Linux vs Windows 11 Comparison: 20 AAA Games Benchmarked [5K]
ฝัง
- เผยแพร่เมื่อ 22 พ.ค. 2022
- - I ADDED EVEN MORE EXPLANATIONS BELOW!-
20 in-game benchmarks (21, if we count Metro Exodus and Metro Exodus Enhanced Edition as different games), comparison graphics and overall results between Windows 11 and Manjaro Linux.
-- Pre-Description --
Please understand that the purpose of this video is to make a driver/OS comparison between Windows and Linux. This is why everything is explicitly set to ultra/extreme and this is also why there are no modifications in either systems. I am aware that there are modified kernels, extreme settings for Linux as well as stripping/debloating scripts for Windows which disabled almost everything. But with the arriving of Steam Deck, people already established that they want the most "plug and play" experience while still staying in control with their systems.
-Timestamps-
0:00 intro
1:03 Dirt Rally (Steam with Proton GE 7.17)
1:41 GTA V (Steam with Proton GE 7.17)
2:17 Bioshock Infinite (Steam with Proton GE 7.17)
2:55 Chernobylite (Heroic Launcher Wine GE 7.11)
3:52 Batman AK (Heroic Launcher Wine GE 7.11)
4:21 Metro Exodus (Heroic Launcher Wine GE 7.11)
4:52 Metro Exodus EE
5:19 World War Z (Heroic Launcher Wine GE 7.11)
6:06 Cyberpunk 2077 (Heroic Launcher Wine GE 7.11)
6:41 Horizon Zero Dawn (Heroic Launcher Wine GE 7.11)
7:14 Shadow of the Tomb Raider (Heroic Launcher Wine GE 7.11)
7:53 Tomb Raider (Heroic Launcher Wine GE 7.11)
8:11 Crysis Remastered (Heroic Launcher Wine GE 7.11)
8:43 The Division (Steam with Proton Experimental)
9:23 AC Odyssey (Lutris Wine GE 7.11)
9:48 AC Valhalla (Lutris Wine GE 7.11)
10:18 Forza Horizon 4 (Steam with Proton Experimental)
10:48 RDR 2 (Heroic Launcher Wine GE 7.11)
11:17 Far Cry 5 (Heroic Launcher Wine GE 7.11)
11:43 Borderlands 3 (Heroic Launcher Wine GE 7.11)
12:04 ME: Shadow of War (Steam with Proton Experimental)
12:26 4K results
12:52 1440p Results
13:10 1080p Results
-- Description --
I wanted to make an "apples to apples comparison" between Linux and Windows for a long time. I used the same computer for both Windows and Linux settings. I first wiped all the drives and installed Windows, after completing the benchmarks I wiped again, installed Manjaro Linux this time.
All the games were installed on the same SSD, one by one. The system is as follows:
CPU: AMD Ryzen 7 3800X
GPU: AMD Radeon 6700XT
MB: MSi B450 Pro 2 Max
RAM: 3600MHz DDR4 16GB dual sticks
SSD: Samsung Evo 860 512GB
I used Manjaro KDE with X11 + Gamescope on the Linux side, and regular Windows 11 with all the possible updates. All have the latest GPU drivers installed. Both systems are set to performance mode.
Linux kernel is 5.16
I used Wine GE 7.11 and Proton GE 7.17 but some games refused to run with them, for those games I used Lutris's own Wine and Steam's own Proton Experimental.
-- The funny story I mentioned in the video --
Apparently Linux is much more power-preserving than Windows. My GPU fans were spinning so slowly, it reached to 115C in the junction many times. So, it throttled to 500MHz frequently. Even than the fan speed was barely around 50%. I set a manual fan curve with CoreCtrl to a fixed 75% value. Never had any throttling anymore.
Another problem I had was about the CPU. Some games use only 2 or 4 cores at maximum. So the CPU thinks it is not a big load and slows down to 1.9GHz from 4.5GHz. This caused massive fluctuations in the results, also reduced the average numbers to even lower than Windows counterparts. So, I set the CPU to performance mode from CoreCtrl and re-done all the tests. After completing Linux tests, I installed Windows back again and re-done those tests again, with "High performance" plan this time. This is how I ensured everything were equal on the hardware side and this was a true apples-to-apples comparison.
