You are not logged in.
Hello everyone!
I am facing a very annoying problem and I had no success finding any solution on the internet.
After installing Arch Linux on my machine I noticed some strange stutter on practically everything: Media player, dragging windows, video playback etc. It can be particularly noticed during camera panning in videos. Furthermore, both Firefox and Chromium fail this test: https://www.vsynctester.com/
I have to point out though, that I do not experience any lags – the moving part of the screen just jitters/stutters during scrolling, camera panning etc. Nevertheless, the gaming performance is absolutely normal and funny enough – I do not encounter any jitter/stutter whatsoever when I play games ???
Testing/attempts done so far:
• changing DE; during tests, all of the following DEs showed the jitter/stutter issue. I could get best results with xfce and compton, but still the jitter was very annoying
◦ KDE Plasma
◦ GNOME
◦ MATE
◦ cinnamon
◦ xfce
• changing distribution; sadly, even on wayland the artifacts occurred
◦ Ubuntu
◦ Fedora 28 workstation (wayland)
◦ Linux Mint
• changing graphics cards; still no improvement
◦ AMD Radeon RX580 (Driver: amdgpu)
◦ Nvidia GT 1030 (Driver: nvidia 390.87)
◦ Intel HD630 (Driver: modeset and i915)
• I also tried several X.org configurations such as enabling and disabling „TearFree“
• changing the monitor with lower resolution ones – still no improvement
• on my notebook (HP ProBook 450 G2) everything works flawlessly with all the listed DEs above and also when I plug the HiDPI monitor in.
• tested games so far:
◦ Metro 2033 Redux
◦ Metro: Last Light Redux
◦ HITMAN GOTY
◦ Rise of the Tomb Raider
◦ RimWorld
Technical info:
• Hardware:
◦ CPU: Intel Core i7-7700k (Stock Speed)
◦ Mainboard: ASRock Fatal1ty Z270 Gaming K6 (UEFI is up-to-date)
◦ RAM: Patriot Viper 4 DIMM Kit 16GB, DDR4-3200, CL16-18-18-36 %(PV416G320C6K)
◦ SSD: Samsung SSD 970 EVO 500GB, M.2 (MZ-V7E500BW)
◦ GPU: Sapphire Nitro+ Radeon RX 580 8GD5 Special Edition, 8GB GDDR5 Generic, DVI, 2x HDMI, 2x DP, lite retail (11265-21-20G)
◦ Monitor: Dell UltraSharp U2715H, 27" (210-ADSN/210-ADSO/210-ADSZ/210-ADZM)
◦ Keyboard: Roccat Ryos MK Pro
◦ Mouse: Roccat Tyon Laser Gaming Mouse
• Log files/configurations:
◦ dmesg: https://ptpb.pw/gF5c
◦ /var/log/Xorg.0.log: https://ptpb.pw/_5wS
◦ journalctl: https://ptpb.pw/xUBl
◦ compton.conf: https://ptpb.pw/JqsM
EDIT: Installed mesa-git, still no improvement though.
Any ideas/suggestions?
Last edited by atmega9 (2018-09-04 14:55:20)
Offline
Interesting test; though I think it may also be dependent on network quality; I'm not sure.
Your system has enough muscle that you should not be expecting issues.
Just stabbing in the dark here, but first, check the output of a tool such as htop. Check for high CPU usage.
Also, check that you are not getting page faults and having your system being bogged down swapping to disk.
I note that your compton.conf file is not based upon Arch's . Arch's has some interesting comments about glx-no-rebind-pixmap that may be relevant.
Have you tried (as an experiment) a simple (non-bloated) DE like openbox or i3wm ?
Are you sending frames to the monitor at the natural frame rate of the LCD? Last time I checked, LCDs panels are designed to run at one frame rate; period. Everything else has to eventually be converted to that rate. If, for example, you are sending the monitor 75Hz frame rate, but the glass runs naively at 60Hz, the monitor has to do the conversion, eventually having to drop frames. This happens downstream of where the compositor can help.
Last edited by ewaller (2018-09-03 14:54:08)
Nothing is too wonderful to be true, if it be consistent with the laws of nature -- Michael Faraday
The shortest way to ruin a country is to give power to demagogues.— Dionysius of Halicarnassus
---
How to Ask Questions the Smart Way
Online
Thanks for your reply!
I did some further testing:
My network connection seems to work without any issues (tested with speedtest-cli):
Download: 18.04 Mbit/s
Upload: 5.40 Mbit/s
During video playback on youtube (1440p 60HZ) some CPU cores spike at 50% but most of the time they sit between 10% and 30%. This is, as far as I can tell completely normal.
During video playback the system occupies 1.7GiB of RAM – swap stays completely empty. The firefox process loads the CPU from 8% to well over 100%, mostly it sits between 80% and 90% though. As far as I know firefox is multi-threaded, so this should be normal? Testings with chromium: practically identical. Chromium seems to do more load-balancing on the cores but this could also be my imagination.
