Ubuntu comes with a couple of different flavors which are largely defined by the desktop environment that’s included in the each flavor, and by the default set of software applications included, to some extent.
While there are about nine flavors (official), targeting various kinds of end-users, I thought making a comparison of the three major flavors, specifically concerning their performance (since I’m more of a technically oriented individual, plus you can easily find their various new features on most other websites anyway), namely Ubuntu 15.10 (Unity desktop 7.3.2), Kubuntu 15.10 (KDE Plasma 5.4.0) and Ubuntu Gnome 15.10 (GNOME Shell 3.16) would come in handy for someone who’s trying to decide which to use.
For this review, I downloaded the 64-bit version of each flavor. The ISO disc image size of Ubuntu 15.10 is about 1.2 GB, Kubuntu 15.10 is slightly bigger, sizing around 1.4 GB and Ubuntu GNOME 15.10 is the slimmest of the bunch, featuring a 1.1 GB sized disc image. And they’re all based on the same Ubuntu core (kernel 4.2, X.org 1.17.2 etc).
Before I begin, below is a brief description of my hardware:
Intel Core i3-2330M CPU, Intel HD 3000 GPU, 4GB RAM (DDR3), Toshiba 7200 RPM (320GB) SATA HDD, Intel N-1030 Wireless adapter, Realtek network adapter ('RTL8168'), LED display with 1366x768 resolution (60Hz/60FPS). It's a Dell Vostro V-131 notebook.
To keep the readings accurate, I made sure to install each flavor into the same partition. For instance, I had previously reviewed Ubuntu 15.10 and I already had the data from it. So I securely wiped the data of that partition first, then installed Kubuntu 15.10 on it, ran the tests, & then securely wiped the partition again, before installing Ubuntu GNOME 15.10 and running the tests.
Also, except while measuring the power usage, I took five samples of the each test (Boot-up speed, Memory usage at desktop loading & Shutdown delay). The reason I didn’t run the power usage test five times was because I use a software application for that and it takes lots of samples and thus takes care of the accuracy automatically as a result. And usually I only run the System Responsiveness test once, though here I ran it twice (more on that later) on each flavor, except on Kubuntu 15.10.
So without any boring details, let’s get on with the number now shall we!.
I start this test when I hit the Enter key at the GRUB boot-loader’s menu, and stop measuring when the desktop is loaded (yes, with the user auto-login feature enabled). And, although some may disagree, when saying ‘desktop is loaded’, I don’t necessarily mean till all the start-up programs are loaded, because as soon as I see the background and other parts of the desktop are loaded and if I can actually start to use the desktop (open a program etc), even if one or two programs are still being loaded from the background, I stop measuring the time. And my reason for this is as follows.
There are some programs that are not necessarily needed to be fully loaded, before the user can begin his/her work. Take the network manager for instance. Sometimes, even after the desktop has loaded all the start-up programs, the network manager still can take about 3-4 seconds for finishing up connecting to a network (in my case, it’s the wireless router). And I think it’s unfair to the desktop to assume that it hasn’t fully loaded, just because the network manager is lagging behind because I can open the file manager or any other program that doesn’t need a working internet or a network connection.
But then again, since most of us primarily use our computers to connect to the internet, some can argue that till the network connectivity is present, the desktop is not actually fully loaded because the user cannot do his/her work. So I know that it’s somewhat a complex issue, but for now, I’ll stick with ignoring the network manager.
As you can easily observe, Kubuntu 15.10 was the slowest of the bunch (81.4% compared to Ubuntu! & 74.2 compared to Ubuntu GNOME!) and unsurprisingly, both Ubuntu 15.10 & Ubuntu GNOME 15.10 were quite close while booting up because despite the names of their desktops, Ubuntu 15.10 mostly rely on the applications of GNOME 3.16.
