MATE MATE performs remarkably poorly in benchmarks

Here's a benchmark I did today that compares desktop and WM graphics performance in Motionmark 1.2:

Mate desktop: https://i.ibb.co/ggZDN02/Screenshot-at-2023-02-19-14-37-09.png
PeKWM window manager: https://i.ibb.co/pjLPHj8/2023-02-19-142216-1920x1080-scrot.png
dwm window manager: https://i.ibb.co/X3Sh1by/2023-02-19-142959-1920x1080-scrot.png

In Xonotic I see 8% performance difference between MATE and dwm. spectrwm and PeKWM were as fast as dwm in Xonotic. i3-wm and awesomeWM were in between dwm and MATE in terms of performance in Xonotic. And now I just ran glmark2 and I have achieved an even more striking result.

MATE desktop: https://i.ibb.co/LC50Wyj/Screenshot-at-2023-02-19-21-40-52.png
PeKWM window manager: https://i.ibb.co/Cb1h1gh/2023-02-19-212447-1920x1080-scrot.png
dwm window manager: https://i.ibb.co/cXXPJYF/2023-02-19-224605-1920x1080-scrot.png

On Nvidia you have no choice but to use a compositor, and as you can understand from the above text, dwm consistently scores highest everywhere and all the time. spectrwm is the closest to dwm in my experience, but almost always super narrowly slower. MATE scores very badly in MotionMark 1.2 and in glmark2 MATE is simply a total fiasco.

Those observations make me think about GhostBSD. MATE is not a bad desktop in terms of how the interface is structured. I used to be a big fan of Gnome 2 in Ubuntu 10.10 But the performance of MATE I see makes it unsuitable to offer as a default desktop in my opinion, it also doesn't give people a good idea regarding FreeBSD graphics performance when they start benchmarking the graphics in GhostBSD.
 
Benchmarking stuff is one thing, the other one to come to conclusions.

I strongly disagree with your conclusions about MATE DE, because what's the real life impact of those? None.
Many apps are indeed more CPU intensive than GPU intensive, so in the majority of cases it's not going to be a big deal. Something that has been true for developers for many years: if you want to be able to jump between desktop and web browsers and phones, you could stick to OpenGL ES 2.0 This technology is something that is very often used for graphics in games and apps. In glmark2, MATE doesn't score half of dwm, and this measures the performance of the popular OpenGL ES 2.0

If you're a laptop user, you're going to have a lot less battery life if you're going to be doing graphically intensive tasks on MATE than dwm. You have websites that are graphically intensive, there are many apps that are graphically intensive and games are also graphically intensive. Apart from the reduced battery life, you will also get much less performance in games. At least it seems to be around 8% performance difference in the best case scenario's for MATE. But when you look at GPUs, the difference between the lowest clocked GPU and the highest clocked GPU for the same model, there's often a $100 price difference between them. And the performance difference is usually around 8% between the two models.

It seems that MATE also shows this underperformance in Vulkan: https://www.reddit.com/r/vulkan/comments/gblf3g/catastrophic_desktop_performance_issues_with/
 
You are not getting my point, which comes from a different angle: most GPUs nowadays are so vastly overpowered when it comes to running a 2D desktop only, this includes chip sets graphics, that a few percent more or less in performance really doesn't matter for most.

And if battery life on a laptop really matters chances are high that you don't run FreeBSD nor Linux. Of course eliminating battery drains there is always a good thing, but people who care about that will most likely stick to a Macbook because neither FreeBSD or Linux are there where Apple is.
 
I don't think frame-rate is seriously related to window manager nor that "mate" will be much slower.
My GPU has 12GB of memory on board so the window manager can use a few...
Playing xonotic ?
 
You are not getting my point, which comes from a different angle: most GPUs nowadays are so vastly overpowered when it comes to running a 2D desktop only, this includes chip sets graphics, that a few percent more or less in performance really doesn't matter for most.
That is true for gamers. According to the steam survey, gamers do have fairly powerful hardware. In the top 14 most used GPUs, the GTX 1050 is the least powerful GPU in between, and it's twice as powerful as my current GPU. This GTX 1050 was my old GPU and I sold it when GPU prices were very high. I noticed that I often couldn't use the GPU's power due to RAM or CPU bottlenecks. I am making a new build, I already have 16GB DDR5 RAM @ 6000 Mhz. I'm waiting for the price of some specific CPU's to drop a bit, I'm thinking about the i5-13600K.

