LINUX vs WINDOWS: complete performance test!
Vložit
- čas přidán 22. 05. 2024
- Download Safing's Portmaster and take control of your network traffic: safing.io
Grab a brand new laptop or desktop running Linux: www.tuxedocomputers.com/en#
👏 SUPPORT THE CHANNEL:
Get access to a weekly podcast, vote on the next topics I cover, and get your name in the credits:
CZcams: www.youtube.com/@thelinuxexp/...
Patreon: / thelinuxexperiment
Or, you can donate whatever you want: paypal.me/thelinuxexp
👕 GET TLE MERCH
Support the channel AND get cool new gear: the-linux-experiment.creator-...
🎙️ LINUX AND OPEN SOURCE NEWS PODCAST:
Listen to the latest Linux and open source news, with more in depth coverage, and ad-free! podcast.thelinuxexp.com
🏆 FOLLOW ME ELSEWHERE:
Website: thelinuxexp.com
Mastodon: mastodon.social/web/@thelinuxEXP
Pixelfed: pixelfed.social/TLENick
Twitter : / thelinuxexp
PeerTube: tilvids.com/c/thelinuxexperim...
This video is distributed under the Creative Commons Share Alike license.
#linux #windows #benchmark
00:00 Intro
00:42 Sponsor: monitor and secure your internet connection
01:46 The benchmark setup
03:04 Resource Usage
06:52 Internet Speeds
08:32 CPU Benchmark
09:17 GPU + Gaming Benchmarks
13:35 Battery Life
14:34 Conclusions?
16:25 Sponsor: Get a device that runs Linux perfectly
17:19 Support the channel
I'll use a Stellaris 15, with an i7 12700H, 16GB of RAM, 1TB of SSD + an RTX 3060. It has a 1440p screen that I'll run at 60hz.
In terms of operating systems, I'll run default windows 11 with all of its updates and the nvidia game ready drivers, and the latest Ubuntu 22.10 with the proprietary nvidia drivers, on X11, and no other specific customization to the default.
Ubuntu uses up 25 gigs. On Windows, the install uses 29 Gigs.
Ubuntu's System Monitor reports 1.9 gigs of RAM used after a cold boot. On Windows, the system reports 3.3 Gigs of RAM being used, after a fresh boot.
Although it does report 4.7 Gigs of RAM being cached, which should represent that preemptive memory usage, while Ubuntu reported about 3.9 gigs of cached RAM. If we add used RAM and cached RAM, Ubuntu uses around 5.8 Gigs, where Windows uses 8 Gigs.
I used Kdiskmark on Ubuntu, which reported read speeds of 3360 MBps, and write speeds of 2706. On Windows, I used Crystal Disk Mark, and got read speeds of 3505 MBps, and write speeds of 2782 MBps.
I ran a speedtest both in Wifi and plugged in using ethernet.
Using Wifi, Windows 11 got a download speed of 108 Mbps and an upload speed of 196 Mbps.
On Ubuntu, the same speedtest over wifi got 154 Mbps for download speed, and 201 for upload.
Using ethernet, Windows got 512 Mbps down, and 483 Mbps up. Ubuntu, plugged with the same cable, got 508 Mbps down, and only 123 Mbps up.
On to the CPU benchmarks, namely Geekbench 6. On Windows, it gave me a score of 2216 in single core, and 10805 in multi core.
On Ubuntu, the same geekbench 6 got scores of 2494 in single core, and 10138 in multi core.
We're going to run Unigine heaven on both operating systems, at High settings, in fullscreen, at the native 2560x1440p resolution, with tesselation and anti aliasing disabled.
On Ubuntu, I got 114 FPS average, with a score of 2878, minimum FPS was around 20, and max at around 202.
On WIndows, running the same benchmark using openGL with the exact same settings, I got 105 FPS average, with a score of 2665, 7.5% lower, and a minimum FPS of 14, and max of 219.
Running the same benchmark using DirectX 11 on Windows resulted in better performance, with an average of 139FPS, and a score of 3513, but minimum FPS dropping even lower at 10, and much higher max FPS as well, at 283.
For Shadow of the Tomb raider, running the in game benchmark at high details, at the native resolution, I got 80 FPS on average on Ubuntu. On windows, I got an average of 87 FPS, with more stable frame times.
Horizon doesn't have an in game benchmark, but playing the same sequence of fighting this thunderjaw, with the game running at the native 1440p resolution, at high settings, with an uncapped framerate, I got a little less than 60 FPS on Ubuntu. It mostly stayed at around 55 FPS for the whole fight.
On Windows, using the exact same settings, and playing the same fight, I got more around 65 to 70 FPS. Although for some reason the game didn't have any audio, not that I would have heard it over the fan noise that went into overdrive on Windows.
Now, for battery life, I used Firefox on both operating systems, and played youtube videos in a loop until the battery died, over wifi, at mid brightness, in battery saving mode, with nothing else running in the background and Both devices running in hybrid graphics mode.
