FreeBSD Screen Shots

[Voltaire]

Honestly most of these window managers are so "bare bones" that XCB vs Xlib barely matters. At most they send ~20 messages to and from the Xserver during even the most invasive recapture.

OpenArena benchmark results between dmw and bspwm

dwm 204.4 - 203.4 - 203.7 - 203.2 - 203.7 - 203.8
bspwm 205.0 - 204.2 - 205.0 - 203.7 - 204.2 - 204.4

I don't want to calculate the average values. In both Xonotic and OpenArena dwm is on average (slightly) slower than bspwm.

In OpenGL games the difference is (usually) minimal.
But in the browser, the difference can be more than 5% in some situations.

https://projects.mini-dweeb.org/attachments/download/4/report.pdf

=> XCB wrong u s a g e ( s y n c h r o n o u s ) D u r at i o n : 0. 0 3 8 6 0 8 s
=> XCB bad u s a g e ( a s y n c h r o n o u s ) D u r at i o n : 0. 0 0 2 6 9 1 s R a t i o : 1 4 . 3 5 7
=> Xli b t r a d i t i o n a l u s a g e D u r at i o n : 0. 0 4 3 4 8 7 s R a t i o : 1 6 . 1 6

Just to be clear, dwm is already a very fast WM. dwm is often >4% faster than awesomeWM and i3 in my tests, because it's written entirely in C.

 

[MasterOne]

Voltaire please feel free to show off more pictures of your bspwm setup, I'd also be interested in which apps you are using and how they look (GTK/Qt?).

 

[kpedersen]

In OpenGL games the difference is (usually) minimal.
But in the browser, the difference can be more than 5% in some situations.

Interesting that you are finding such a big difference for browsers. I would have thought that the interaction with the WM would have been fairly minimal.

I wonder what X11 calls it is making that is resulting in a 5% difference.

 

[blackbird9]

Or we could look at this another way ;-)

Window Manager	Looks like Next(tm) Window Manager?
cwm	no
dwm	no
fvwm	maybe with some hacking! Afterstep??!
i3	no
bspwm	no
twm	no
spectrwm	no
openbox	yes
wmaker	yes

And we have some new winners

I like motif too, in a different incarnation I used to write GUI code on AIX using motif. That was a long, long time ago in a galaxy far away, though hahaha. Not sure if the hungry programmers are still in business

Thinking of programming drag and drop in motif still makes me feel queasy. Can GTK or Qt really be any worse?

 

[blackbird9]

I don't want to upset anyone but I never really got the whole tiling thing. I can appreciate some sysadm's and testers might like it for a long running setup, and it might be a good fit for the trading desk. But what's wrong with having a z-order anyway? If you set sloppy focus, aka focus follows mouse, etc, then you avoid the tedious click to surface a window. Why restrict yourself to tiling? I can use screen or tmux if I want tiled terminals...

 

[kpedersen]

Thinking of programming drag and drop in motif still makes me feel queasy. Can GTK or Qt really be any worse?

I find DnD functionality is always messy. I think the lesstif guys never did implement it before Motif was GPLed. It isn't that elegant on Gtk or Qt either in my opinion. The nicer API I have seen is in Fltk but Motif drag and drop is actually not so bad, once you get past the segfaults

I can use screen or tmux if I want tiled terminals...

I tend to agree. These days, after starting a maximized xterm running tmux, I sometimes wonder if I even need a Window Manager anymore. Specifically the default unaliased/bitmap font on OpenBSD's kms/modesetting console (Spleen) is also nice enough to use without X11 entirely. So most days I just use that along with a raw framebuffer VNC client for web browsing.

In some ways, it also ensures that if the Wayland madness ever does kill everything we know and love, we can still just have a single maximized terminal and get on with our lives; regardless of the terrible workflows and visual tackiness, the compositor throws our direction.

 

[Voltaire]

Interesting that you are finding such a big difference for browsers. I would have thought that the interaction with the WM would have been fairly minimal.
I wonder what X11 calls it is making that is resulting in a 5% difference.

It is certainly not something that most people would expect, that in some situations a gap of 5.5% can still be observed with a super lightweight WM like dwm that is also written entirely in C.

I see the difference mainly in the Chalkboard HTML5 test where dwm is always significantly slower than bspwm (on average).

Today I found another browser test where dwm is also significantly slower on average.

bspwm

dwm test #1

dwm test #2

The performance difference here appears to be approximately 5.9%

 

[kpedersen]

The performance difference here appears to be approximately 5.9%

I don't know too much about these browser tests. Perhaps part of the test is to open up popup windows or programmatically resize the existing window? That could explain why there is interaction with the Window Manager.

I wonder what the performance would be like with a compositing manager (i.e xcompmgr) on both dwm and bspwm. But if it is graphical, I can only assume that dwm is creating windows with an unmatched i.e color depth and this is not perfectly optimal for the web browser display. Just guessing at this point!

 

[Voltaire]

I don't know too much about these browser tests. Perhaps part of the test is to open up popup windows or programmatically resize the existing window? That could explain why there is interaction with the Window Manager.

I wonder what the performance would be like with a compositing manager (i.e xcompmgr) on both dwm and bspwm. But if it is graphical, I can only assume that dwm is creating windows with an unmatched i.e color depth and this is not perfectly optimal for the web browser display. Just guessing at this point!