And if battery life on a laptop really matters chances are high that you don't run FreeBSD nor Linux. Of course eliminating battery drains there is always a good thing, but people who care about that will most likely stick to a Macbook because neither FreeBSD or Linux are there where Apple is.
This is mainly because Safari is more efficient than Chrome and other browsers on Apple hardware. Safari and a number of other core apps have been extremely optimized by Apple. But I wonder why FreeBSD would consume more than macOS. I also highly doubt that this would be the scenario in my case. Suppose I use dwm, and I deactivate Conky, the CPU is at 0% idle all the time. Suppose I take all lightweight apps, such as feh/viewnior for image viewer, claws-mail as email, Thunar as file manager, Emacs/Geany for text editing, zathura document viewer, mpv media player, etc. None of those apps really use any significant CPU when they run. These apps startup in a blink of an eye. How can FreeBSD draw more power than macOS, which has a hybrid kernel (less performance than monolithic) and has more bloatware and heavier apps?
I don't think frame-rate is seriously related to window manager nor that "mate" will be much slower.
My GPU has 12GB of memory on board so the window manager can use a few...
For most gamers your logic holds, gamers usually have powerful hardware. But overall my benchmarks are really relevant.
Integrated GPUs make for nearly 76% of all GPU sales.
Which integrated GPUs are much faster than my Nvidia GTX 650?

Playing xonotic ?

I was more of a Dota2 player in the years 2015 to 2017. But at some point Valve replaced the entire design team of Dota 2 with another team, and with every thing this new team did, I found the game to be in quality decrease. At some point I didn't like the game anymore.

Some other games I like are Nier: Automata, The Witcher 3, The Talos Principle, Tomb Raider, Portal 2, Dishonored, Kona, Hitman, A Plague Tale, Ghostwire: Tokyo, ALIEN: ISOLATION, Hellblade: Senua's Sacrifice , SEKIRO: SHADOWS DIE TWICE, Oblivion, Firewatch, SCP: SECRET FILES.

I also like racing games sometimes, and shooters sometimes too. Xonotic is something I enjoy playing and probably find better than most commercial games. I also like 0 A.D. and a number of other open source games. Lately I mostly follow YouTubers instead of gaming myself :)
 
For most gamers your logic holds, gamers usually have powerful hardware. But overall my benchmarks are really relevant
And here I disagree mostly, because even most chip set graphics is more than overpowered to draw a normal 2D desktop, so if e.g. normal frame rate would be around 200+ fps and 8% less due to some programming - who cares.
 
And here I disagree mostly, because even most chip set graphics is more than overpowered to draw a normal 2D desktop, so if e.g. normal frame rate would be around 200+ fps and 8% less due to some programming - who cares.
There are plenty of people who will find the 8% difference important. As far as I know Intel was more popular with gamers for years, not because it was 8% faster than AMD in games, but because it was +- 4% faster. The differences in MotionMark and glmark2 also seem to indicate that the differences can often exceed 8%, and I'm not talking about synthetic benchmarks. I will test this in more depth in the future, so that we get a broader picture of what the differences are in reality.
 
And here I disagree mostly, because even most chip set graphics is more than overpowered to draw a normal 2D desktop, so if e.g. normal frame rate would be around 200+ fps and 8% less due to some programming - who cares.
I found the time to do a first additional non-synthetic benchmark. This time 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. Below you can see the results. Compton is enabled in PeKWM and dwm and in these window managers I do not have any screen tearing, just like in all previous benchmarks.

MATE desktop: 10.6 seconds 321.8 fps 1.0/3.1/8.0/1.2ms
PeKWM window manager: 8.9 seconds 383.0 fps 1.0/2.6/7.0/0.9ms
dwm window manager: 8.7 seconds 388.8 fps 1.0/2.6/7.0/0.9ms

In percentage terms, we can say that dwm is 20.82% faster in OpenArena than MATE.
20.82% performance difference is more than the difference between the RX 570 4GB and the RX 580 8GB

I will test other games in the future as well, and I might also have to do a synthetic test from UNIGINE, experts see it as a good stress test.
 
How can FreeBSD draw more power than macOS, which has a hybrid kernel (less performance than monolithic) and has more bloatware and heavier apps?

