEM] Unhandled Exception Results Failed To Upload.
Dota 2 major 2022
My first setback came when realizing that my Dota2 installation seems not to be able to load the demo (dota2-pts-1971360796.dem) for the old PTS test profile pts/dota2. Demo was running fine last year, so this must have happened at some point last year, with some of the game updates.
If I use a different demo, it won't be the same results anymore, so I may as well create a new test profile. This new profile will be basically same as pts/dota2 but running a new demo, which is compatible.
The name of the test profile will be "local/dota_major_2022".
The demo used will be "6582379623.dem", Game 4 of the finals of Dota 2 Major in Stockholm, 24 May 2022, 19 minutes in, 6 minutes of gameplay.
Can be run manually this way:
steam -applaunch 570 +con_logfile <logfile>.log +timedemoquit replays/658237962 3 +demo_quitafterplayback 1 +cl_showfps 2 +fps_max 0 -nosound -noassert -console -fullscreen +timedemo_start 60000 +timedemo_end 66000 -autoconfig_level 3 -testscript_inline "Test_WaitForCheckPoint DemoPlaybackFinished\; quit"
Total War Warhammer 2
My second setback came when TWW2 was crashing, after doing the update to Ubuntu 22 (segmentation fault). Steam is unable to run the game either. Since both the test profile and the game were running fine on Ubuntu 21, and I didn't change anything from the configuration, I count that run as a "failed".
I am sure that probably the Proton version may be working for this game, but it is reasonable to expect players to be running the Linux native version, as this game has a lot of single-player content.
Otherwise, the game runs fine on Fedora/Nobara and Garuda.
Company of Heroes 2 - Proton
Like the games previously mentioned, Company of Heroes 2 remains still a very relevant and very popular game today. This hasn't changed due to the launch of new Relic titles Dawn of War 3 nor Age of Empires 4, and it is very unlikely to change after the launch of Company of Heroes 3.
Now, CoH2 is an example where perhaps the opposite of TWW2 applies. Some knowledge of the game and a quick glance at its ProtonDB page will reveal that most players go for multiplayer mode and, since cross-platform play is not working, they run the Proton version.
This was made worse by the fact that the game must be run on Steam Linux runtime anyway in some distros, already with worse performance than Proton, and performance is what I was here to measure.
Taking all this into consideration, I will create a custom profile local/coh2_proton. Again, this is a copy-paste from original test profile, but the game is running under Proton.
The main difference to consider is that the results will now show up under pfx directory created by Proton:
Code:
pfx/drive_c/users/steamuser/Documents/My Games/Company of Heroes 2/LogFiles/pts.csv
Other benchmarks
Unigine Superposition and GCC compilation performance are not that relevant per se, but these two happen to get me the highest values in GPU and CPU temperatures respectively, so I am mainly interested in temperature values in this case.