KDE (Kubuntu) on the other hand is a completely different desktop environment that has no relation to GNOME whatsoever and also loads a couple of applications when loading the desktop which should be partially held accounted for the delay as well.
I also noted that, from the GRUB menu, Kubuntu 15.10 took about 20-22 seconds before it started to run the display-manager (I know this because even though I had enabled user-auto-login, it’s only when the display manager is executed the desktop is started to load, and in KDE it is indicated by displaying a splash screen) and if you do a simple math, it means that KDE plasma 5.4.x roughly took about 30 seconds, just to load the desktop session. That’s about 58.2% of the entire boot-up time!.
Memory Usage Upon Desktop Loading…
Now unlike while carrying out the above test, here I make sure to wait till all the programs are fully loaded, before start measuring the memory usage.
As you can see, comparatively, Ubuntu GNOME 15.10 is the slimmest (about 4.4% compared to Ubuntu) and Kubuntu 15.10 used about 39% more memory compared to Ubuntu 15.10 and 45.5% compared to Ubuntu GNOME 15.10. But, as mentioned earlier, KDE does load a certain number of applications at the desktop start-up and I’m certain that by disabling some of these you can bring these memory usage readings down.
Now I wiped out its partition and installed Ubuntu GNOME 15.10 before it came into my mind, but this is not the first time I’ve reviewed KDE Plasma 5. For instance, earlier this year I reviewed KaOS 2015.02 and according to the data, it consumed about 442.6 MiB once the desktop was fully loaded which is quite closer to Ubuntu 15.10’s readings.
Power Usage at Idle…
When measuring this, I make sure my laptop’s Wi-Fi adapter is turned ON (connected to my wireless router), Bluetooth is turned OFF, brightness is set to its maximum value, and any other power related features that can intercept with the accuracy of the readings (such as screen dimming, screen-savers etc) are disabled. Then I let the OS to idle and the only running program is the one that’s measuring the power usage.
As you can see from above screenshots (as a proof), Ubuntu 15.10 was actually the better one where it consumed about 3.6% less compared to Kubuntu 15.10 and Ubuntu GNOME 15.10. These are not big numbers, but when spanned over a couple of hours (depending on your laptop batter’s capacity), they do make a difference on the overall battery life.
I however do not usually stop here. To see if things can be improved, I usually try installing a power usage optimizer (I prefer ‘TLP’) to see if it further improves the readings. So after dissatisfied with the readings (because I’ve seen other distributions doing far better), I installed TLP (
sudo apt-get install tlp) on all three occasions and below is the result:
As you can see, even here Ubuntu 15.10 was slightly better, but all in all, I was happy with the improvements (there is a .10 difference between Kubuntu 15.10 & Ubuntu GNOME 15.10, but I think that was caused by a GNOME’s power related notification).
CPU Usage at Idle…
In Ubuntu 15.10 when left to idle, for longer periods, no process interacted with the CPU and leave it alone, although the system monitor process itself kept consuming 2-3% of the CPU usage which is pretty common for it.
In Kubuntu 15.10 it was even more impressive (something that I’ve praised even in the past) where it scored 0% CPU usage activity when let to idle. The KDE’s system monitor pops in within every couple of seconds, but even that happens it only consumed about 1% CPU. This is by no means is a deal breaker, but it’s impressive.
As mentioned in the beginning, being based on the same set of applications, CPU usage at idle on Ubuntu GNOME 15.10 was pretty much the same. There too the system monitor kept consuming about 2-3% CPU, constantly (greedy bugger! 😀 ). Other processes did not interact with the CPU for long periods.
Hardware Recognition & ACPI…
Being based on the same Ubuntu core, all the three flavors were able to properly configure my hardware (except for the proprietary fingerprint reader for which there is no proper driver) and were also able to successfully restore the previously set brightness and the ON/OFF states of the network hardware (Wi-Fi, Bluetooth…) upon desktop loading.