-- Side note about Min/Max 5% --
Some may wonder what Max 5% and Min 5% mean. Imagine it this way: a benchmark has completed and a total of 100 frames were drawn. Instead of finding the fastest and the slowest one, I take the fastest 5 (hence 5%) and the slowest 5 frames. The average of the fastest 5 frames is Max 5% and the average of the slowest 5 frames is Min 5%.
I sincerely believe the fastest or slowest frames alone can be misleading. Imagine them in a 1 hour gameplay session at an average of 60FPS. That is roughly 216000 frames drawn on the screen. Taking only one of them and claiming it the "fastest" or the "slowest" can not be a performance indicator. However, taking the slowest/fastest 5% of the all frames and giving their average tells a story. If I tell you your Min 5% is 48 FPS, this means you spent approx. 3 minutes (5% of an hour) playing your game at around 48FPS. - เกม
Incredible work. Thank you for doing this!
Thank you! It must have been a lot of work, so the biggest appreciations :)
I don't usually write comments, but the video is really well done. All information is also well noted in the video description
Thank you so much for this comparison, this represents so much work!
THANKS YOU
you've done a great work for the community, it deserves my share !
You got my sub bro, thank you for all the effort that you put into making this vid...very well done
Great video. It was obviously a lot of work to do so thanks for that.
Good insights, thank you for sharing!
Great video! Thanks for all the work you put in.
Timestamps
0:00 Intro
1:03 Dirt Rally
1:41 GTA V
2:17 Bioshock Infinite
2:55 Chernobylite
3:52 Batman Arkham Knight
4:21 Metro Exodus
4:52 Metro Exodus Enhanced Edition
5:19 World War Z
6:06 Cyberpunk 2077
6:42 Horizon Zero Dawn
7:15 Shadow of the Tomb Raider
7:54 Tomb Raider (2013)
8:12 Crysis Remastered
8:43 The Division
9:23 Assassin's Creed Odyssey
9:48 Assassin's Creed Valhalla
10:19 Forza Horizon 4
10:48 Red Dead Redemption 2
11:17 Far Cry 5
11:44 Borderlands 3
12:05 Middle Earth: Shadow of War
12:25 Summary
13:48 Outro
I have them with greater detail in the description section. However, for some reason, the video segments are not visible in the timeline.
Wow! Great video!
That's amazing. I never thought there would be such a difference. Thanks for the video
Metro Exodus has a native Linux version, sold on Steam, with working ray tracing on NVIDIA.
Also, ray tracing simple isn’t implemented in VKD3D yet, so it will not work no matter what. Only if the game uses Vulkan, like DOOM Eternal
@@IshayuG it does now! :D
Solid video. Most needed too.
great video man
Everytime I see these comparisons I'm amazed by how back and forth it manages to be. The DXVK, VKD3D, and Wine devs have brought Linux gaming farther than most people ever believed it could come.
Also, it's especially interesting how quite a few games generally perform better on Linux but then it falls behind hard in 4k. I wonder what causes that?
Edit: I typed that while watching and now later in the video I'm seeing games that perform far better in Linux than Windows at 4k. This is some weird inconsistency.
Most of the inconsistencies are because of a simple truth: Vulkan is a Global solution, whereas DirectX is a business one! It is not only about Microsoft, with each DirectX version I have watched AMD and Nvidia adopting new approaches, helping creating new DirectX features and with each new future one introduces, the other fell behind.
For example, Microsoft introduced DirectX 11 with certain CPU limits and Nvidia circumvented them. This lead to CPU bottlenecks for that era and Nvidia got rid of them by changing their GPU design and created a software scheduler for their cards, instead of proposing a more general solution. Then, in the driver level they made DirectX 11 calls multithreaded and solved the CPU problem. This was a forceful update and they forced all the studios working with them to surpass this limit. That way many games using DirectX 11 worked terrible on AMD cards and better on Nvidia cards, regardless of the GPUs potentials. AMD proposed their own solution after a short while but the damage was done. DirectX 11 advanced with two different style and one worked better for one GPU vendor and worse for the other and a unified solution never came. Later, AMD helped with DirectX 12 and they got the lead. But this time, for some unknown but guessable reason, many games didn't want to switch to DirectX 12 for a very long time. When Nvidia introduced their new Ray Tracing tech, everyone jumped to the DirectX 12 ship.
