Steam Deck and Linux officially hit over 12 000 games playable and verified

[SchumannFrequency]

I mean yeah you can cherry pick and say "if you do this and that" but it's kind of dishonest to the average user. Most users linux experience won't be with a stripped down wm, and phoronix was pretty clear that the average user experience lags in their benchmarks, as stated.

I have never understood this that Linux gamers often do not use WM for gaming. I mean dwm or bspwm install is literally 5 or 6 seconds work, and then you can select it at your login manager. I'm not saying you should use it as your daily driver, but if you're going to be gaming for an hour or more, why not just select dwm or bspwm during from your login screen. Often dmenu is already pre-installed and you can start your game immediately after logging into the WM, without having to configure anything. What prevents any gamer from using a WM without significant effort?

Phoronix is not a team of 1,000 specialists. In my opinion, it is only one person who is human and has limited time and not unlimited knowledge of things like everyone else. In his Linux tests, there are many popular technologies that he never tests. Also in his gaming benchmarks I always notice something, which is that he often never tests the most popular games, but sometimes rather fairly obscure games. Let's take a look at seven unusually popular games, something Phoronix keeps forgetting to do:

Dota 2 Benchmark - Linux vs Windows

Dota 2 Benchmark - Linux vs Windows

You can find the results here:https://flightlessmango.com/benchmarks/F2xv7n0gklwBenchmark Info:Resolution 1920x1080Linux Driver: 440.66.17Windows Driver: 451...

www.youtube.com

World of Warcraft BfA Benchmark - DXVK vs VKD3D vs Windows [Nvidia]
https://www.youtube.com/watch?v=V2_LlSaYNUo

Although Phoronix virtually never tests these games, these are the PC games most played in the world, the absolute best in terms of popularity.

 

[R-T-B]

I have never understood this

That mirrors most users feelings towards bspwm.

You aren't going to convince anyone by questioning one site and then linking random youtube benches.

 

[Super Firm Tofu]

That mirrors most users feelings towards bspwm.

You aren't going to convince anyone by questioning one site the authority on linux benchmarking and then linking random youtube benches.

FTFY

 

[Nordic]

FTFY

As good as phoronix is, there test suite is not beyond reproach. Their test suite is dated

 

[Aquinus]

As good as phoronix is, there test suite is not beyond reproach. Their test suite is dated

You could also say that it's consistent and suites that don't change can be used for comparisons using previous test data. That's not necessarily a bad thing. It definitely has grown with regard to the tests and suites that it provides. If it has aged badly in any way, it's the lack of easy to use interface. However Michael is a one man show. He's done a lot all things considered. For people who want to automate testing, it's really not a bad solution.

 

[mechtech]

Definitely a milestone vs 10 years ago.

I said it before, I would pay a Windows 10 Pro OEM price for a stripped down version of windows, make it like LTSC with full control, and even less bloat and sell it as Windows 'gaming' OS

Be nice if steam put forth more of an effort to make their Linux distro a true windows replacement.

 

[Nordic]

Definitely a milestone vs 10 years ago.

I said it before, I would pay a Windows 10 Pro OEM price for a stripped down version of windows, make it like LTSC with full control, and even less bloat and sell it as Windows 'gaming' OS

Be nice if steam put forth more of an effort to make their Linux distro a true windows replacement.

The Steam Deck is a culmination of 10 years of work.
Valve had those failed steam machines years ago that lead to the Steam Deck.
Valve iterated on the Steam Controller with the Steam Deck.
All of Valves hardware has given them ever more experience in manufacturing that lead to the steam deck.
Valve has been investing heavily in Proton and other softwares to make gaming on Linux viable.
Valve has promised to eventually release the same SteamOS on the deck for public use.
Thousands of others work on Linux in general.
The true windows replacement is coming

 

[SchumannFrequency]

If it has aged badly in any way, it's the lack of easy to use interface. However Michael is a one man show. He's done a lot all things considered. For people who want to automate testing, it's really not a bad solution.

Have you tried running the Selenium tests on FreeBSD, for example? I can tell you that you are not going to have a pleasant experience.
The Phoronix tests suite (or a specific part of it) simply does not work on many platforms.

That mirrors most users feelings towards bspwm.

What percentage of Linux users have ever tried bspwm?
It is the second most popular tiling window manager and almost as popular as the number one, which seems to suggest that those who give it a chance like it very much.

You aren't going to convince anyone by questioning one site and then linking random youtube benches.

Those tests are just the opposite of random.
Each of the games I test there is most likely in the top 10 most popular PC games of all time.
I think Minecraft has more players than the sum of the number of players from the games Phoronix is testing.
Which makes the test result of this one game many times more relevant than the average of all Phoronix's gaming results.

Nor is it a small difference when it measures 22%.
For example, the performance difference between the RX 6600 XT and the RX 6650 XT is 3.9% which is much smaller than the difference between windows and Linux in this unmatched popular game

Valve has been investing heavily in Proton and other softwares to make gaming on Linux viable.
Valve has promised to eventually release the same SteamOS on the deck for public use.
Thousands of others work on Linux in general.
The true windows replacement is coming

You now have two decent alternatives.
Apple is also trying to make macOS more convenient for gaming.