On Ubuntu, the laptop lasted for 6 Hours and 52 minutes before it died, a little bit lower than what I got when using it with Fedora when I reviewed it.
On Windows, it only endured for 5 hours and 36 minutes. - Věda a technologie
Download Safing's Portmaster and take control of your network traffic: safing.io
Kind of unfair to test windows 11 even though nobody uses windows 11
@@WohaoG I use it
Hzd does have an in game benchmark
@@DryPaperHammerBro You are not a person
@@DryPaperHammerBro You are an alien hired to corrupt and assassinate all humans
Personally more curious as to how other Linux distributions compare when it comes to performance, wonder if we can get that comparison in the future.
We might!
It is really time consuming tho, he would need like 2 or 3 of the same computer so it's faster. Maybe making 1 video a week about a different OS would be a great way to do it and as weeks advance, we get more and more comparison. But that would be really interesting, maybe we could debunk those riced up distro to show they're maybe not that good
Linux is linux
It is probably more down to the kernel, like the clear linux kernel might work better with intel.
While distros would certainly be interesting, I feel like DEs (and WMs) on a distro that implements them well would be more interesting. The only distro I'm specifically curious about would be a highly Optimized gentoo setup compared to some other more commonly used setups. Idk maybe Ubuntu vs fedora gnome vs arch with a light weight WM like i3 or sway vs gentoo (with idk what setup) vs kde neon vs opensuse tumbleweed with kde. That would be so much work, but would certainly answer a lot of my questions.
I think the real beauty in this is that they were generally so close. The takeaway being that you're going to get generally good performance on either OS. Very cool. :)
Yep. It’s kind of insane that an OS where some drivers are developed by individuals is that close, or surpasses windows
@@TheLinuxEXP Some Linux drivers come from the community, but many are professionally produced, especially if you will find that particular chip in a server. most corporations depend on Linux and need to get the most out of their hardware investment
@hello "UsE LiNuX" are you guys not fighting for freedom? Isn't it that you guys believe in others choice? How the tables turn? Nobody is dictating nobody period
@@TheLinuxEXP Also, your windows test was on a fresh install. If it was done on a 1 year old windows installation, it would probably be slower. (or if you used a minimal meta-distro like Puppy, it would bully windows even on a fresh install)
@@balala4641did you test that?
I'm pretty shocked that battery life was better on Ubuntu than on Windows. I remember battery life usually being worse on Linux though the last time I've used Linux on laptops is around 5 years ago (I built a rig). I guess the power management has greatly improved throughout the years and as a current Zen 3 user, I hope the EPP driver brings more improvements in Kernel 6.3.
Yeah, it was surprising to me as well!
The battery is better on Most GNOME based distributions and often supports the ability to open with dGPU. I thought Windows gives less battery for me because it spins my Secondary Drive ( HDD ) with that Windows Defender. I can hear the fans crying randomly.
Windows 11 has absolutely terrible battery life on most laptops. I tried Lenovo Yoga i7 for around 1600 USD and the laptop had around 4 hours of battery life while doing relatively light tasks such as programming in VS Code
I have a ThinkPad from work with a GPU that I installed Ubuntu 20.04 on and the battery life is atrocious (~2 hours)
I think Pop_OS goes even further, by default, since they sell laptops and they setup their OS based on that. I remember I used to get over 4 hours on my HP an050nr laptop and that had a small battery (2400 mAh if I am not mistaken)
To me, the advantages of Linux outweight any possible performance differences. Linux doesn't decay over time, so I don't have to reinstall it periodically. Migrating to different hardware is generally much easier. I can move the entire OS, while with Windows you typcially have to do a new install, then install the applications on it.
every os decays over time. linux does too. they might be better than windows but that doesnt mean linux is immune to the issue.
Windows doesn't degrade over time in current versions. This is an outdated myth. This might have been an issue in Windows XP times. As long as you don't install shady stuff and don't go too experimental on the Windows Registry - you will be fine. The migration point is a definitely a big plus for Linux. Just "dd"ing your OS within few minutes is awesome - this came in handy when I recently swapped in a new SSD into my SteamDeck.
When was the last time you had to do a reinstall of windows because of that "decay" of yours?
You can migrate windows 8 and above without issues but the decay and age rot (and the lovely weight gain in ram usage when it have been running for a few days 24/7) will be there. It is a good advice to reinstall while you are moving so you can remove the age rot and the decay.. and a few bad google searches or installs O.o
Windows in general need to be reinstalled after 1 year of use, using the same system longer than that is a bad idea.
Linux can be used longer than a year but as a bald tech dude with glasses i would recommend that you reinstall once a year anyway. It depends on how important the computer is or what its used for.
I used to run a server and that system was old, maybe 2 years old but it was easier and better to keep the system as it was compared to reinstall a headless server.. Didnt notice any performance drops or issues and was able to log in and out as i pleased withouth issues or slowdowns..