Such stories goes as back as DirectX 8 if I remember correctly. So, DirectX is not just being developed, it is being re-shaped over and over again for the advantage of one side and even for the disadvantage of the other side. Vulkan, on the other hand, is being developed for perfection without being toyed by business politics. So, Linux devs definitely have the right arsenal on their side!
@@gamerhobbit i think people didn't switch to directx 12 for a long time because of the fact that dx12 is just much more lower level and would take longer to get working right. Hardware compatibility on directx 12 isn't great either
Great video mate
Loved this video man! Very well done hope you do more it is great to see Linux getting attention from gamers! I have a similar setup to you, and use Linux Fedora 36 KDE and love it! I use Steam and the Heroic Games Launcher a lot :) I'm still a bit of a noob at the Linux side of computing but have been a Windows user since Windows Vista and I have to say I really love how much customisation can be done with Linux :D
Love this comparison! Thank you for showing off Linux! I'm a new user and have switched from windows because gaming is actually viable on Linux now!
It was very useful, thanks a lot.
Extremely polished, in-depth and helpful video.
You are extremely underrated.
I'm leaving a well deserved subscription
Great Video Brother 💓
Great job :)
What an amazing video, I greatly appreciate the effort put in!
I've just bought a new PC and considering not even installing Windows.. But sadly some games I play such as Fifa and Warzone aren't supported :(
Thankfully every year seems to be getting better for Linux, and with Steam deck being as popular as it is, I think developers will put more effort into adding support for it!
Really nice professional comparison, love it!
One can decide for theirselves, I would personally not bother switching my OS for these amounts of gains/losses. But then again, my resolution seems to favor my current OS so maybe that's a bit biased.
Is it 4k 120Hz with and AMD GPU? :)
@@gamerhobbit Ultrawide, 3440x1440 with an RTX card. I forget whether you mentioned anything about amd vs nvidia in the video since I watched it a while back.
If 1440p is bacically equal on average and 4K is better on windows then I'm somewhere in between.
@@ZenTunE- I believe it is more closer to 4K territory as ultrawide also triggers higher FOV. This video is about AMD GPUs but it is showing its age, now. Since then, AMD GPUs have much better drivers which mostly put them on equal or better positions and they are also Ray Tracing capable now, even though RT performance is visibly lower than Windows. On Nvidia's side, everything should be equal to Windows as long as you do NOT want to use more than 1 screen while gaming.
@@gamerhobbit Oh, alright. I guess that could be true about fov. Most of the time if it is adjustable I use a low/normal number, high fov honestly looks bad on ultrawide, since it's supposed to be cinematic and immersing.
In any case seems I don't need to worry 👍🏻
Thank you very much for tests! Did you use gamemode here?
Lutris enables it automatically so it was enabled for sure. As for Steam, I don't know it uses it automatically.
Beautiful work. I'd be interested in seeing how newer and older hardware fits in here. What if I'm building a Linux machine today or repurposing some 5 year old parts I have lying around?
I have another desktop with a Radeon 5700XT and an old Intel i5 6500. I also have a laptop with Nvidia MX130 and i5 8xxx CPU. They can be count in the 5-6 year old category. So, I'll see what I can do.
Great comparison. Thanks for this contribution to linux gaming!
win won most
Hmmmm, I see you are using the MESA drivers under Linux, but I would also like to know what Kernel you used in Linux and whether you played the games through Proton and DXVK or via their Linux native ports that some games offer (Dirt Rally, Tomb Raider, etc.) Also what Proton/Wine version was used - if so (Proton stable version?, Proton experimental, Proton GE, Wine stable, Wine testing, Wine GE, etc.?)?
Did you start up all games through Steam or use Lutris? Did you apply runner options like "game mode", "DXVK Async" or similar?
Would be nice to have that info included on the results screen.
wolfgang made a video about that, the margins are very small.
Also steam does launch with lutris optimizations out of the box.
@@enzowahlberg4072 Am I supposed to know who Wolfgang is and were to find that specific video?
I knew I forgot some details. I used Steam for some games and Lutris or Heroic for others, depending on the launcher. I'll add all these details to the explanation section.
thank you for posting this, i also apreciate the extra details you added in the description. proton has come along way in the last few years to make linux very viable minus anti cheat issues
Hopefully the anti-cheat issues will be resolved with Steam Deck, as well.