 

[Aquinus]

The Phoronix tests suite (or a specific part of it) simply does not work on many platforms.

I understand and there are tests that won't run on my Ubuntu install either. Now answer this and this is very important; what solution is a better option? The answer to this question tells you everything you need to know about why things are still the way that they used to be, aside from Michael being a one man show.

 

[unwind-protect]

I understand and there are tests that won't run on my Ubuntu install either. Now answer this and this is very important; what solution is a better option? The answer to this question tells you everything you need to know about why things are still the way that they used to be, aside from Michael being a one man show.

I just discussed that with other people, too. As imperfect as phoronix' test suite and its runs might be, they beat the alternative - which is nonexistent.

 

[R-T-B]

What percentage of Linux users have ever tried bspwm?

What percentage of linux users want a "tiling window manager" and not a desktop environment? Most do mixed work not just games, so I have a strong feeling the percentage is low.

Those tests are just the opposite of random.

Your methodology is described somewhere I assume? You aren't just "running some games" and treating that as a scientific bench?

 

[SchumannFrequency]

Your methodology is described somewhere I assume? You aren't just "running some games" and treating that as a scientific bench?

Instead of asking critical questions (that don't actually add any new info) you could have long since tested the games yourself under the same conditions and see if the results are correct.

Since you don't seem to have tested anything yourself we can conclude that your statements are based on nothing.

ComputerBase (one of the most reliable sources out there) did a comparison between gaming performance on windows11 and Linux less than two weeks ago:

Linux-Gaming: Distributionen im Vergleich: Testergebnisse und Fazit

Linux-Gaming: Testergebnisse und Fazit / Spieleauswahl und Grafikeinstellungen / Benchmarks: Dreimal Linux, einmal Windows / Fazit

www.computerbase.de

According to their results, Nobara is 6% faster in average FPS values than windows11.

 

[R-T-B]

Instead of asking critical questions (that don't actually add any new info)

This is ESSENTIAL info my man.

Since you don't seem to have tested anything yourself we can conclude that your statements are based on nothing.

I don't have the time to do a proper bench, got me. Until you prove you can I'll defer to Phoronix.

(Keep in mind, I'm not "out to get you," I started this debate THINKING the amdgpu driver was faster).

 

[SchumannFrequency]

I don't have the time to do a proper bench, got me.

I once measured the gaming performance of FreeBSD on old hardware and that performance was very good using bspwm.
What I've also noticed is that the popular MATE desktop (with compositing enabled) performs noticeably lower than bspwm + Compton both in synthetic tests and in real games.
I have discussed these major performance differences in detail here: https://forums.freebsd.org/threads/mate-performs-remarkably-poorly-in-benchmarks.88108/

Therefore, if you play games (with a compositor), I recommend a window manager rather than a full desktop environment.
To be perfectly honest, I could even determine which window managers were faster than other window managers by testing for a long time, and taking average values.

If you are a noob, dwm combined with dmenu is probably the easiest option.
dwm also performs very well, though slightly slower than bspwm. The window managers with the highest performance: spectrwm, evilwm, fluxbox, openbox, PeKWM, bspwm, ratpoison, StumpWM, 2bwm, FrankenWM and xmonad.
These are the window managers that are the most optimal in terms of gaming performance.

Until you prove you can I'll defer to Phoronix.

What I'm trying to teach you is that Phoronix percentages are often unreliable for making general conclusions.
He measures mostly, >90% of the time on very modern hardware. If you then go and do the same measurements on hardware that is six years old, or older, you very often come up with totally different results than Phoronix. Furthermore, it happens frequently that he does not measure things correctly.

ComputerBase and FlightlessMango are of similar reliability to Phoronix. ComputerBase says that in their test, Nobara was the fastest in average FPS values. But what ComputerBase also says is that Arch Linux scored best in the 1% Low values, and that in their test Arch Linux was the best operating system for gaming, better than windows11.

 

[R-T-B]

What I'm trying to teach you is that Phoronix percentages are often unreliable for making general conclusions.

Their methodology however is reproducible and consistent, which should be number one requirement for any benches general usability/comparability. I guess what I'm trying to get you to answer is: is your bench ALSO directly reproducible? If so please describe your runs.

I unfortunately tired of waiting for your answer a while ago, but here's to one more try.

 

[SchumannFrequency]

Since Arch Linux is too "unfriendly" for many beginners, ComputerBase recommends using EndeavourOS for gaming, since it should achieve the same performance as Arch Linux.

I guess what I'm trying to get you to answer is: is your bench ALSO directly reproducible? If so please describe your runs.

I see no reason why my benchmarks would be more unreliable, and of course you could reproduce them if you were using (exactly) the same hardware and software.
I repeated them many times and then took the average values to be 100% sure. And every test I did always told the same story, and I did many different tests.