There is a few programs that show a update popup but its not required to do it and linux wont update unless its allowed.
linux have no backdoor installs or other "hey look at this, install this" popups.. I usally notice that my computer have been running for 7-10 days 24/7 when using linux. My record is 6 months+ and i got it with a server i was running, worked great and i was able to log in within seconds and do stuff whenever i needed ;)
My windows cant be running for more than 5 minutes before i get a install from the backdoor and using it for more than 4 days in a row is a nightmare.. the ram goes up and up.. even if i try and close programs.
@@lokelaufeyson9931 I had to reinstall windows last time like twenty years ago. Experience might vary I guess. But the current one is running well 5+ years. And it is just another one in the line of several other notebooks, which also ran well (I mostly had to replace them because of motherboard issues - Windows experience for me has been generally good since like XP service pack 2). Not saying that windows does not have other issues. But stability issues or slow downs I do not notice.
Phoronix maintainer has already done some amazing work comparing Windows and many Linux distros performance for all kinds of applications. He has a full suite of automated benchmarks based on real use cases, from gaming to HPC applications.
Sqlite also did some benchmarks, to show that SQLite can be faster than accessing the filesystem
Yes, and some times Windows wins :).
Back when I still dual booted, which was around a year ago, I usually got 80-90% of the gaming performance in Linux, that I would in Windows.
But something like 5 years ago that number was closer to 50%, and getting the games running to begin with was a hassle.
Gaming on Linux is such a better experience today.
And new triple AAA titles Iike Atomic Heart run perfectly out of the box at same performance as windows!!
@@AndRei-yc3ti I played Spider-Man: Miles Morales on my Steam Deck, at around PS4 settings, while pulling well under 50 watts from the wall.
I even have a few people from my old place of employment who have never used Linux, who have gotten the Steam Deck and had no idea it ran Linux until I told them.
Your wording implies you are only on Linux. Any advice for someone considering jumping in with both feet and totally abandoning Windows?
@@Queldonus Not as such, as I have been using Linux for nearly 20 years and have kind of forgotten the onboarding process 😝.
The biggest thing though is that distributions (distros) don't really matter all that much, the desktop environment is more important, as that is how you interact with the actual machine.
I run KDE (desktop environment), which is also what the Steam Deck uses, mostly because it is very easy to customize.
For distro I run Manjaro, but for most of my time in Linux I used Debian.
@@Queldonus Get a beginner friendly distro (Ubuntu or Manjaro for example) and if you can choose your desktop environment, KDE is the smoothest transition when coming from Windows. I personally jumped from Windows to only Linux 2.5 years ago and went with Manjaro + KDE Plasma. Haven't looked back since, but I changed my distro of choice to EndeavourOS.
Don't be afraid to distro hop a lot in the beginning. Sometimes it takes a while to find one that really clicks for you. I had been trying to go full Linux mode on Ubuntu for a while but Ubuntu never clicked for me. Manjaro really ticked all my boxes, and the Arch User Repository is a big reason for it.
But the best way to switch is to just do it. Just back up your files on an external drive or something, wipe your disk and install some Linux distro on it. No more dual booting and force yourself to use Linux, and you'll find that every problem you Google has a solution ;)
I appreciate you essentially dualbooting on the same laptop to get optimal comparison. Other comparisons of os tend to use different hardware and can be at least slightly up in the air how accurate tests were. But here, there's no debating same tasks on the same device with same hardware. The results are solid.
Interesting results.
What generally surprised me were the read/write speeds, since I didn't expect Windows to pop off like that.
Also quite interesting that Ubuntu takes up that much space by default nowadays.
Thanks for the video 😉
Well, do a git clone of a large project with many small files on a well tuned windows machine vs some linux box. The read/write test basically tested mostly the hardware, as many of those tests did. This is quite interesting to see, but there are other situations where the os matter much more.
Executing many programs at once. I think windows will do quite well there.
Handling many small files. Linux would just dominate here in my experience.
Applications that use a lot of RAM constantly starting and stopping in the background and how this affects overall performance over a longer period of time. Can't even start to predict performance differences on that one...
I use alternative OS like arch, mint or any other os that sounds fun to use.. I dont like the UI in ubuntu. Too much "touch pad, tablet or phone based". I can touch my screen at home until i have to cut my nails but it will only give me a really dirty screen with alot of fingerprints..
Specially considering NTFS is utter garbage
Jup, was also surprised by the SSD performance. Funny how much space Ubuntu uses now :O
@@HowToLinux ubuntu have a strange development history,, i think it start and become a bit "too much" now.. being unique is good but being too unique can be a downside aswell..
I'm pleased that you didn't make the tests more complicated than they had to be
One reason the Linux Multi-Core score might have fallen a bit is the newer architecture from Intel using efficiency cores and performance cores. IIRC Intel helped Microsoft with the scheduling on their new processors. I would be willing to bet an AMD processor would have had improved performance - even if a small amount - in both single and multi core performance when going from windows to Linux, probably why the open data from Blender benchmarks have Linux as the dominant OS for AMD processors. Another metric that would be interesting to see would be total system power draw while completing benchmarks.