Nice work thks! very well organization and methodology. The only thing i would add is run each test 2 or 3 times and calculate the mean... and the total number of frames.
Thanks. I did all the tests 4 times and used the last 3 of them (the first ones may be different due to cold starts and cache building). So, all the numbers in the end are averages of 3 rounds.
Güzel video.
vkd3d-proton ( DX12 ) oyunları için RADV RT env vari yanında ayrıca VKD3D_CONFIG=dxr11 de gerekli olması lazım.
En azından çalıştırmak için, performans anlamında yine sınıfta kalacaktır şu aşamada.
Bu arada Bioshock Infinite native port muydu yoksa Proton ile mi çalışıyordu? Native port kötü olabilir performans bakımından.
Teşekkür ederim. İki environment variable'ı da ekledim ama yine de oyunlar Ray Tracing'i görmedi. Oyunların tamamı Proton ya da Wine ile, bu arada.
I don't think 1080p is "low" resolution. The most common resolution by Steam's survey is till 1080p by far (59%). I'm using Linux since 20years. I don't have time to play games since ~10 years. Actively, ~13years. Back then, directx games was rarely even run on linux, and is started, many was unplayable. Native AAA games was practically non-existent. So seeing this result is an amazing achievement. I guess Valve/Steam is a great influence, catalyst in this. They invested so much to Linux Gaming, I can thank them enough. I would never imagine that Linux will be more popular to games than everyday desktop users, but today, this seems to be the case (of course, Linux is still the most used for servers and devices, such as Android phones, IOT, etc.)
Not all heroes wear capes. Thank you.
that was unexpected :o
we need more of this but I'd also like to see the live frametimes not only some min max avg numbers graphs
I have switched to Nobara Project and Mangohud works with Gamescope there. I'll make sure to include the results in my next video.
How is your cyberpunk 2077 performance so good on linux? What tweaks did you use? Any tips?
No tweaks. Just installed the game and ran through Heroic Launcher.
Did you use DXVK as it's a major implementation of directx to vulkan?
I believe DXVK was always on on every game which did not use Vulkan natively. I am definetely not an expert on Linux and did not set DXVK to on explictly but I %99 sure Steam and Lutris set it for me.
Would be interesting to see a comparison with all the extra performance enhancements enabled on both Windows and Linux. I can help you with the Linux optimisations for the games I own.
That would be great! Thank you for the nice offer. For starting, I am switching my PC back to Linux this weekend and thinking of Fedora or Nobara this time. I have only used Ubuntu and Manjaro so far, and a little worried about getting used to part. Also my USB Wi-Fi card requires a DKMS package which is easily available in the AUR, to function. What do you think? Is it possible to apply the same fixes and gaming tweaks on Fedora/Nobara?
@Kauã Azevedo I haven't enabled DXVK_Async yet. I have read it could get you banned. Is it still a risk?
@Kauã Azevedo Thank you. I have a B450 motherboard. When I enable one, the other is enabled automatically. What does "nggc" do?
What's the link for the benchmarks sheet?
Amazing work! If you want to go even further with the testing, try out one of the gaming optimized kernels (liquorix, zen) and if your equipment supports it maybe the new p-state driver for AMD.
From my testing, Linux gives a more consistent framerate thanks to fewer system processes running in the background (no Windows Defender, Windows Update, etc) and if one's going for the absolute performance you can strip out components to get lighter
Thank you. Would you share with me easy-to-understand guides for those? I can only change kernels through Manjaro's kernel manager and it only offers official kernels.
Liquorix has packages available through the AUR and zen should be on the Manjaro repos (these are the package names: linux-lqx, linux-zen)
I don't know about the AMD p-state driver, but there should be a kernel/tutorial out there for getting it.
Both of these kernels also tune audio latency alongside other things. Liquorix fixed an audio stutter i had with VLC, I'm using it on Debian and it flies.
you can turn those off on win
its only gonna get better from here on out :)
Do you think that using Wayland ( I guess is better to say XWayland ) will give better performance?