Chalkboard HTML5 tests scaling, panning and zooming.
Maybe this particular thing is something that dwm has more trouble with than bspwm.

For the BMark test, the results sometimes fluctuate quite a bit if you take the test many times in a row. I don't know how reliable that benchmark is.

You also have the following benchmark https://luic.github.io/WebGL-Performance-Benchmark/
Here I also see large fluctuations that I find difficult to explain, except by assuming that it is simply not an accurate test.

In MotionMark 1.2, results often fluctuate more than 4% in the same window manager, with nothing running in the background.

There may still be no demanding graphical browser benchmark capable of accurately measuring exact performance. It's strange that no one ever thought of putting time into it.

 

[Ogis]

Sorry, maybe this is off topic, but here are my test results. The test was performed using sway and with open source nvidia drivers.

 

[kpedersen]

Hmm, wayland or x11 wm aside, I would still assume that almost all of the stress test is on the browser which should be consistent per machine.

Strange idea; what score do you get running Xorg without *any* WM?

Ogis Do you have any other Wayland compositors (i.e Weston)? Do you notice different results between them?

 

[Alexander88207]

I guess we should move this topic to its own.

 

[kpedersen]

Yeah true; I completely overlooked that this was the almighty screenshots thread. Apologies for the (inquisitive) noise everyone!

 

[Minbari]

 

[Ogis]

Do you have any other Wayland compositors (i.e Weston)? Do you notice different results between them?

Hello. I have labwc. Same results.

 

[Voltaire]

Sorry, maybe this is off topic, but here are my test results. The test was performed using sway and with open source nvidia drivers.

You really throw away a lot of performance by using the open source Nvidia driver.
It's your choice, but I would always use the proprietary driver with your type of hardware.

 

[kpedersen]

You really throw away a lot of performance by using the open source Nvidia driver.

I found in the past that for older cards, the performance is very comparable. This is because Nouveau is actively developed and optimised. The old nvidia legacy drivers are stagnated and receive no maintenance.

Obviously for the newer hardware, the vendor drivers are likely faster (although in the past they have had issues with Wayland; but I think resolved now). Though many of us don't need performance. We just need pixels to display on screen!

 

[Voltaire]

I found in the past that for older cards, the performance is very comparable. This is because Nouveau is actively developed and optimised. The old nvidia legacy drivers are stagnated and receive no maintenance.
Obviously for the newer hardware, the vendor drivers are likely faster (although in the past they have had issues with Wayland; but I think resolved now). Though many of us don't need performance. We just need pixels to display on screen!

It might depend on which specific GPU you're using. I use a GPU of the Nvidia 600 generation (Kepler series) and I still see a reasonable gap there. It also makes sense because Nvidia itself does not really contribute to this open source driver. Three years and a few months ago, the open source driver was very slow compared to the proprietary driver.

The Open-Source NVIDIA/Nouveau vs. NVIDIA Linux Driver At The End Of 2019 - Poor But A Lot Of Hope - Phoronix

www.phoronix.com

 

[ThomasAdam]

If it has few developers and few users, then few people can notice security issues. I don't think FVWM2 is actively being developed anymore. Maybe in the past, but not anymore. I would use FVWM3 instead of FVWM2. But I wouldn't really trust FVWM3 because it has few developers and especially few users, so almost nobody is going to notice security problems in the code. Or is FVWM still the default WM in OpenBSD? What version of FVWM are they using? I haven't installed OpenBSD in a long time so I wouldn't know if FVWM is still default. I also don't know if that matters because the first thing 99% of OpenBSD users do is install another WM or desktop as soon as possible.

Oh dear. This is very subjective, immeasurable, and ultimately not true.

I understand if you don't want to use fvwm -- that's fine. But to say that fvwm3 has "few developers and especially few users, so almost nobody is going to notice security problems in the code" is bullshit.

 

[BaronBS]

 

[hbsd]

BobSlacker Extremely beautiful. But I noticed that you use red color a lot, especially for texts. Red color may be a little annoying for the eyes. I suggest that you use anti-radiation glasses to avoid damaging your eyes.
I fell in love with your desktop. Your file manager is really beautiful.

 

[BaronBS]

BobSlacker Extremely beautiful. But I noticed that you use red color a lot, especially for texts. Red color may be a little annoying for the eyes. I suggest that you use anti-radiation glasses to avoid damaging your eyes.
I fell in love with your desktop. Your file manager is really beautiful.

Thank you for the tip, but now I'm confused. I always heard that the blue light is the one that damage the eyes and the red light it is the safe one.
And IDK if it is placebo or not, but I can feel the change in my ability to focus when using redshift at night.

 

[Alain De Vos]

I think color is what pleases you most.
The bottom-bar is that cairo-dock ?

 

[BaronBS]

The bottom-bar is that cairo-dock ?

x11/plank

 

[hbsd]

Thank you for the tip, but now I'm confused. I always heard that the blue light is the one that damage the eyes and the red light it is the safe one.
And IDK if it is placebo or not, but I can feel the change in my ability to focus when using redshift at night.

No, I didn't mean to. As Alain De Vos said, color is a personal matter.
I just had this question, do you use red color even when reading a book or doing things like programming?
Honestly, I'd really like to see your Emacs theme!