Presumably because it has better power management. A few years ago I read some articles about people experimenting to see how far they could push down the power consumption on PC hardware. They were all testing under Windows, with various things disabled, because they couldn't get anywhere near the same performance with Linux (the BSDs were all worse). Things might have changed since then, but it doesn't look like it.
 
Presumably because it has better power management. A few years ago I read some articles about people experimenting to see how far they could push down the power consumption on PC hardware. They were all testing under Windows, with various things disabled, because they couldn't get anywhere near the same performance with Linux (the BSDs were all worse). Things might have changed since then, but it doesn't look like it.
The new MacBook Pro is the best Apple has right now, offering 20 hours between charges in many cases
A StarFighter running Linux gets 18 hours of battery life: https://be.starlabs.systems/pages/starfighter
So what hardworkingnewbie claimed earlier is simply incorrect, if you optimized Linux for the same M2 chip it would probably have longer battery life than the latest MacBook Pro with macOS.

Interesting gone install openarena. icewm awesome openbox
I am again very amazed by this game. If you set the graphics to 1920x1080 and set all other settings to maximum, it actually looks quite modern. Not every map is equally fancy, but there are nice maps if you unlock all content. What surprised me is the sound quality of OpenAL in this game. If you set it to high sound quality and set the in-game volume to maximum, it really is better sound than anything I've ever heard coming out of Linux and windows.
 
The new MacBook Pro is the best Apple has right now, offering 20 hours between charges in many cases
A StarFighter running Linux gets 18 hours of battery life: https://be.starlabs.systems/pages/starfighter
So what hardworkingnewbie claimed earlier is simply incorrect, if you optimized Linux for the same M2 chip it would probably have longer battery life than the latest MacBook Pro with macOS.
This is irrelevant to what you were saying earlier about quiescent power. It's a very different situation.
 
This is irrelevant to what you were saying earlier about quiescent power. It's a very different situation.
But no one has provided the slightest bit of evidence that macOS is more energy efficient than Linux or FreeBSD. I do believe that Safari is more efficient and better optimized than Chrome and Edge. But I also believe that macOS has a strong disadvantage in that it is a hybrid kernel that is less efficient than the Linux and FreeBSD kernel. macOS has a lot of bloatware installed by default and running in the background, more so than Linux and FreeBSD. So it would be very illogical for macOS to be especially energy efficient. The kernel and bloatware play a strong role in general power consumption.

As far as I know, Apple hardware has never been fully open, and a lot of Linux driver code for Apple hardware is 'reverse engineered'. That's why it's hard to compare the two systems fairly. However, I think that if you compare the performance per watt of apps running on a Linux desktop (with hardware well supported by Linux) to the performance per watt of the same app running on macOS, then macOS is going to get the short end of the stick. Have there ever been any such objective comparisons made?

I think so:
https://www.phoronix.com/review/macos1015-win10-ubuntu/2
 
Remove the battery & place an ampere-meter(tang) on the 220V cord ?
You can perfectly measure how much the laptop consumes, but what I've been trying to make clear all along is that it's not a fair comparison.

Apple's hardware is not open. Linux and FreeBSD can therefore not properly optimize for this hardware because they don't have the slightest documentation about certain components, while Apple knows the hardware perfectly and developed it themselves.
 
Remove the battery & place an ampere-meter(tang) on the 220V cord ?
What you also often see is that people then benchmark Ubuntu Linux, although Ubuntu has never been a leading distro in performance or efficiency.
A StarFighter running Ubuntu gets 18 hours of battery life.
But the small nuances are always important.

By replacing Ubuntu with Clear Linux you will be able to get 20 hours of battery life on this Linux laptop, or as much as the battery life of the most efficient Apple laptops.
Source: https://www.reddit.com/r/archlinux/comments/qsvl2p/clear_linux_kernel_is_more_power_efficient_while/
 
But no one has provided the slightest bit of evidence that macOS is more energy efficient than Linux or FreeBSD.

I don't really care. You asked how it was possible, I told you.

I also believe that macOS has a strong disadvantage in that it is a hybrid kernel that is less efficient than the Linux and FreeBSD kernel.

It's kind of the point of a hybrid kernel that that isn't so.
 
I don't really care. You asked how it was possible, I told you.
Good, but I always like to see source references. What studies show that macOS has better battery life on fully open hardware? Is macOS even installable on a system that only uses open source hardware?

Why does Apple go to such lengths to keep certain hardware components obscure to Linux and FreeBSD developers? Has Apple ever had an attitude of, we're going to give Linux and FreeBSD developers exactly the same knowledge about MacBook hardware as all the knowledge we currently have?