I don't think so. Wayland is not about performance increase, it is more focused on solving the age-long problems which have been piled up with old X11. And it somehow brings its own problems (like Nvidia refusing to fully support it or screen sharing on Discord) while solving them :)
Great
doing god's work
I have an APU and in every single game I would get like 3x the perf
Later I learned that it was because Linux was changing the acpi power controls which would boost the tdp to 30 watts and windows didn't, but either way, it makes it super easy to get very good battery life on linux with the low power mode and very Good performance when plugged in
but if i will start to use linux but for playing games i will use virtual machine for windows to play games ???
i still mainly use windows for gaming because not all features are supported in the games. Warhammer 40000 Soulstorm and dark cruisade are my old favorite titles, which run absolutely smoothly under windows, but under Proton (Steam) often have massive performance drops when many enemies are to be seen. there are also fewer soundtracks (voices) supported under linux. but through steam has done a lot in this point, which I find very good. there are also enough titles that have been optimized directly for linux. how did you install gta v under linux, are there any instructions?
I know Linux feels like "almost there" when it comes to covering all the expectations in gaming. I personally am waiting for Reshade support on native Vulkan games, like Doom and RDR 2. BTW, I did not do anything special, just installed GTA V through Steam and enabled Proton on the compatibility tab.
@@gamerhobbit I have discovered lutris, with it many games also run under linux .... crazy
eline saglik
you are lucky to have bought a AMD GPU and not a Nvidia GPU
and you could use the ACO Shaders for a bit more peformance on Linux
I thought they were automatically chosen by both Lutris and Steam?
1. completely baseless and false.
2. nvidia drivers are completely fine, people complaining about them are: a) FOSS freaks b) had an incident with nvidia 10 years ago.
IF ANYTHING, nvidia drivers are better performance wise than AMD drivers. I've never had a single issue with my 1070 and neither has 99% of the userbase
how would you do gamescope with heroic's games?
The same way I make Steam to use Gamescope, by adding the command "gamescope -f -e -w 3840 -h 2160 -r 144 --" on the command prefix part. (If game throws errors, I delete -e part)
To be fair, Ray tracing is currently more or less broken on AMD on Linux.
For Ray tracing benchmarks I suggest using nVidia GPUs instead :/
it doesnt work that good on Windows either
yep it also on windows and nvidia :p it destroy performance. 4K is faster than 1080p with raytracing activated.
Just 6 or so months ago there would be a 30% difference in performance in almost all games that are not on vulkan, and now this, awasome, HUGE props to valve. 11:13 btw, in this instance linux wouldn't cut it for the 4k, since it dips to 25 fps, while the lowest on windows is 29.5 and here 12:13 I would much prefer to play on linux, since the 5% low is much better
Thanks. Conclusion from this:
The older the game, the better in runs on Windows.
LOL
Considering AMD and Nvidia are competing each other for per-game driver optimizations as well as all types of upscaling & frame generating algorithms, I can say the newer the games, the worse they run everywhere :)
Where 5k?
As many times as i switch to linux, it'll never be as good as a modified windows version like ReviOS
Linux wins at every CPU bottleneck scenario, a well expected result. Thanks!!
Ty ⁉️
Modding on Linux brings too much headache if it's outside the steam workshop.
Thanks for the work.
09:05
You don't even find it a LITTLE BIT suspicious that 1080p win11 is slower than 1440p win11?
Problem seems to be something else here.
In essence from your benchmarks: Linux seems to be way better in the CPU Limit. Windows is better in the GPU limit.
Still , the differences are a little to big to be trusted out of the box. Why? Thing is. When Linux is so much better in the CPU limit why is the minfps in 4k worse than Win11? Example Metro Exodus (4:42). Because minfps are usually improved with a better CPU. If Linux is better at CPU handling why is the minfps worse with higher avgfps than win11?
This doesn't really make sense. Something is off in some of these benchmarks and thus I can't trust them entirely.
But I will test this when I try out the "Nobara Project" Linux which is intended for gamers.
Thing is of course I have a Intel CPU and nVidia GPU so the results may be very different. But on of my friends got a AMD/AMD PC too. So I can do those tests on two different platforms and maybe I can find the culprit.
Have a nice day!
So sorry for the late reply. Such comments make me wish I had more people around me to discuss about details like these :) Here my answers:
1) I did not find it suspicious because at that time AMD was known to be extremely under-powered on Dx11 titles for long years and they fixed this with a driver update one week after this video went online. As for the 5 FPS difference between 1080p and 1440p, I can not explain this entirely but those are the averages of 3 runs so, I think there may be a slight throttling of CPU when the game was at 1080p.