One note on your point regarding the amount of blender users using AMD is higher in Linux because a higher amount of blender users use AMD cpus because AMD has focused their design architecture on multi-core for longer than intel, whom until recently has focused mainly on single core performance.
Pretty interesting!
Would be good to see how the major distro families fare as well.
In general we as a community need more benchmarking of stuff.
I didnt expect at all linux winning in battery life here. Interesting.
Me neither!
Nick, I think I see a series coming. Benchmarking multiple Linux distributions in head to head competition. Do it bracket style.
That's an excellent comparison of performance... Great hard work for this time consuming video making ! Linux is getting better and better...
Thanks! It took a while to make 😂
Not realy because linux have much better distros than ubuntu, super fast and lightwaight like Arch Linux and this will be a desaster for windows
@@innerfinder100 True, but also remember there are Windows tweaks and lighter Windows ISOs out there that can give you better performance as well. Still, Linux's performance is impressive.
Very nice video! I would have appreciated a written side by side comparison of the various results put on video to better appreciate the differences, side by side
Thank you for taking the time to do this series of tests.
Hey Nick, thanks for doing this! As Steve Burke from GamersNexus frequently points out, *cooling equals performance.* The boosting behavior of modern CPUs and GPUs makes them work harder when they're cooler. I wonder how your results might have changed if you had found a way to ensure both setups were using the same fan speed curves.
I was thinking the same thing. If the temps are at 86, I would be checking whether the GPU is thermal throttling.
I'm actually unsurprised about the battery life and the wifi speed differences. Using less background resources should lead to a longer battery life, and Windows update likes to reserve up to 20% of your bandwidth for itself if you don't go in and customize that setting. I am shocked by the Ethernet upload on Linux though, that's pretty awful and might need looked at.
my guess is that it's a particular problem with the onboard NIC and you won't really get it on other devices.
@@adwaitagnome That's surprising given that it's a device from a manufacturer focused on Linux compatibility. Hopefully the cause gets ironed out regardless of the cause, since that's a pretty bad look for Tuxedo.
@@unlucky1307 It would be a bad look, but Tuxedo isn't the OEM for their laptops. I'm pretty certain they use Clevo for their laptops who I'm pretty sure also designs the motherboards and decides what NIC to use.
Realistically, Tuxedo did as much as they could to make it a good laptop for GNU/Linux. There's a good chance that even other Tuxedo laptops (hell, even different revisions of the same laptop) wouldn't have this particular issue. Nick might have just gotten unlucky here.
To my knowledge, the only GNU/Linux laptop maker who actually manufactures their machines is Starlabs, so they get even more free range to put in good hardware for GNU/Linux than Tuxedo does.
I have been using my Linux desktop pc with Ethernet for more than 4 years, and no difference in upload speed for Internet. Definitely, there was some kind of problem with the machine or some misconfiguration.
Thanks for another great vid. Have you done any tests for battery life with different DEs by any chance??
Cool to see thank you!
As someone who has run Linux on my laptop I've noticed huge battery life improvements with the past few years of kernel updates.
I wonder if some of your performance differences could be due to thermal throttles?
Nice hand-on comparison. Just one note:
Under Linux the cache memory is considered to be *free memory*. While under Windows the cache it is part of the *used memory*.
That's not how free displays memory use here. Linux considers RAM being used as cache as being used but will free it up for other use if needed.
That's not true, cache on Windows is a part of the standby list which isn't included in the total that task manager shows
I was going to say that the multi-threaded performance on Ubuntu may be explained by having a pre-5.18 kernel but 22.10 has a 5.19 kernel. 5.18 is the kernel where "Thread Director" (Intel Hardware Feedback Interface) was integrated. Intel did a lot of work with Microsoft on Windows 11 to prepare it for Alder Lake. They did significantly less work with Linux (except for their own variant of course) and it's still ongoing. A newer kernel _might_ help.
I don't get the difference in upload speeds over ethernet however. It'd be interesting to just do the test over a local ethernet connection: you could do it with rsync set up as a client and server and a set of large files, for example, although that wouldn't be exactly the same test on Windows *unless* you did it on WSL and WSL isn't significantly affecting performance. But that's not exactly the idea is it? It would give you a ballpark. Ideally, there'd be the same app compiled for Windows and Linux, just using a single socket and TCP connection and maxing it out. There are various settings that can affect TCP connections such as buffer size, packet length, how acknowledgements are being handled, window management (transmission window not graphical window), routing, and so on. The network interface should be fairly self-tuning but you never know. There's no reason (that I can think of off the top of my head) why you should get worse transfer speeds over a wired ethernet connection (not that bad anyway) than you're managing with WiFi.
At one point I had a throttling problem with my cable internet connection and it turned out to be a setting I'd made on the router for supposedly prioritizing game and streaming media traffic. It did nothing except throttle the throughput to less than 20% of what it should have been. I can't see why anything like that would affect Linux more than Windows however. There may be differences in terms of routing based on IPv4 vs. IPv6... You'd have to do a full forensic analysis. It's pretty antagonizing!