It's kind of the point of a hybrid kernel that that isn't so.
That's new info then, I thought all kernel experts have known all along that hybrid kernels are always slower than monolithic.
Furthermore, I also thought that there is evidence that this is still the case: https://www.reddit.com/r/linux/comments/tj12vw/hugo_runs_twice_as_fast_in_asahi_linux_than_macos/

Since I haven't learned anything about this matter yet I'm going to leave it at that, I'm not a laptop user for several reasons so it's not really something that concerns me.

One last thing I want to give you is that longer battery life is often at the expense of performance. What I mean exactly by that is that your Linux system can perfectly get better power efficiency, but then what you're going to see, and that's interesting, is that with this power efficient system you're not going to be able to get the same high performance as with the system that was less power efficient. Source: https://www.phoronix.com/review/apple-m2-december/2

If taking the geometric mean of all the benchmarks conducted then and now, Asahi Linux now is running 95% the speed of where it was when the support premiered this summer.
One of the possible explanations is more power management work happening that is leading to better power efficiency for the Apple Silicon on Linux, albeit at the cost of performance.
 
The future of some CPU's will be the same as the future of the HURD-kernel i.e. going nowhere.
I personally wonder if Raptor Lake that came out a few months ago is still Nehalem. It is strange that in such a competitive area as CPUs you base all your products on the same design for 14 years: https://linustechtips.com/topic/1114884-nehalem-is-10-years-oldand-intel-are-still-using-it/

You can say Intel is a sad company, but I think Apple is worse. Their customers are very proud of having 20 hours of battery life, the equivalent of a StarFighter running Clear Linux using a CPU architecture that is 14 years old. Is that really something Apple users should be proud of?

And shouldn't it also be said that their MacBook is so efficient just because it's slow as a snail?

Source:
We benchmarked the base M2 MacBook Air’s storage and (surprise!) it’s slow

Apple M1 Ultra’s 64-Core GPU Gets Smoked By NVIDIA GeForce RTX 3090 In Compute & Gaming Benchmarks
Despite Apple's claims, M1 Ultra GPU is not as powerful as RTX 3090.
With Geekbench 5 OpenCL benchmark, the NVIDIA GeForce RTX 3090 scores 215,034 points while the Apple M1 Ultra SOC only manages 83,121 points. That's a 2.6x gain for the GeForce graphics card on the desktop PC. Even if we used an Apple optimized API such as Metal, the RTX 3090 still ends up 2.1x faster.
 
I think apple&iphone is over-hyped due to alot of marketing.
PS: & note general someone running a freebsd-server does not care about cpu's in power-saving-mode.
 
I think apple&iphone is over-hyped due to alot of marketing.
PS: & note general someone running a freebsd-server does not care about cpu's in power-saving-mode.
My sister has a MacBook, the generation before the M1 chips. It had cost something like 1800 EUR. I can tell an anecdote about this MacBook.

I can compare this expensive MacBook perfectly with my FreeBSD desktop with the following hardware: Intel i3-3240 + 4GB RAM @1600MHZ single channel DDR3 + NVIDIA GTX 650 1GB + EVO 850 500GB

So that MacBook's default apps launch something like 2 to 3 times slower than the lightweight alternatives I use on the aforementioned cheap FreeBSD desktop. The super cheap FreeBSD desktop feels much snappier and more responsive.

Her MacBook's UI is irritating if you're used to XFCE, it feels stupid and I keep asking myself why, why is everything less intuitive and obvious in macOS. I have to do a lot more work and mouse clicks to achieve the exact same things.

Then I want to install a free photo viewer. I search for half an hour, but find nothing acceptable, after one hour of searching I find XnView MP, which is super annoying, but apparently the best free photoviewer for macOS.

Then I install RawTherapee on macOS. I know RawTherapee's algorithms are better than those of the expensive Lightroom program. So you can have more detail in your RAW results than what you can achieve with Lightroom, and RawTherapee is further than Lightroom in many other areas as well. I also know RawTherapee has never crashed on FreeBSD. But what I notice about macOS: it is very unstable and crashes all the time. Which I have never seen on FreeBSD.

What my sister told me last week: her macOS suddenly had a bad crash and she couldn't fix it herself. She had to give 1050 EUR for the repair. My FreeBSD desktop has never crashed.
 
Back
Top