X high load and performance issues

Forum / NoMachine for Linux / X high load and performance issues

Viewing 15 posts - 1 through 15 (of 18 total)
  • Author
    Posts
  • #11421
    sspj
    Participant

    Hello

    We have installed the NoMachine in our environment but we have a lot of performance problems. Please, could you tell me how I can fix them? Here you find some details:

    – The OS is SLES11 SP3 64 bits with GNOME

    – We are using NX, port 4000, and we opened the ports 4000 [tcp] and 4011-4999 [tcp & udp]

    – Version of NoMachine: free version, 5.1.24

    Problems observed:

    Response to mouse clicks and typing has a noticeable delay. For example: type “xclock” and the delay to the screen is 3 to 4 characters behind typing speed. Over ssh (outside of NoMachine) the response time is the same as you would expected, if you were sitting in front of the machine.

    When I connect using NoMachine, X using 100% CPU – disconnect NoMachine and the CPU returns to nominal.
    – If I reduce the Display resolution to 640 x 480, the CPU usage drops to 48% – and response time to mouse clicks and keypresses is no difference to accessing via ssh. It seems that X server CPU usage is proportional to display resolution when using NoMachine – anything over 1024 x 768 pushes the CPU usage above 98%.

    Thanks in advance

    #11427
    fra81
    Moderator

    Hello.

    This problem has been observed in the past with specific video drivers. See: https://www.nomachine.com/forums/topic/xorg-high-cpu-utilization-matrox-bug.

    Can you tell us more about your graphics card and installed drivers?

    #11431
    sspj
    Participant

    Hello!

    Thanks for reply.

    We have the same problem in virtual machines with default OS drivers, in workstation with nvidia graphic card and nvidia drivers, and in servers with ATI graphic card and default OS drivers.

    Thanks

    #11583
    fra81
    Moderator

    Hi.

    We couldn’t reproduce on any harrdware we tested on. Could you tell us something more specific about the cards and the drivers?

    As far as I understand, you are observing these performance problems running NoMachine within a VM. If so, which VM type, software vendor and specific version?

    #11586
    sspj
    Participant

    Hi

    We have the same performance problems in three different environment:

    – In a workstation, with nvidia K600 graphic card and driver NVIDIA-Linux-x86_64-361.42

    – In HP DL380 server, without any specific driver

    – In a VM running with vmware player with vmware tools installed

    All the environments has the same OS: SLES11 SP3 64 bits with GNOME Desktop

    Thanks in advance

    #11687
    fra81
    Moderator

    Hi,

    we tested again SLES11 in our labs without observing any performance problem. Problems with drivers could be expected though on the HP or in the vmware machines, but it is strange you have performance problems in the workstation with nvidia’s proprietary drivers. Are you sure the symptoms are the same on all machines, with X using a lot of CPU?

    For further investigations, please send us the following info from the workstation with the nvidia card:

    – the ‘/etc/X11/xorg.conf’ file

    – the output of the ‘lspci -d 10de: -v’ command

    From the other machines, the output of:

    – ‘lshw -C display’ command

    #11721
    sspj
    Participant

    Hi

    I’ve attached the requested logs for

    – A VM

    – A server

    – A workstation with nvidia card (and nvidia drivers installed)

    All are running SLES11 with NoMachine 5.1.24, and all has the same performance problem.

     

    Thanks a lot for your support

    #11724
    Bilbotine
    Participant

    Hello sspj,

    There’s an upload error, can you forward the logs to forum[at]nomachine[dot]com using ‘ X high load and performance issues’ as subject ?

    Thank you

    #11725
    sspj
    Participant

    Done!

    #11765
    sspj
    Participant

    Hi

    Any news with this issue? Did you find anything that could help us? thanks

    #11773
    graywolf
    Participant

    Hello, high CPU usage looks a problem with Matrox video cards. This could be issue with VM and server. Try to reconfigure Xorg (edit /etc/X11/xorg.conf) in order to use generic “vesa” driver in place of “mga” (Matrox) driver. Reboot after changes.

     

    Things are different for the workstation with nVidia card. No performance issue is know for that case. I see the workstation has two nVidia cards and each card has two video out. So that up to four monitors can be attached to the workstation. Can you tell me if that is your case? Testing with only one card (and one monitor) is possible for you?

     

    • This reply was modified 8 years, 5 months ago by Britgirl.
    #11798
    sspj
    Participant

    Hello

    Thanks for reply. I’ll check with “vesa” driver and I’ll write you back with the test results.

    For the workstation, yes, we have two nvidia cards with four monitors attached to it. I’ll try to test the performance with one monitor.

    Thanks!

    #11949
    sspj
    Participant

    Hello

    We have 3 monitors attached for workstations with 2 nVidia cards.

    There appears to be no issue when using NoMachine to access a workstations with one card and monitor

    Thanks

    #12259
    sspj
    Participant

    Hi

    Any news with this issue? Did you find anything that could help us? thanks

    #13326
    graywolf
    Participant

    With some configurations the video device doesn’t provide a GPU-accelerated screen capture, so that you could get spikes in the CPU usage.

    It’s also because of that we are always in search of different (and more efficient) ways for capturing the screen.

     

Viewing 15 posts - 1 through 15 (of 18 total)

This topic was marked as solved, you can't post.