Yeah he should have used ClearOS as the Linux distro.
Finally! I've been wanting for a big channel to make a new benchmark video comparing these two OSes for a long time now! Thank you!!
Awesome video!! Might be good to add the numbers (such as fps) on the screen when comparing the two operating systems so we can just see the difference at a glance
Thermals: on windows this often relies on the OEM drivers for specific chassis - which makes sense as different laptops with otherwise identical components will have different thermals. That being said, on my Dell XPS I have the same issue: you can tune it to "super noisy fans all the time" or "run so hot you not only throttle the CPU and GPU but also burn my knees" - and there's nothing in between. On Ubuntu it is much saner and even in "performance mode" you only get occasional boosts of full fan.
I have to wonder if there's maybe a bug in the driver for that Ethernet NIC, interesting find!
Yeah, it’s weird, and it has the same result on my desktop using Fedora, which doesn’t have the same Ethernet hardware
Thank you for doing these tests on default installs. That’s a valuable point of reference for someone like me that’s looking to move my gaming system to Linux.
It’s just a matter of if I do it now, or in a couple of years at the end of Win10 support
Very good video, thanks for it! And very interesting and kind of unexpected results on some tests :)
A comparison between Linux Distros/desktops would be quite interesting I think.
I have two 10 year old laptops and battery life was doubled going from Windows to Ubuntu for the one laptop and I got the same result on the other laptop with Mint. So I was surprised to hear that Linux usually does worse. So it really can be very different from device to device I guess. The Mint laptop is actually my mother's machine and since she spends 99% of her time in browser it was a no-brainer to put her on Linux and now I don't have to worry about Windows slowly mutating in size over time. These old machines also respond better with Linux than Windows did, but on modern hardware it's hard to tell a real difference between the 2 in day to day use... at least in my experience ;)
Great video, Nick!
Really appreciate this type of content.
In my opinion the slight performance deficit of Linux does not outweigh all the negatives Windows has in comparison to Linux.
If you love an open and free (in every aspect of the term) OS, which doesn't spy on its users, Linux is for you.
I've made the switch almost 3 years ago now and haven't looked back, Linux is amazing!
Thanks for this excellent video!
Great video, the results are fairly interesting! I would say that many of these tests would vary greatly from one computer to another, due to driver optimizations for each component on each OS. For example, on my previous laptop, I got Geekbench results that were within margin of error between Linux and Windows, but battery life was way worse on Linux. Also, Windows 10 used up only around 2-3GB, whereas both 10 and 11 use 4GB on every new computer I checked, including my new laptop, and like you noticed on yours. In addition, my friend's old laptop with an older Nvidia discrete GPU has very wacky drivers on Linux, whereas newer Nvidia GPUs that I've tested work way smoother.
Awesome Video, could you please add a comparison graph for next time though. I had to keep rewinding the video back to compare scores. Would love to see performance compared using something like Handbrake between these two OS's. Thanks for posting!!
Now this is some quality video from you
Would be incredible if you could test with Wayland
I think this a good start and a very interesting concept. I've been waiting for some reviewers to test performance between Linux and Windows. That said I guess the result are inconclusive partially because it's being tested with crippled hardware to begin with. I'd be much more interested in the results testing with a more powerful desktop computer, with room to stretch its legs. As a side note, including some comparative graphs would make the numbers a lot easier to read and digest.
hi there. just wanted to suggest adding side-by-side comparisons visually at the end of each section if you ever do something similar again. either graphs or text. would make it easier to digest after the commentary. either way, thanks for doing this! i liked the video.
I really miss using Fedora, but it broke after an update for some reason. I'll probably switch back when I get home. Windows definitely had an edge with gaming on my Intel Iris Plus, but then I don't game very often so it's alright.
Same, anything other than a Ubuntu-based distro keeps causing issues now and then. Arch-based Manjaro simply felt like a beta OS to me, everything ready to crash and burn the moment I do something a little 'different'. Yes I'm a Linux noobie, and it's certainly not ideal for a beginner. Currently using Mint and it's working fine enough. But no Freesync is killing it for gaming. I always have to crawl back to Windows for some reason or the other, so it's still my main.
Not to mention, Dolby Atmos supported games are insanely good sounding, so that alone is enough for me to never delete Windows.
I'm most surprised by the ethernet results! My experience is completely different - windows was a bit slower, but both were pretty close to my top internet connection - no variance big enough to matter.
I have tested it on 3 diferent PCs, and on several different networks, in several different cities.
I'd say the problem lies in your specific setup. Very weird. Maybe you're using an incorrect ethernet driver? Maybe swapping from r8168 to r8169 (or back) would help? Most welcome apps for most linux OSes have an option or that
Regarding the GeekBench results - I know you're talking about defaults, but this is easily solved using a ZEN kernel, which is designed for everyday desktops - another easy fix. It should be the default on many distros, really.