2) At (4:42), what surprises me is Linux having 5 FPS higher on max FPS section at 4K. The lows can always be about extra Proton layer or a momentarily interaction by the OS. Linux may be lighter than Windows in general but it still runs in the background.
Anyway, both OSes have had newer drivers in the meantime. And I have switched to Nobara Project. AMD has released a Directx 11 fix for Windows Drivers, Mesa 23 is about to come with Ray Tracing support with AMD. I basically should re-do these benchmarks with 2023 conditions :)
Cool as soon they get HDR working as well as it does on windows I'll switch over.
I personally use Reshade and always create my own LUTs for every game I play. I liked HDR as an idea in the beginning, it was supposed to be the next step in pixel standards but companies turned it into a scam. They refused to meet in a standard approach and everyone started implementing their own so-called artistic aspect. Since all the HDR screens and games are far from a standard and they are merely the subjective visions of their companies, I prefer to use my own subjective vision :) As long as game developers are not careless enough to make their games in limited color ranges (I'm looking at you Sekiro!) 8 bits with 4:4:4 color is more than enough.
@@gamerhobbit HDR is a noticeable improvement in some games for me on my monitor. It's been my experience that Reshades HDR doesn't compare. I don't know exactly how it works but it's my understanding that Reshade HDR is not the real thing. As much as I like Linux, I wouldn't use it for my daily driver if I can't get the full benefit of my hardware. I also make my own LUTs with Lightroom for most games.
@@Ryan-vc4lr you are right, for some games HDR looks really exceptional. But for the most, it's just higher contrast + more brightness. As for Reshade's HDR, it has a nice idea in it's code but the result is definitely not what true HDR aims for. Calibrating your screen and making your own LUTs is always the best way to maximize the details and making your games more artistic.
@Linden Reaper That isn't what HDR is actually...
@Linden Reaper It's my understanding that HDR has to do with local dimming, The monitors hardware actually has sections that change the brightness. Without the proper profiles for the monitor the hardware isn't going to do local dimming.
Awesome video. thanks. It really shows that PC Gamers who use ONLY windows are doing themselves a disservice... HOWEVER it also means that LINUX-only gamers are missing out, too. You have to dual-boot to get the best of both worlds right now.
This is what I think as well. Users shouldn't feel the need to choose one, they should choose the right OS for the right game. I believe, there will be more games which would drive more people into Linux world in the future.
I'd like to see this with a newer nvdia GPU like RTX 30+
I haven't been able to make videos for a very long time. Last year I bought a laptop with RTX 3060 so I am really hoping to make a new version of this video as soon as possible :)
No PCIe 4.0? Unfortunate.
5k?
That was before TH-cam has deleted 5K option. Now, there is only 4K.
AMD is clearly ahead of Nvidia in some titles here... I know how some games perform on Nvidia and it's no where near as good as it does in Windows. I'm talking about ~30 FPS less.
Linux is a lot less taxing on the CPU
If you have amd gpu, gaming on linux is fine.
I am about to discover the Intel and Nvidia side of it. Hopefully, it'll work as good!
very interesting that windows or linux wins based on the title
I remember trying linux gaming over two tears ago and it was very poor with most of the my games not working at all. Things look much better and with steam deck this will improve.
it is still poor
But it's getting better every day. When you think about the manpower put into Linux, it is evolving suprisingly faster than Windows. All thanks to Valve and many individuals who work countless hours on Linux.
win is just so awesome , best os especially as you can just install games and dont have to worry with boring timesink testing like on linux
It is because all games are made for Windows in mind. Linux tries to "carry" this to itself with a layer between the OS and the game. But, benchmarking is still a valid work on Windows as well. Because nowadays, games are more buggier and less tested before they are actually released. Plus, there are cases where using Linux helps. Crysis Remastered series is a good example of this with its much higher FPS on Linux than Windows.
linux not worth the time and headache
Some people say the same thing for Windows.
Funny, now Horizon Zero Dawn is just as fast on Linux as on Windows 11, have not used
Windows 11 now almost 2 years, and i love it, no personal stuff being stolen by Microsoft
or BING AI seeing everything you do, Microsoft made it easy for me, to go to Linux.
Great video mate