But, in my Ubuntu 15.10 review I complained that on twice, while using Google Chrome and VirtualBox (both I had installed manually), X.org server crashed and that I was forcefully logged out of the desktop and lost all the work that I was previously doing before the crash.
I came across no such issue in Kubuntu 15.10, but, within a couple of minutes of logging into the desktop on Ubuntu GNOME 15.10 (pretty much every time), an error appeared saying that X.org server has crashed. Luckily though, unlike in Ubuntu 15.10, nothing really happened except for the error message. And I had not installed Google Chrome or VirtualBox when this happened on Ubuntu GNOME 15.10 either.
Other than that, it was all good.
In its essence, this test is very simple. I try to get a sense of the OS’s responsiveness by making the hard disk drive busy (I do this by copying a reasonably large file, about 1.5 GB, within two locations in the ‘Home’ folder) and then trying to play a multimedia file and while at the same time trying to open a couple of programs as well.
If most of the programs get opened up and the multimedia playback carries out, before the file finished copying, and while all this is happening, if the mouse pointer doesn’t loose its sensitivity by that much, I say that’s a responsive operating system.
To make a long story short. I carried out the test in Ubuntu 15.10 and at the first attempt it sucked!. But from my previous experiences I knew it’s probably due to the default I/O scheduler (that which manages read/write requests send to your hard disk), of Ubuntu 15.10, scaringly named ‘deadline’ 😉 . So I changed it back to a one that’s more fair, called ‘CFQ’, rebooted the computer, ran the test again and this time Ubuntu 15.10 rocked it!. Very responsive.
I was a little reluctant while trying it in Kubuntu 15.10 because in my past experience, KDE have had a major issue at being a responsive desktop (except once if I remember correctly). I used to freak out thinking that I was the only one facing this issue, but I’ve seen others complaining about it too. Some have even reported that the Plasma desktop with compositing is more responsive, though I don’t remember weather it was turned ON or OFF under the KDE distributions I have so for reviewed, because I was not aware of it until recently (I’ll keep that advice in mind for future reviews).
Anyhow, I carried out the same test under Kubuntu 15.10 and found that it was very responsive, just like Ubuntu 15.10!. Pretty much all the programs got opened before the file was fully copied and multimedia playback was though interrupted 2 or 3 times, they were short lived (happened in Ubuntu 15.10 too) and the mouse pointer sensitivity was excellent.
However, since Ubuntu 15.10 uses the ‘deadline’ scheduler I assumed Kubuntu must be using it too and wondered how it rendered such a great result. But when I checked, I found out that Kubuntu actually by default uses ‘CFQ’, not ‘deadline’.
Ubuntu GNOME 15.10…
Ubuntu GNOME suffered the same consequence as Ubuntu where it sucked under ‘deadline’. But then I changed it to ‘CFQ’, rebooted the computer and ran the test again, it had completely fixed the situation!.
So all in all, all the three flavors performed really well (though two weren’t, by default) as far as their responsiveness was concerned.
As you can see, both Ubuntu 15.10 & Ubuntu GNOME 15.10 were quite fast (and pretty much equals) while shutting down. But compared to them both, Kubuntu 15.10 was unfortunately 183.3% (roughly) slower to shutdown. Yuk!.
As you can see, both Ubuntu 15.10 and Ubuntu GNOME 15.10 were quite similar, performance-wise. Because again, they have a lot in common. While Kubuntu 15.10 comes with the KDE Plasma desktop, which uses a different set of technologies, and performance-wise, it didn’t perform that well, though it was very responsive. And, as a desktop, KDE offers a lot of features and crowns the user the king, and I have a great deal of respect for that. Also, if you’re somewhat experienced in GNU/Linux, then I think you should be able to fine-tune KDE to achieve better performance (I’ve done it in the past) as well. But when all is said and done, as far as the overall performance is concerned, Ubuntu 15.10 & Ubuntu GNOME 15.10 are the top contenders.