Concerning openbox/i3: No, I did not had time yet to test them as they are not really an option for me. But I will definitely try them out just to see whether it works or not.
Frames get sent with 60HZ, which is exactly the refresh rate of the monitor. In all the DEs I tested, I set the refresh rate from “Automatic” to “60HZ”
I have a more recent Notebook (Acer TravelMate B117-M) with a Pentium N3710 and Intel HD405. I will also try an installation on that machine.
EDIT: Installed openbox and tested; Sadly, the jitter issue persists. Firefox remains completely the same. Chromium does better than on a "normal" DE but if you move the mouse the framerate drops significantly.
Last edited by atmega9 (2018-09-03 20:01:42)
Offline
I did another test with different hardware:
Mainboard: ASRock B150M Pro4V (90-MXB2W0-A0UAYZ)
CPU: Intel Core i3-6100T (Intel HD 530)
Nevertheless, everything stayed the same. On the Acer TravelMate I mentioned above, everything is totally fine. Any ideas where the differences could be? I am starting to think that this is a bug as several hard- and software configurations do not change anything whatsoever... But where and how should I submit a bug report?
Offline
Hi,
I also have this annoying issue with short stutters and discovered your thread after my Arch update today. In my case at least, I have isolated the problem to the kernel. For my hardware, kernels <=4.2.5-1 work fine; newer kernels exhibit this issue. I even submitted a bug to the kernel maintainers at one point and it was purportedly fixed, but I haven't noticed any improvement.
I've been meaning to have a go at bisecting the kernel again to try and isolate the bug, but if you have the inclination and time you're very welcome to do it and share the results.
Offline
Hi, frukt,
good to hear that I am not the only one facing this issue. Since my initial post, I did not stop researching and found out a few new things:
Currently, I am using KDE with the amdgpu driver. If I enable compositing and use any sort of OpenGL, everything stutters as described above and also as you mentioned it. Nevertheless, the stutter gets better as soon as I switch to Xrender or completely disable compositing. Furthermore, windows, icons and everything else lags behind the cursor when dragging. This gets fixed as well by disabling OpenGL. The same applies to the compton backend section.
Wayland / X11 seems to have nothing to do with this issue. I tried a Wayland Session on Plasma – couldn’t even tell the difference.
The first Distribution I used was ubuntu 14.04 which used the 3.13 kernel. I can’t remember any problems at all.
I found quite a few bugs for pretty much every DE. Users reported stutter during scrolling, dragging etc. Apparently, the only thing that is the same among all machines is the kernel, which supports your theory of the kernel being the problem.
If that wasn’t enough, I also seem to be a victim of this bug: https://bugs.freedesktop.org/show_bug.cgi?id=106175. Kernel 4.21 will get some fixes in the DRM. In the meantime, I disable PageFlip for amdgpu I guess…
Sadly, I do not have the time to dig around in the kernel. Have you tried a BSD distribution already? Do you remember the bug report you submitted? It would be a shame if we can’t get this fixed.
Offline
This stutter is not there for me in XFCE's xfwm4 and its compositing. This is with an RX480 and amdgpu. The only special thing I've configured is I use amdgpu.dc=0 on the kernel command line.
In any other compositor I tried, there was stutter in that Firefox vsynctester.com website experiment. The next best thing to xfwm4's compositing I could find is compton started like this:
compton --config /dev/null --backend glx --vsync opengl --sw-optiThis "--sw-opti" option is not supposed to help according to the documentation, but it does.
This compton setup is not perfect, the stutter shows up in Firefox when moving a window around. Meanwhile with xfwm4's compositing things really seem to run absolutely perfect. The color of that "VSYNC" text on the website stays gray even when for example moving a terminal window around.
Xfwm4 is also really good about latency, when moving a window, the window stays super close to the hardware mouse pointer's position.
A problem I found with xfmw4's compositing is, it seems not all games might like it. I noticed in Civilization 6, there's tearing. It's not there when disabling compositing.
Offline
The purported fix was specifically in the i915 driver and I haven't tested if the issue appears on anything else than Intel graphics hardware. My workaround thus far has been staying on kernel 4.2 (even though this will become more difficult, for example during the latest update I had to substitute the vanilla qt-base package for a custom one that wouldn't use the statx syscall (only present in kernels >=4.11)). Bisecting - recompiling - testing the kernel to isolate the bug isn't all that difficult and once this issue becomes too painful I'll try to get around to it. No need to muck around in the internals.
Speaking of compositors there's no ideal one. I generally use the xfwm4 compositor at work (plays much better with GTK2, for example) and compton at home (better for video and gaming in my experience). I've whipped up a script to switch between the two.
Offline