I had the same issue on 2 different computers with 2 different distros :/
@@octia2817 people configuring and building their own custom kernels should be the default.
@@1pcfred It will not, do not fool yourself. The majority isn't, doesn't want to nor needs to be technical. And even though it's not very hard, it's very technical. And most people don't even need it. Slap zen if you're using your pc for anything more than basic tasks and you're good to go. People with slower PCs would just get frustrated, because a process that lasts hours would result in no real benefit
Compile your own kernel only if you really need it. Don't claim that everyone should do it, there's no need for that, and it just scares people away.
Also, if I remember correctly, if you remove support for different architectures, hardware upgrades will screw up your install. And who wants to deal with that? Who wants to remember about that? Not most people. And they shouldn't need to.
@@octia2817 you don't screw anything up. You just chroot if all else fails. You can have as many kernels installed as you like too. You can only run one at a time though. Learn how to drive that PC.
Thanks for putting so much effort into this!
I would love to see a "trimmed" test as well where you do your general first install optimizations for both windows and linux and add those to the comparison as well, just to see what the peak of what the average user can get on both windows and linux
Seems like the better gaming performance on windows is (at least partially) due to lower temperatures and thus higher clock speeds - the OSes seem to handle fan management differently. It would make sense to manually set fan speeds to the same level on both OSes for a fair comparison.
Yeah, Windows BLASTED the heating-up components with air while Ubuntu used the fan more conservatively.
Power Management is also a part of what makes an OS & Kernel better. So I think it was fair.
Games perform better on Windows because developers optimize better for Windows. Why would they expend the same amount of effort on a fraction of a percent of their users? Linux gamers isn't even 1% of the market. So you're damned lucky it works at all.
@@1pcfred this is not true. Generally, games running on proton/wine run about as well as on windows, some better and some worse.
@@runef3356 I don't do Wine. The sulphates give me a hangover. I don't do Windows at all either. It's native binaries or nothing at all.
Vanilla stock software settings is what 99% of users do so correct to do in stock settings
Very interesting. I didn't expect the differences in write and network performance.
My OS uses 820MiB of RAM when fully loaded. 😅 I‘m using Hyprland (it‘s a quite heavy Compositor) and I have an authentication agent and some other small tools running. Although these 820MiB include about 30 MiB by Terminator and about 10 MiB by Bashtop.
You're a brave man making this video, got my pop-corn in the microwave ready for the flame war to go down.
a question : when you updated windows 11 did it install "killer intelligence center" or similar software? It limits your internet speed. generally sits in the startup apps. I've personally tested this on Dell G15 5520. The internet speed is almost same.
Have you tried a different user agent while uploading with Linux? Sometimes test-pages seem to throttle bandwidth on their side for random reasons for Linux.
[8:10] forget the speed tests with a browser over the internet. They are not reliable. Maybe the lower upload speed is a setting of the browser. Perhaps it's a driver issue. The tool iperf3 is a good start for network speed tests. You can run an iperf3 server on your local network. I did the test and got over Ethernet the full upload speed, but my upload is only 50 MBit/s.
That is fascinating, thanks for sharing
Nice to know there isn't much difference overall. Linux has way less spyware, trusts me to customise things more, comes with Python installed. And, I think KDE Plasma looks way nicer than Windows 11.
My favourite DEs are Cinnamon (as on Linux Mint 21.1, in dark mode with green accent colour) and Unity (as on Ubuntu Unity 22.04 LTS with theme Yaru-unity-dark, making the accent colour purple).
I took the plunge into PopOS as my gaming PC OS... It was a bit of a struggle at first, but the improvement in the OS and Steam has been exponential in comparison to previous years. Issues like the cursor not locking on to the primary monitor(out of 3), or Halo MCC settings crashing the game, are gone.
What is the linux distribution you are using in the 3:14 of the video? I really liked the look. 😄
Probably PopOS!
Maybe that fan blast also helps a little on windows
I'm sure battery life difference was because of Microsoft making sure you're safe from getting worse ads 😏
Could you also try benchmarking Vulkan somehow?
I’ll try!
Wondering how ubuntus choice of ext4 affects the disk test, would be interesting to see fedora with btrfs
Thanks for another great video!
There are so many variables in a laptop from the specific components and their drivers to the layout inside the laptop and so on, but my main takeaway is, that, apart from the battery, there really isn't much of a difference. This means, in my opinion, that we can compare the systems on other things such as privacy, availability of software and other parameters.
For me, in all cases of personal use, it will be Linux. For work and studies, it will be a choice depending on the software that I need to run, which for now means windows. This just means that I have an older laptop for personal use and programming for my own projects, and one with windows for cases that demand it
For me the main takeaway is encouraging - both systems are pretty much on par with each other, so you can use whatever you like without fear of losing too much in terms of performance. I usually prefer Windows for my everyday work but I love watching your videos :)
I think a great game to test the performance would be Doom Eternal, as it uses Vulkan both on Linux and on Windows. In case of Shadow of the Tomb Raider Feral Interactive ported it from DirectX to Vulkan, which could result in a worse performance.