 

[AusWolf]

If Windows 12 continues the evolution that 10 started and 11 perfected, that is, becoming a gigantic, always online spyware, I might give Linux gaming a go as well.

 

[R-T-B]

I see no reason why my benchmarks would be more unreliable, and of course you could reproduce them if you were using (exactly) the same hardware and software.

Different levels, different scenes, different fps. It must be consistent to mean much of anything in a comparison.

 

[SchumannFrequency]

Different levels, different scenes, different fps. It must be consistent to mean much of anything in a comparison.

This is what I tried to tell you before. All my testing has been completely consistent.
If you were to read the two pages of content from my link you are going to notice the following things:

1. MotionMark 1.2
2. In Xonotic I see 8% performance difference between MATE and dwm. (this is about the standardized benchmark that is part of this game and that you can call up via the in-game console)
3. glmark2
4. OpenArena, the 'TIMEDEMO 1' benchmark. The resolution I chose for all tests: 1024x768 in fullscreen mode. The graphical settings of the game: Very High Quality.
5. CS:GO benchmark (https://steamcommunity.com/sharedfiles/filedetails/?id=500334237)
6. World of Padman, the 'TIMEDEMO 1' benchmark BIGBALLOON-DEMO.DM_68. The resolution I chose for all tests: 1920x1080 in fullscreen mode. The graphical settings of the game were the highest settings.
7. Unigine Tropics, with all settings on the lowest setting.

What I hope you now understand is that my method is exactly as reproducible and consistent as Phoronix's.

LXQt often uses picom, which I think should be as fast as compton. LXQt should normally get higher FPS than MATE in games when using a compositor.
MATE performs very underwhelmingly in games, or at least on FreeBSD with my old Nvidia card.

 

[R-T-B]

This is what I tried to tell you before. All my testing has been completely consistent.
If you were to read the two pages of content from my link you are going to notice the following things:

1. MotionMark 1.2
2. In Xonotic I see 8% performance difference between MATE and dwm. (this is about the standardized benchmark that is part of this game and that you can call up via the in-game console)
3. glmark2
4. OpenArena, the 'TIMEDEMO 1' benchmark. The resolution I chose for all tests: 1024x768 in fullscreen mode. The graphical settings of the game: Very High Quality.
5. CS:GO benchmark (https://steamcommunity.com/sharedfiles/filedetails/?id=500334237)
6. World of Padman, the 'TIMEDEMO 1' benchmark BIGBALLOON-DEMO.DM_68. The resolution I chose for all tests: 1920x1080 in fullscreen mode. The graphical settings of the game were the highest settings.
7. Unigine Tropics, with all settings on the lowest setting.

What I hope you now understand is that my method is exactly as reproducible and consistent as Phoronix's.

LXQt often uses picom, which I think should be as fast as compton. LXQt should normally get higher FPS than MATE in games when using a compositor.
MATE performs very underwhelmingly in games, or at least on FreeBSD with my old Nvidia card.

Ok, I appreciate that. I must have missed your link sorry (I don't have a lot of time to dedicate to this to be completely honest).

 

[Nordic]

If you were to read the two pages of content from my link you are going to notice the following things:

I hate to be this guy but with today's attention spans on the internet, it is unreasonable to expect someone to follow a link to another website and read two pages of content. You might expect more from an enthusiast forum like this one but it is still the internet. I did not click that link and it is unlikely most will. You probably would have had better responses here if you copy and pasted the relevant material.

Having only now clicked your link, it is two pages of discussion on another forum. That is absolutely unreasonable to expect us to go read that.

I more or less agree with what you are saying. Linux does perform better than Windows in many circumstances. The way you are presenting your information is at best insufficient and at worst unnecessarily hostile.

@SchumannFrequency I am curious about your methodology for performance testing on linux. For Windows I use CapFrameX to do before and after testing. I plan to do some testing on Linux too. Where is your methodology easiest to read?

 

[SchumannFrequency]

Linux does perform better than Windows in many circumstances.

I plan to do some testing on Linux too. Where is your methodology easiest to read?

There are many games that have built-in benchmarks. Some examples include Xonotic, World of Padman, OpenArena, CS:GO, Tomb Raider 2013, etc. As long as you use the same graphics settings for the tests, the results are directly comparable. Although of course built-in benchmarks are not always reliable to compare e.g. AMD with Nvidia GPU performance, as has been discovered several times in the past. But for comparing the performance of operating systems or kernels they are fine.

There are multiple kernel options for Linux.

• Linux 6.6.7
• Liquorix 6.6.7
• XanMod 6.6.7
• TKG-CFS 6.6.7
• Zen-Kernel 6.6.7

Of these five kernels, Liquorix 6.6.7 is faster than all the others in just about every game. This is also something to consider. My advice is to use bspwm + compton/picom in combination with e.g. the Liquorix kernel (or in combination with Clear Linux). For the filesystem I recommend XFS based on my measurements of game load times.