The NVIDIA drivers on Windows also obviously get a lot more attention. I would be very interested in seeing a comparison while using a modern AMD GPU.
Also rdr 2, there are probably more games with vulkan on windows. Oh and a fun fact, vulkan can run more stable on w10 with vulkan than dx. And that was on my 1050ti gpu a few years back. Vulkan was much smoother than dx for some reason.
Appreciated your work❤
Excellent video Nick!
I have two similar systems, fairly high performance.
Windows at work and Manjaro at home.
With similar startup apps on both machines i can get 3.5GB used on Manjaro and 6.5GB used on Windows 11, after logging in.
Manjaro can also boot up in about 4 seconds, whereas Windows 11 takes about 15.
That's your problem you're using manjaro. Manjaro is super bloated.
@Ransacked did you even read the comment??
Troll
@@keepanopenmindlookatallthe2540 lol here comes the manjaro fanboys. I'm not trolling it's the truth. Manjaro is the ubuntu of arch based distros. Bloated.
@@ransacked yes but you said "that's the problem", except there is no problem, Manjaro beat windows in everything.
I also think Manjaro is bloated and not a great distro but there is no reason to mention it now, not like this
If anything you should say "if Manjaro can do this imagine something better like Arch"
@ZeroKun265 lol ok that's still his problem and your comment doesn't change my mind.
I don't know the type of testing you will do yet.
But before I start watching, I'll say for CPU intensive Multi threaded tasks, Linux Wins.
Graphics intensive tasks, Windows.
Since Linux is still targeted at Servers.
Edit: I wasn't far off..
I was surprised… Windows seems slightly better at CPU multi threading
@@TheLinuxEXP Same. Last time I heard that Windows had a hard time dealing with more than 16 threads. Maybe the difference wouldn't be there if the core count scales.
But then that would be a server / workstation comparison.
Great video. Thanks.
I made popcorn to view this comparison.
Tnx for time and video.
Great
no you all no first
Thank You for this research, just started using windows 11 and i can see they made Streaming Privacy a concern with some of the changes they made with file seaching and folder viewing
Thanks! Particularly for the egpu wayland script
I didn't know keeps old files in the background. I always wondered why after using a computer for the longest time that no matter what I deleted, I could never go back to the same space as I started on initial install
We want part 2 in comparison with professional apps like compiling a code or performance during video editing etc.
I think the reason of huge upload speed difference was firewall settings on ubuntu, if it was selected as low then could be roughly same
Awesome comparison 👍🙂
It could be awesome to see the comparison with KDE and PopOS vs Windows 🙂
Which disto are you using with KDE? 🙂
HZD does have a benchmark in the settings, look at the bottom for the keybind to press to start it.
Really? How did I miss that??
I have actually been using WSL2 recently rather than Dual Booting after I had to format my PC after it into RMA. To be fair, the current Linux experience with my hardware is still subpar. From what I've tested the Linux 6.1 kernel has improved a few things (I did a Fedora 37 test install just for fun) and I will probably install Ubuntu 23.04 once it comes out.
One of the reasons I still prefer Ubuntu over Fedora is that is has fractional scaling on X11 and with NVIDIA hybrid graphics X11 still seems to work better. I also like that it comes with a Dock by default, whereas with Fedora I always risk that Dash do Dock still does not support the latest GNOME version (e.g., right now it seems that it is not compatible with the latest GNOME beta).
when doing these types of comparisons. would it be possible to put comparison numbers side by side? bit of a pain having to flip back and forth through the video
I would guess that the fans running loud for Windows may mean that the processors (CPU, GPU, etc) were being more fully utilized, or the fan curves were affected, or other things were running in the background, or any combination of these things. Maybe something I haven't thought of as well.
Interesting watch, thanks for making the video :)
Thanks for the video!
Nice video and was surprised by the battery life.
I wouldn't complain about the distros to much, my interest is more with the Kernel version I wounder how much performance there is between something like 4.x.x kernel and a 6.x.x kernel.
Really liked the video and of course you weren't able to do all of these tests on 10 distinct systems by yourself, but I would guess that they would all fall sort of in the same ball park. Windows with a slight edge in terms of (gaming) performance, and as long as that's the case: I'll stay with fedora which is an actual pleasure to use (used Windows since Win'98 up until a year ago). If Windows can up the performance of my system by 2 or 3 times, I might reconsider, but I don't think I'd ever want to go back and seeing how Asahi Linux even manages to squeeze better performance out of Apple's ARM chips than MacOS does, I'm fairly confident that we can have the best experience AND performance on any platform, as long as we invest the time in it/support the amazing developers who make all of this happen,
Thanks for video! As usually, very interesting. However, I have some feedback.
First, it was hard to follow the numbers in the head while you were reading them out and comparing them myself. I think it would be better if you first show the common settings of the particular software on any system (either Windows or Linux system), and when the actual benchmark results are discussed splitting the screen such that e.g. left=Linux, right=Windows with the respective numbers. This would make it much easier to follow, since there are a lot of numbers in the video.
Also, the title of the video might be a bit misleading. As you're testing Ubuntu vs Windows here to be precise. I agree that for most benchmarks it doesn't make much difference, but for example for memory usage and especially storage usage it's an important point. You also mention it earlier in the video, but adapting the title might still make it more clear for inexperienced users.
Thanks again!
You can always go back in the video and write things down.
Thanks for the video 👍
Now that you compare one OS to another, maybe compare Fedora to Nobara (GloriousEggroll's Fedora fork customized for gaming)? Would be interesting how significant is custom patched kernel and other improvements that come with it, aside from convenience.
13:09 Horizon on Linux runs on VKD3D, for DX12 to Vulkan convertion.
DXVK is DX9-10-11 to Vulkan
Ok this was really interesting. I worked on the OpenGL driver for AMD for years and it just went through a massive revamp (for windows only, maybe? Actually I will look into that) in the last year. I would like to see a Vulkan head to head perf test. Talos Principle has a Vulkan and DX benchmark built in. It previously had an OpenGL benchmark as well but that was discontinued.
The internet speed blows my mind. It's the same hardware. This shouldnt happen. It would have to be a base linux driver issue, I think? This is super weird to me.
Otherwise, these numbers were really cool to see and VERY surprising. I do love that you went with windows 11 instead of 10.
My main reason for starting the migration from windows is the new adware bloat they are forcing onto the OS.
Thanks Nick!
I really like your videos.
Interesting result on speed test, though I get 997.89 down and 999.75 up. It varied by just a few megabits but this was the best result for me at least. On the geekbench I remember I got 5-6% faster on single core on Manjaro while multicore was just about a few points higher than in windows 11 very debloated.
there is also a program called tlp which increases battery life on linux laptops, it would be interesting to see a comparison with tlp installed and without
Which is the distribution that you show at 3:16 ?
Networking can have different reasons. It could be that your ISP provided you with less speed by the time you were testing your Windows installation. It could also be that the network driver is way better on Ubuntu, but it might be hardware dependent.
did the laptop come with ubuntu preinstalled or did you installed it on it by youtself? because who knows if maybe system76 made some optimizations or put some drivers on the oem installation
I installed it myself
Overall - and mostly subjectively - I'd say your results pretty much agree with my experience.
I have both Linux and WIndoze machines for work. The Windows machine has much more horsepower, but the Linux box uses a fraction (and I use that word advisedly) of the resources. Subjectively the performance is about the same. I run Manjaro with OpenBox, a status bar, and a lot optimization. It idles at a comparatively paltry 350MB of RAM, while Windows consumes about 4GB at startup. Something that helps greatly on the Linux box is a fixed-size swap file of 8GB. I tried it and the performance was noticeably better. I also run WSL on Windows with Arch ("I use Arch, BTW...") for light maintenance work on our servers, and - with considerable tweaking - it works OK.
I also have a suggestion for Windows users - Atlas OS (I'm not affiliated in any way with them), which is a very debloated version of Windows 10 aimed at gamers. It's perfectly legal, not cracked or hacked (you need to activate it with a legit license key), the installer having been optimized using Microsoft's own tooling to install the absolute minimum that can still do pretty much everything you would need. I run it as a VM on my puny Linux box and it runs great.
4:40 I think, the problem of Ubuntu is the snap package: It retrieves the complete dependencies which uses much disk space.
For the graphics tests: there is clearly driver related differences. The difference in fan curves is the best proof to this. Most of (practically all) these days CPUs and GPUs are thermally limited and controlled. If your silicon is cooler, the clock speeds will be higher, ending in higher performance. That 7 °C higher temperature on its own can resulted the lower gaming performance under linux, because your GPU ran at lower clock speeds.
For me the fact that in Manjaro I get regular updates to the entire system that last around 5 minutes with reboot is enough to never look at the Windows update screen again.
The Portmaster URL in the AUR package is broken gets 404 not found
What the hell, the D/U speeds. I just did a test on my system (running Tumbleweed) and got 800 down/900 up
and man...the Horizon test makes me realize how amazing Proton is, damn.
Battery life can change dramatically between different Linux distros. I didn't actually think it would be that good in Ubuntu. I've also noticed differences in speed with FTP uploads to my server between Linux distros
Great video Thank you
It would be interesting to see the ethernet up/down to a device local to your LAN and remove the complication of your ISP, router etc... Maybe fio or a different test too. Seems like there must be something odd going on there.
Not saying it's not valid, it's just an interesting quirk.
Thank you very much! Great stuff! I think you should have run some performance tests while off the charger for the sake of completeness. Windows and Ubuntu might use very different power profiles with Ubuntu (maybe) throttling the CPU performance on battery more aggressively.
As for the multicore Geekbench score. Core 12th Gen CPUs have those different efficiency and performance cores. Windows 11 might be better at supporting this heterogeneous setup.